Practical Uses of Artificial Intelligence in Medical Laboratories and Healthcare

Summary

Highlights

• Panelists discussed AI's role in enhancing laboratory processes and improving test accuracy. 🤖
• The conversation covered ethical considerations such as bias in AI algorithms and the need for transparent operations. 🤔
• Experts pointed out AI's potential to streamline data management and enhance pre-analytical workflows. 📈
• Discussion highlighted AI's role in potentially reducing healthcare costs and increasing patient accessibility. 💡
• The webinar addressed the training and workforce impact, emphasizing AI as a supportive tool rather than a replacement. 🧑‍🔬

Key Takeaways

• AI is revolutionizing healthcare by improving lab workflows and diagnostic accuracy. 🚀
• Experts stress the importance of AI as a tool to augment, not replace, human capabilities in labs. 🤝
• The integration of AI into healthcare poses ethical and regulatory challenges that need careful management. ⚖️
• AI can streamline pre-analytical processes, making data entry and validation more efficient. 📊
• Personalized medicine may become more accessible as AI helps decipher complex data patterns. 🔍

Overview

Chapters

• 00:00 - 00:30: Introduction of Guest Speakers The chapter introduces guest speakers who bring diverse perspectives, experience, and knowledge to the conversation. It highlights Abhi Bhosale, the co-founder and CEO of CrelioHealth. He has a strong background in engineering, particularly from pic pun, and is a computer science graduate who turned into a Healthcare Tech leader. With over a decade of experience, he has played a key role in growing CrelioHealth globally.
• 00:30 - 01:30: CrelioHealth and Lighthouse CIO Introduction CrelioHealth provides services to over 1,500 medical laboratories worldwide, handling millions of medical records daily. The chapter also introduces Vitalii Khvatkov, the CIO at Lighthouse, who brings over 20 years of experience in technology innovation, consulting, and managing large IT projects. His recent work focuses on designing and developing predictive analytical technologies that enhance lab payment processes using big data.
• 01:30 - 02:30: Jihoon Baek and Dendi Introduction The chapter introduces Jihoon Baek and Dendi, focusing on their work in analytics and computer intelligence to predict payment risk by insurance payers. This technology was also applied at Bio Reference Laboratory. Jihoon Baek works at Lighthouse on a product called RCM Spotlight utilizing this technology. Previously, Vitali served as CEO of a software company that developed cloud-based cancer diagnostic applications with deep learning analytics. Vitali also has experience managing large-scale consulting projects for Fortune 500 companies like Enture, specializing in similar fields.
• 02:30 - 03:30: Housekeeping and Webinar Recording Details This chapter introduces two key individuals involved in business process transformation and the development of specialized software solutions. Vital is highlighted for his extensive experience in implementing SAP systems and creating 3D analysis solutions for NASA, with academic qualifications including a Master's in Computer Engineering and an MBA from Rice University. Meanwhile, Jihoon Baek is recognized as the founder and CEO of Dendi, a company aiming to be the comprehensive operating system for Diagnostic Reference Labs. Jihoon is noted for his diverse background in software development.
• 03:30 - 04:30: Buffering and Q&A Details The chapter discusses the application of AI and machine learning in financial analytics within the healthcare sector. It highlights the contributions of Jiun, who started Dendi in 2020, and has since assisted 60 reference labs in enhancing their operations. A housekeeping note suggests closing bandwidth-heavy applications to minimize buffering and confirms that the webinar will be recorded and accessible without needing to request it.
• 04:30 - 05:30: About Lighthouse Lab Services Chapter Title: About Lighthouse Lab Services This section explains the logistics of a webinar hosted by Lighthouse Lab Services. Participants who registered for the webinar will automatically receive a link to it via email. Audience members are encouraged to submit questions at any time using the Q&A section on the right-hand side of their screen. The organizers will attempt to address some questions during the presentation if they are pertinent, and the rest towards the end.
• 05:30 - 06:30: Introduction to AI in Healthcare Discussion The chapter introduces Lighthouse Lab Services, a company involved in medical laboratory consulting and recruiting. It mentions the use of polls during the webinar to gather real-time feedback from participants. Lighthouse Lab Services is noted for building over 50 CLE labs and providing revenue cycle management services.
• 06:30 - 08:00: AI Affecting All Aspects of Laboratories The chapter discusses the role of AI in impacting various aspects of laboratories, particularly in supporting clear Labs across the country. It emphasizes the assistance provided by AI in validating laboratory-developed tests (LDTs), gaining market access, and resolving issues related to payment by commercial payers or network integration. The chapter encourages lab professionals to seek assistance if needed, indicating that AI can offer solutions or point towards one.
• 08:00 - 09:00: AI-Generated Questions on AI and Labs The chapter discusses the integration of AI in the healthcare sector, specifically in laboratory settings. The speaker notes that while discussions about AI are quite common, the topic's significance can't be overlooked. Therefore, the focus is on understanding the impact of AI on healthcare labs, with insights from experts possessing substantial laboratory knowledge.
• 09:00 - 10:30: AI Improving Accuracy of Lab Results The chapter discusses the impact of AI on the accuracy of lab results, indicating that its influence will be more extensive than initially expected. It suggests that AI will affect all aspects of a laboratory, not just the technology team. There is ongoing discourse on this topic, with major headlines often presenting polarized views. Some claim AI will revolutionize healthcare, while others argue it might have detrimental effects.
• 10:30 - 12:00: AI as an Assistant Quality Manager The chapter explores the role of AI as an assistant quality manager, delving into a variety of perspectives on how AI can impact different laboratory areas. The discussion kicks off with addressing several questions about AI's implications, specifically focusing on questions generated by the AI tool, ChatGPT.
• 12:00 - 13:30: Correlations and Regional Data in AI This chapter titled 'Correlations and Regional Data in AI' delves into the impact of AI on laboratories and healthcare. It begins with a discussion on AI's potential to enhance the accuracy of laboratory results, highlighting the importance of appropriate dialogue regarding AI in these sectors. The speaker introduces the topic by posing a question about AI's role in improving laboratory result accuracy, kicking off with expert insights from participant AB.
• 13:30 - 14:30: AI in Laboratory Workflows The chapter 'AI in Laboratory Workflows' discusses the impact of artificial intelligence on laboratory operations, particularly focusing on changes in lab workflows and reporting. It addresses how AI can influence the accuracy of medical records or test results and explores the extent of this impact. The narrative emphasizes improvements in these areas attributed to AI advancements.
• 14:30 - 17:00: Human-AI Interaction in Labs In the chapter titled 'Human-AI Interaction in Labs', the narrative explores the role of AI as an ever-present assistant quality manager within laboratory environments. The technology functions as a continuous layer of quality control, diligently monitoring instruments and correlating this data with lab results. Through this integration, the AI facilitates a deeper understanding of individual test accuracies and outcomes, ensuring a more efficient and error-reduced lab operation.
• 17:00 - 18:30: AI and AI-Generated Content The chapter discusses the role of AI in identifying abnormal or unexpected results within data, particularly in contexts where instruments or systems may not behave as anticipated. Artificial Intelligence can learn to recognize and identify instances when an instrument goes beyond its linear range, acting as a layer of error detection. This capability emphasizes AI's broader potential in enhancing system accuracy and reliability.
• 18:30 - 21:00: Pre-Analytical and Post-Analytical AI Applications The chapter discusses the role of AI in pre-analytical and post-analytical processes. AI can identify and understand human errors, such as typing mistakes, and detect when results are altered from what the instrument reported. The chapter explores how AI can assist in contextualizing and correlating test results to improve accuracy and reliability in these stages.
• 21:00 - 22:30: AI Technology Adoption and Challenges The chapter discusses the adoption and challenges associated with AI technology, particularly in the medical field. It highlights how AI can assist in identifying discrepancies in results, such as a metabolic panel showing a certain set of results that correlate differently with other tests. This technology can enhance systems or workflow solutions used by medical directors, suggesting the potential for significant improvements in medical diagnostics and decision-making. The emphasis is on AI's role in augmenting human capabilities rather than replacing them.
• 22:30 - 24:00: AI in Training and Workflow Optimization This chapter discusses the utilization of AI in enhancing training and optimizing workflow processes. It emphasizes the role of AI systems in assisting individuals to make better decisions. The excitement surrounding various use cases is evident, as there are numerous scenarios that the speakers are eager to explore and test. Additionally, it is noted that customer-driven requests are a significant aspect of addressing use cases in AI deployment.
• 24:00 - 25:00: Patient Impact and Personalized Medicine This chapter discusses the use of regional or specific geographical data in personalized medicine, particularly in analyzing microbiomes. It highlights how these localized results can contribute to understanding patient impacts more accurately. The focus is on using specific regional characteristics to enhance the precision of medical tests and treatments, acknowledging that certain tests might yield unusual results depending on the regional biological factors.
• 25:00 - 26:30: AI's Role in Data Utilization The chapter titled 'AI's Role in Data Utilization' explores the significant role that artificial intelligence plays in managing and analyzing large datasets. It highlights the human propensity for error and suggests that AI can serve as a safety net to catch anomalies or outliers in data processing. The chapter underscores the potential for AI to improve data interpretation and reliability, enhancing overall accuracy and trust in data-driven insights.
• 26:30 - 28:10: Regulation and Ethical Considerations This chapter focuses on the role of AI in improving laboratory workflows and touches on the importance of reference ranges and criteria for determining test results.
• 28:10 - 29:30: Pre-Analytical AI Applications in Depth Chapter Title: Pre-Analytical AI Applications in Depth This chapter discusses the potential of AI to redefine traditional scientific reference ranges by utilizing demographic and population data. It highlights the arbitrariness of current reference ranges and suggests that AI can prompt a reevaluation of these standards. AI's ability to interpret vast amounts of data may uncover new insights and raise important questions about existing scientific practices.
• 29:30 - 30:00: Q&A and Practical AI Applications The chapter titled 'Q&A and Practical AI Applications' discusses the influence of AI on lab work processes. It explores the notion that outcomes like test results being 'within range' or 'out of range' are often arbitrary. AI may prompt us to rethink these traditional approaches and improve the systems currently in place. Although the speaker isn't a clinician, they emphasize AI's potential to bring about a shift in how we've been operating in laboratory environments. This discussion underscores the questioning of conventional methods and encourages embracing AI-driven solutions.

Practical Uses of Artificial Intelligence in Medical Laboratories and Healthcare Transcription

• 00:00 - 00:30 all right so I'm going to go ahead and I'm  going to do some intros to some of the guest   speakers we have today we've got a great lineup  of people that bring uh different perspectives   and some great experience and knowledge uh to the  conversation so we have Abhi Bhosale, he's the   co-founder and CEO of CrelioHealth. He pioneered  streamlining new healthc care Diagnostics through   technology through his engineering Roots at  pic pun and computer science graduate turned   Healthcare Tech leader with over a decade of  experience growing CrelioHealth globally. Today,
• 00:30 - 01:00 CrelioHealth serves more than 1,500 medical labs  globally and transacts to million medical records   every day. We also have Vitali Khvatkov who is  our CIO here at Lighthouse and he is a technology   expert an entrepreneur a computer scientist with  over 20 years of experience managing Technology   Innovation Consulting large scale IT projects his  recent experience includes design and development   of predictive analytical technology for better  lab payments for RCM uh that utilizes big data
• 01:00 - 01:30 analytics and computer intelligence to predict  payment Risk by Insurance payers uh he did this   for Bio Reference Laboratory as well he has a  product here at Lighthouse called RCM Spotlight um   that utilizes that technology uh previously Vitali  served as a CEO of a software company developing   cloud-based cancer diagnostic applications with  deep learning analytics his prior experience also   includes managing large scale Consulting projects  for Fortune 500 companies like enture specializing
• 01:30 - 02:00 in business process transformation and  implementing of SAP systems as well as developing   3D analysis solutions for NASA space show  program uh vital received his Masters in Computer   Engineering and an MBA from Rice University. We  also have Jihoon Baek, he's the founder and CEO   of Dendi. Jihoon uh started the company with the  mission of being an all-in-one operating system   for Diagnostic reference Labs he has a diverse  background in software develop Vel M product
• 02:00 - 02:30 management applications of AI machine learning and  the financial analytics of healthcare sectors jiun   started dendi in 2020 and has uh since helped  60 reference labs in the country grow their   operations so thank you guys all for being part  of the conversation and being with us um a little   bit of housekeeping here to reduce your chances of  buffering please close any heavy bandwidth using   apps that you might have running or things in the  background uh most common question we get is will   this webinar be recorded can you get a copy of it  yes is the answer do you need to ask for it no uh
• 02:30 - 03:00 so don't worry about that it'll automatically  be mailed out to you through email if you've   registered for this webinar which you wouldn't  be here if you hadn't so you'll get a link don't   need to do anything um you don't need to wait till  the end to ask questions so if you have questions   we're going to be using the Q&A um section which  you'll find on the right hand column of your   screen you can uh push questions there and we will  try to field those towards the end of the webinar   we may uh May grab a few of them as we go as well  if they're relevant to what we're presenting at
• 03:00 - 03:30 the time we will run some polls during this  webinar we encourage you to participate that's   our feedback system to understand what uh you guys  think too so we'll push those out and you guys uh   have the ability to respond and real time and see  what the peers and other people on this webinar   think all right so Lighthouse Lab Services just  real quick plug for what we do because that's how   we keep paying the bills and keep the lights on we  are a medical laboratory Consulting and recruiting   company we do re revenue cycle management U we  build Clea laabs we build 50 plus CLE Labs every
• 03:30 - 04:00 year uh from scratch and uh we provide the medical  directors for a lot of clear Labs through the   country on a monthly basis we help validate ldts  um get Market access if you're having a hard time   getting paid um by commercial payers or getting a  network so if there's a issue you're up against in   a c clinical lab good chance we have a solution  or could point you towards one so keep us in   mind for that and we'll give you a chance to ask  for assistance if that's something you want to
• 04:00 - 04:30 connect with us about but we're going to jump in  today to the conversation around Ai and healthc   care this webinar is probably overdue um everybody  was doing talks on AI so I didn't think it was too   exciting but it's just uh too big to ignore as  well so I wanted to really try to focus it in   on uh how AI is going to affect Healthcare but in  the laboratory specific so all of our guests have   really deep lab knowledge today and so they're  going to be able to tie it back and um I think
• 04:30 - 05:00 it's going to be a little broader than maybe you  think you might think that it might affect your   technology team I think it's going to affect um  probably all all aspects of the laboratory in   your company so this is some of the conversation  that's happening out there um right now and as   you can see as you read through these headlines  from you know major major sources um day-to-day   you'll read a headline about how AI is going to be  a wonderful thing that's going to fix Healthcare   and the next headline will be how it's going  to destroy health care and um cost you your
• 05:00 - 05:30 job potentially so a lot of lot of different  viewpoints on it maybe all of them could be   correct you know there could be a little bit of  all these things so we're going to jump into that   today and here different perspectives on how AI is  going to affect different areas of the laboratory   so uh we're going to start with tackling some  questions I've got maybe 12 15 different questions   um these questions were all generated by uh chat  GPT so these are all AI generated questions when
• 05:30 - 06:00 I asked them how we should talk about uh AI  affecting laboratory and Healthcare this is   what uh AI came up with so I thought that would be  appropriate for the the day so we're gonna we're   going to go with that and we're gonna start with  our first question is how is AI gonna improve the   accuracy of laboratory results and I'm going to  direct that one to AB thanks Sean uh so thanks   for the introduction and so um you know yeah we  uh at K basically kind of Tinker with some of
• 06:00 - 06:30 these questions and uh think about how um how  is that going to change like lab workflows and   how is that going to change like reporting and a  decent part of that is actually uh can we impact   um accuracy of medical records or test results  and to what extent so you know um some of things   that we um we see really um getting better with  AI is I think first and foremost like it it it's
• 06:30 - 07:00 essentially a um like an assistant quality manager  always available for you so and um so it acts like   a layer that is always present for um checking qu  control of your instruments as well as correlating   that with uh lab results and and ultimately kind  of you get to understand if uh a particular test
• 07:00 - 07:30 result is um like abnormal or uh not expected to  be uh in the linear range of the instrument so if   the instrument kind of doesn't behave correctly  um the a systems can learn to identify that and   I think in overall it kind of extends to like a  greater goal of like being like a layer of error
• 07:30 - 08:00 Management in some sort so uh it acts like um like  it can it can understand um human errors it can   understand when results get sort of changed from  what the instrument kind of reported versus maybe   there's a typing mistake so these sort of things  kind of get uh you know you can train AI pretty   well um we also believe that it can help you  contextualize and correlate test results um so if
• 08:00 - 08:30 a if a metabolic panel is um having certain set of  results and there are other tests which would kind   of correlate very differently uh so the AI can  really kind of assist or suggest that there is a   discrepancy or there is a difference in these two  results I think there's a lot of um augmentation   that could be done on uh what medical directors  use systems or workflow Solutions like ours or you
• 08:30 - 09:00 know others uh to you know really augmented to a  level where the systems ultimately help them make   these decisions better and uh we see a tremendous  amount of um use cases that I think um uh we are   pretty excited to kind of go out and uh test out  um you also there's one other use case that that I   would like to talk about I think there are the  requests we get from some of our customers to
• 09:00 - 09:30 kind of use these results to add um correlations  uh based on Regional or specific geographical uh   results of patient so um if um for example how um  like you can you can actually look at microbiomes   in a particular region and then use that to see  if you know this particular test has a very odd
• 09:30 - 10:00 microbiome detector so um yeah yeah that's all  yeah so many exciting parts of that and to think   about being able to digest some of those massive  amounts of data and find some of that correlation   I think the safety net mentality also is something  that everybody could get behind that um you know   we are prone to making errors as humans and  if we can put a safety net behind that to to   catch those out liers that seems like something  everyone should be able to get get behind um AB
• 10:00 - 10:30 touched on workflows a little bit uh Jun I wanted  to kick over to you what are your thoughts on how   AI can help improve workflows in the laboratory  yeah uh yeah thanks for having me uh John and uh   I actually want to add to what AB was saying as  well with regards to test results um if you look   at things like reference ranges and uh look at  results in general what's determined as a positive   versus negative result um those things are all  determined with a lot of stats and a lot of
• 10:30 - 11:00 science uh and the background um but even things  like reference ranges I mean at the end of the day   and in terms of scientifically if you've chosen  for example um demographically you know valid um   numbers as as a kind of like guide posts um to  determine reference ranges but something like AI   um could really help redefine those things based  on population data that we have or or even uh   things that you know these things are arbitrary uh  an AI May kind of resurface uh some questions as
• 11:00 - 11:30 to how we determine a a test result is a certain  outcome uh and so if if you know out of range   within range like those are those are things  that are arbitrary at the end of the day and   so maybe we'll find something different uh with  the data that's coming out from AI but um might   it make us question why we do what we do a little  bit too yeah exactly exactly um in terms of how   AI can improve the lab work Flows In general um  so I'm not a clinician so you know I I just talk
• 11:30 - 12:00 touched a little bit about how they can improve  uh test resulting uh to some degree um but there's   actually a lot of uh I think loow hanging fruit  when it comes to especially the clinical lab um   pre analytical workflows um that's something  that I see a lot of value in for example one   of the biggest challenges for all clinical Labs U  especially independent reference Labs is getting   paid uh and so there's a lot of of uh tools that  you know blow hanging fruit that that can be
• 12:00 - 12:30 solved or or at least improved with AI workflows  uh for example pre accessioning uh you have   tools like that already exist for example uh like  optical character recognition OCR for scanning in   insurance cards requisition forms um things like  that uh you can do AI for demographics validation   um you can do you can use AI for essentially  anything that has um that uses statistics uh to   to determine an outcome that you can use AI for  uh and I think a big part of it really the low
• 12:30 - 13:00 low hanging fruit here is is uh pre-analytical  workflows um getting the requisition or order   into into a system like an Lis or getting it  getting it out from an EMR um there's a lot of   we we see a lot of data issues um for labs uh you  know e either you have missing patient demographic   data or or or you know Mal malformed patient  demographic data or or you have you know you're   missing a lot of information that that allows  uh the lab to make money and you know improve
• 13:00 - 13:30 revenues um there's a lot of that going on um and  and at the other end the post analytical too um   you have a lot of uh tools that are available for  post an like revenue cycle for example there's a   lot of opportunities to use AI there um so so yeah  overall I I think there's just a lot of lwh hang   fruit right now uh in terms of workflows it can be  done and even in the in the actual lab itself um
• 13:30 - 14:00 probably there's I think Abby mentioned on it  taking QC data U using AI to run QC data from   an aggregate perspective that's an opportunity  uh a lot of opportunities there as well but um   really what I see is the lowest hanging fruits  are pre-analytical workflows okay yeah that's   interesting smarter more accurate makes makes a  lot of sense getting it right on the front end um   say we're sold on okay AI is gonna help we should  bring it into our workflows we we want to do it   uh we signed we signed up with someone to to bring  it in houseal what's the challenge of implementing
• 14:00 - 14:30 it we want to we want to use AI to improve our  workflows what what are the obstacles to making   that happen right so all the things that you  just said about great about AI they true all   the terrible things that you've said also true  it really depends whether or not we can utilize   it right so the way I look at it is AI is like  a jet engine right it it can moves your business   at the speed of light it can get you to the point  8 be very fast but if you put your engine in your
• 14:30 - 15:00 lab without appropriate framework for that it's  like putting a jet engine on a horse bug you're   not going to go anywhere fast you probably will  not go anywhere at all right so that's the whole   that's the whole thing so um when we when we  look at it then sort of can we utilize AI I   mean that's a wonderful thing but can you really  take advantage of it so from that standpoint the
• 15:00 - 15:30 framework the other processes the automation that  the lab needs to be in place is if we still doing   copying and pasting from Excel spre spreadsheets  if we still doing manual entries like of a patient   data as you know been mentioned before and we're  making human mistakes doing that if we downloading   and saving files in five different places and  trying to reconcile them if we emailing each   other you know on on on on the results and patient  data then there's real use from that wonderful jet
• 15:30 - 16:00 engine that we can we can offer and advantages  that it it offers so um at lighthouse we serve   hundreds of customers we build labs for people  right that's what we do and my job uh Frankly   Speaking is to make sure that we build the lab  we have the infrastructure so we can put the jent   engine and we already doing it for some aspects  like you know predicting being paid or you know   where it can be paid uh but there is much more  extension to it uh but first thing first we've
• 16:00 - 16:30 got to make sure we have an appropriate sort of  vehicle to put this jet engine on so we can we   can really Propel the business that's basically  in a in a in a nutshell that's the problem you   have to have you have to put the basics in kind  of framework in place to use the AI and take   advantage of it Mak since um as we start to think  about implementing this I think that there's a   natural human tendency towards fear of what does  this mean for me is it going to take my job that
• 16:30 - 17:00 you're talking about AI doing some of these things  that I do does that mean that I'm being replaced   um how should scientists in the lab maybe even  Pathologists think about uh think about that   is AI the boogeyman coming for their their role or  uh is it going to help them to to you know be able   to do their job better anybody want to grab that  I'd say it depends it's it's both right if you're
• 17:00 - 17:30 doing your job well you can really take advantage  of a especially like pathologist is shiny prime   example uh this if if pathologist I mean they  have a tremendously uh challenging job using the   incredible cognitive skills you know tens of years  of training to find out you know the the malignant   patterns but then they have to spend a lot of time  quing their reports and you know working diagnoses   things that have nothing to do with with their  Prime specialty so think about the we're already
• 17:30 - 18:00 using AI to extract and code the pathology reports  it's a tremendous savings but you have to be kind   of skilled in understanding how it can how to use  it um and so on and so forth though it's a it's a   kind of an infinite brain for you that can bring  you any cases any patterns all the knowledge in   the world but you got to be able to use that  knowledge so from that standpoint if you're   not capable if you're not keeping up you very well  may be impacted negatively by by this development
• 18:00 - 18:30 however if you if you keep up with your job and  kind of stay current with it it's a tremendous   help it's it's it's enormous tremendous help so  that's my take on that yeah like f said uh go go   oh like please so I agree actually what he said  like absolutely agree there's tremendous value in   terms of how people can use it for day-to-day  basis uh everything from like reducing your
• 18:30 - 19:00 day-to-day work especially the Redundant pieces  of your work uh to kind of U helping summarize   it maybe create Trends better or create like  reports uh like internal lab quality reports   or internal lab like um uh end of the day job  reports these things could be very well automated by yeah exactly and uh yeah to what everyone  was saying I agree with what everyone was
• 19:00 - 19:30 saying um every time there's been a huge  technological shift the internet mobile   devices you know all this stuff coming out  in the past 20 30 years um if you want to go   really far back the Industrial Revolution um  and uh anytime there's a huge paradigm shift   in technology human beings have been pretty  quick to adapt obviously there's people that   don't adapt and those people get left behind  oftentimes but um you know I think with this   um AI is a tool U we don't have generalized AI  at all we're not even close to it at this moment
• 19:30 - 20:00 um I think some people might argue that we are  getting closer and closer but um today we're not   that close to generalized Ai and so until that  happens and uh they take over the world um I   think I think we should look at AI as a tool uh  in your toolkit and I think it just um it helps   to further abstract your job uh so you don't have  to do the boring stuff um you can do stuff that's   meaningful uh so I think in overall I don't  think it poses that much danger uh I mean
• 20:00 - 20:30 right now there's been a lot of effort made to um  made to for example like in other Industries like   full self Shing cars for example right um' been  working on that for years and years and years   but we're not even close to it yet uh and so same  thing here in the medical field um it's much more   regulated here uh and so I think yeah we're still  years and years away from it replacing anyone's   job in so far as there's um physical labor needed  to perform the job itself yeah I think I think
• 20:30 - 21:00 you're right I think we're ways out you bring up  an interesting point on the the self-driving cars   too I was thinking about um this the other day  that just the nature of being inside of Health   Care has such a high bar and such a low margin  for risk tolerance um that I think it's going   to be one of the more challenging places for AI to  be accepted um whether it's capable or not I think
• 21:00 - 21:30 self-driving cars probably see that same thing and  there's there's also this perception that even if   the AI was more accurate and less prone to error  than a human I think there's a there's still a   resistance to um you know the errors made by a a  computer versus by a human and I think we're GNA   face that same challenge inside of healthcare when  you have people's lives on the other side of these   results that even if it's 99.99 uh accurate we're  going to focus in on maybe those those few cases
• 21:30 - 22:00 when something slips through um even though as  humans we also have this and so that's H that's a   challenge and we'll have to see what what Society  deems acceptable in terms of error risk because I   don't think anything's ever error free but let's  move on to this uh this next slide here um vital   I'm going to direct this first question to you  how will AI power Technologies impact cost and   accessibility for Diagnostic testing for patients  everything going to be cheaper if so who benefits
• 22:00 - 22:30 did the labs pocket the extra margin insurance  companies the patient um you know is it gonna   get cheaper and if so all of right all the way  above right so very dear to my heart this question   I think about it every day in what I do um so  really the impact that's tremendous different   for different type of types of lab though  for clinical lab it's a process Improvement   the for clinical lab the technology to make  to to actually do the clinical test is pretty
• 22:30 - 23:00 amazing nowadays those tools those instrument as  chemistry analyzes they they they fantastic they   can do test in a very very quick and economical  way so what makes them long and expensive it's   all the processes around them that's all the  inefficiencies around you know entering patient   data validating results QC uh what not all  those things so if you strip all that away   then you can deliver diagnostic test for clinical  lab probably for pennies or dollars on a test no
• 23:00 - 23:30 problem so the rest of it comes from all the other  processes so that's where I can help tremendously   to reduce this cost now who will pocket that cost  a much bigger discussion we probably need another   webinar on that so in the current in the current  day um in the current day it seems like it seems   like maybe everybody or nobody body but who  should who should really pocket the cost is a
• 23:30 - 24:00 patient in my opinion so the system needs to move  to a different models that really really allow the   patient to be benefiting from smart decisions  about Healthcare as I said very different very   different but that's when maybe I can help maybe  I can help patients to be smart about their health   decisions and they test and so and so on so forth  so that's for clinical lab for Pathology Lab it's   very different but still pretty tremendous impact  so for pathologist the problem is that the great
• 24:00 - 24:30 at what we do we have shortage of them we have  shortage of good Pathologists in a country in   in the world for that matter so that's why I can  help a free out pathologist from some things that   not productive for them like you know coding and  and working with reports and and those type of   things but also create this extended brain  with the world knowledge of pathology that   can that that they can use at the fingertips and  actually really really improve the quality and   ability of pathology Diagnostics around the world  much kind of higher sort of road to travel uh
• 24:30 - 25:00 again because the the medical or says do not harm  so that's the kind of the whole medical mentality   hinges around that uh so that's really has to be  proven that does not harm first before anything   else uh but still um the fundament the impact is  tremendous of mainly from accessibility standpoint   for pathologist because the human still has to be  involved it's a full medical do so I wouldn't say   the cost should be a major factor but quality and  accessibility should for pathology that's that's
• 25:00 - 25:30 my take on that yeah I think um I read a quote by  Gary Kasparov who would play chess against deep   blue and there was always this conversation about  what will be smarter who will win the computer   or the human and what he said which I thought was  interesting is you know it might go back and forth   whether a human will beat a computer computer beat  a human but um a human working with a computer
• 25:30 - 26:00 will win every time right so when when AI is used  to augment and paired with a human brain um really   big things happen and uh that's probably where  the the rocket fuel is when you mix those two   things so that that's a good Bridge into our next  question for Jun um how will it help staff like   it will it help with the training we've got this  shortage going on we don't have enough scientis   well it helps you know people be more productive  they can get to more um what are your thoughts   on how AI might augment or assist with laboratory  Staffing shortage or just the lab tech themselves
• 26:00 - 26:30 yeah I think it's as far as the the lab job of  a lab tech I don't know if it changes all that   much because even if a lab lab tech is actually  working in the lab with the AI tools um it's so   abstracted at that point um so oftentimes you may  not even know that you're using AI in a lot of   cases uh but yeah it's very common in in I get we  call it human in the loop that's what we call it   um where in Industries where it's a high risk  or have a lot of liability uh you use AI like
• 26:30 - 27:00 John said you pair it with the human that has  expertise in that space uh and then you get   the best of both worlds uh right now so that's  already going on with pathology um you're seeing   some softwares out there that that are I think  procia is doing that um with with some ai ai   power pathology tools um the market but um yeah so  in terms of um how can labs more better implement   uh AI I think really it comes down to  understanding uh what AI can and can't
• 27:00 - 27:30 do um and so I think there's a lot of confusion  when I talk to folks about what it is and isn't   um especially what it isn't uh and so what AI is  right now is is a probabilistic model it's not a   it's not some kind of deterministic algorithm  that that tells you exactly what to do every   single time uh it just it just gives you Based on  data uh a probabilistic uh outcome that's really   what it gives you and so you can't just kind of  uh there there's a phrase when people say when
• 27:30 - 28:00 all you have is a hammer everything looks like a  nail and so you really really want to avoid that   I think um and so it's not some kind of Panacea  and so I think you as as a lab Lab person or lab   owner executive you have to have a really clear  understanding of value um it's like what are you   trying to do with the AI Tool uh and what are the  most important use cases that we're trying to get   done and so I think those are the questions that  you have to ask yourself um when you're trying to
• 28:00 - 28:30 implement any kind of uh tool for training staff  or optimizing workflows it's like what what are   the outcomes you're looking for um and then when  you find a tool that that because there's G to be   a lot of tools out there that come out in the next  couple years uh with you AI tools that come out um   understanding okay like maybe we shouldn't throw  the whole kitchen sink at this thing um maybe we   shouldn't maybe we should iterate one step at a  time and implement and um take small steps towards
• 28:30 - 29:00 adopting AI uh I think that's probably the right  approach uh and so for example even today um you   know we talk a lot about Ai and the implications  of AI um but I don't know how many people on this   webinar today have actually implemented or at  least used chat GPT once um and so I think taking   baby steps and hey just chat try chat GPT once and  see how you like it and see how that affects your   potentially affects your workflows I think you  start with that and then you can take other baby
• 29:00 - 29:30 steps or maybe integrate something other other  workflow or I don't know use it for marketing   purposes or something like that and then kind  of get your feet wet that way um I think trying   to come up with a grand like corporate uh what's  your AI strategy uh kind of a master plan is not   the best uh plan for kind of implementing the AI  tools that are available yeah I think you're right   taking it in bite-sized pieces I'm G to bring us  to the results of a poll that we ran um how close
• 29:30 - 30:00 do you think your lab is to implementing AI tools  and we had uh 24% of people said a long way years   um 31% said we're close maybe months um another  30% said we're already using it and um 15% said   no plans at this time not sure so mixed bag but  there's it's out there we got at least a third of   people already using it or third of labs rather  and another third um on Deck so um not not too   far out mons we're saying um AB I'm going to go to  you with this next question how does it affect the
• 30:00 - 30:30 patient we've talked about the business side of  the lab how it's going to affect the scientists   how it might affect our profitability as a lab  owner um what about the patient personalized   medicine does uh does medicine get better I think  right now Us's Health Care outcomes are pretty   pitiful especially for what we spend are we going  to start closing that Gap and uh really seeing   the patient benefit or what are your thoughts  on that yeah I think you know we all have been
• 30:30 - 31:00 like um like trying to take like steps towards uh  having like personalized Healthcare and I think   that has been in discussion for like a decade now  uh but I think AI is that technology that really   brings us closer uh without having to make it too  complicated um um I I see a lot of use cases on   um like um helping patients like really understand  the results uh better uh and also you know getting
• 31:00 - 31:30 a sort of um like a virtual consultation on  their results um maybe extending it to like   helping them decide what sort of tests they should  do on a routine basis uh but you know that really   depends on what state and you know what sort of  uh insurance you have we also do see um like big
• 31:30 - 32:00 relevance of like uh personalization happening uh  on the uh bailing side where uh patients can not   only understand uh their qualifications of their  insurance better but also you know um understand   what what sort of test or what sort of costs they  would be incurring and you know just get a general   sense of of um you know what what is it going to  look like from a uh like an xense or Healthcare
• 32:00 - 32:30 cost time Point um so there are there are a couple  of areas that I think the most exciting area that   we find for personalized medicine really is like  helping patient understand those tests better and   like kind of elaborating more on these areas uh  but you know there are these other set of like   scenarios that you know if it works out it'll  be it'll be crazy sure yeah I think um there's
• 32:30 - 33:00 so much data out there and maybe some of these  correlations might be a big part of that um kind   of brings us into our next question and I've  evolved in my thinking over this over the last   couple years um I've always been an advocate  for the value of labs inside of medicine you   know everybody knows that quote that's probably  overly quoted uh 75% of medical decisions are   uh based on laboratory data I thought great AI  uses tons of data it needs data that's the gas
• 33:00 - 33:30 and the fuel for AI we produce it as the lab this  is our moment we get to step into the spotlight   here and really see the value we drive because  uh laboratory currently only uses I think it's   like 2.6% of healthc care spend goes to the lab  and we're producing 75% of the data behind the   decisions I thought this is this is it this is AI  is going to be the thing that really um puts us   uh puts us in the Showcase light um more recently  I've started to think that that data that we've
• 33:30 - 34:00 produced May uh may just get digested used by  others I don't know I can't see a clear path   to how the laboratory benefits maybe the type of  data isn't the type that um really drives AI so   I'm gonna kick it over to Jun because we had  a little bit of conversation about this that   I thought was enlightening on the difference  between maybe structured data and unstructured   and some of the more valuable types of data for  um for AI how how does the lab benefit or does
• 34:00 - 34:30 it um from ai's appetite for data yeah so um as a  lot of people know uh I think there's tremendous   demand for dat like diagnostic data that comes out  of the lab uh you know a lot of biofarma companies   purchase that um they can't seem to get enough  of it uh and you have entire companies uh for   example flat iron Health uh they have the the enle  EMR and they had a huge data play on the oncology
• 34:30 - 35:00 space uh and they sold the data to biopharma  and whatnot and um and so there's we do know   there's a huge demand for that and and that labs  are the source of that data so that theoretically   theoretically you should be able to get uh some  kind of monetary value from the data itself um   as it pertains to a use cases in Ai and whatnot  um so far it hasn't really materialized the the   way that I think a lot of people expected it to  um even the lab is the source of the data and   the lab is the one driving so much decisions um  by the way John I asked my wife about that that
• 35:00 - 35:30 75% thing she thinks it's more like 90% of decis  she's a physician so she's telling me maybe it's   more like 90% because I I I look at quantitative  data all the time uh but yeah so anyways um uh   yeah there's there definitely an appetite for it  but so far it hasn't really materialized because   you see a lot of the software the data resellers  and people that aggregate data um they're getting   the data from the EMR uh the patient chart which  then combines the lab data with the all the other
• 35:30 - 36:00 uh encounter data from The Physician and so  they're selling all those all the uh kind of   combined data to the biopharma companies out  there that want the data and so from a kind of   the perspective of the lab um it's been a little  limited but I'm not I don't think it's necessarily   tapped out in terms of opportunity because the  lab has information uh diagnostic information that   perhaps the the emrs don't have U maybe it  has you know assay related information that
• 36:00 - 36:30 matters U maybe there's QC data that matters maybe  there's more uh maybe the lab is actually a better   aggregator of overall um patient uh data overall  like uh qualitative and nonquant qualitative um   for example like as we get into genomic sequencing  um right now like people have talked about okay   how do we integrate genomics testing into the EMR  right now it's really hard to um so maybe there's
• 36:30 - 37:00 applications in you know NGS that pertains to  uh you know being able to sell I mean and it's   very sensitive topic because obviously genetic  data is uh very sensitive but um there could be   opportunities there for the lab uh when they get  into more and more esoteric or high complexity   testing uh but certainly I think there there is  still value that that we not may not be seeing   right now uh to be able to capture value from the  data that that the lab produce every day and I
• 37:00 - 37:30 think it was helpful for me when I spoke to you  about the fact that like a a lab value if that's   the data um maybe AI is not really needed in order  to run a formula right like we've been able to do   that with spreadsheets or whatever for a very  very long time and so um just having numbers um   probably isn't the type of data that's really  going to be monetized it needs to be whether   it's imagery or it's non-structured or it's um  different types of data being bundled together   through the EMR I think that's where really  the value starts to get unlocked and uh good
• 37:30 - 38:00 for us to think about I know I talk to lab owners  regularly that have a a data play where they're   going to sell their data and it normally means  I have a bunch of lab values with decimal points   and I'm hoping somebody pays me for those um and I  don't I don't know if that paycheck's coming um so   starting to think about ways to u to put your data  in a a format that's going to be more valuable   by pairing it with something else or maybe it's  some of that imagery data for those groups that   are doing AP um I think 23 and me was the one  that everybody was pointing to as they they had
• 38:00 - 38:30 made this big data play and it's not not playing  out real well for them recently if anybody's been   following that in the news so um okay we're gonna  move over to our last slide just in the respect of   time um and the first question is going to go to  Abby uh what steps are being taken to ensure that   the AI algorithms used in diagnostic testing are  transparent unbiased clinical validated how do you   make sure that there's not this uh you hear people  talk about it this you know bias that's built into
• 38:30 - 39:00 the training set that carries over and has maybe  bad uh outcomes for patients yeah joh um this is   a question that I think keeps all of us up at  night and I think you know that really ensures   that whatever these Solutions get buil uh you  kind of make them uh as unbiased as possible and   as validated as possible um so one of the ways I  think um you know really make sense is one to have
• 39:00 - 39:30 a feedback loop in place so uh you have um like  any sort of AI suggestion or an action kind of has   a sort of um like a thumbs up thumbs down at the  least which kind of factors in a minimal feedback   loop at at least on top of that um you could offer  like flagging Solutions where people can really
• 39:30 - 40:00 flag unexpected uh scenarios and ultimately I  think um the more relevant training data that   you can actually feed into um the the more like  closer to expectation the AI system would be   so um I think we really believe that um at the  stage it is today uh AI solution should be sort
• 40:00 - 40:30 of suggesting or sort of aiding or guiding and not  really performing the action itself uh so I think   that would be the right sweet spot where you kind  of still get the benefit of um like a more quicker   workflow but um still are cautious on deciding  when um to let AI do it automatically yeah
• 40:30 - 41:00 absolutely um that those ethical questions lead  us into kind of regulatory um the groups that are   going to be responsible for that um I'm going to  direct this question to Vitali what do Regulators   whether it's uh CMS or FDA who knows who it's  going to be these days what do they need to think   about as they try to write the rules for this I  know we're already working with several uh liquid   biops companies that have ai algorithms built into  how they're um reporting out their test results a
• 41:00 - 41:30 little bit of the Wild West right now certainly  wasn't contemplated in 1988 when the most recent   set of clear regulations was written um what are  the what are the themes and topics that Regulators   should be thinking about as they try to draft  the rules of the game for using AI in the lab   yeah that's a billion dollar question right um so  the thing is that Healthcare Market is La Market   is not Market in a true sense it's not driven by  market forces it's driven by regulations uh makes
• 41:30 - 42:00 it special for better for worse um and um those  Regulators in my humble opinion are hopelessly   outdated hopelessly outdated by decades um and  they driven by best intentions patient safety   right kind of do not harm that's the that's the  rule number one but but they have to come up with   the ways to actually catch up with the uh with  the world and with the current current state of
• 42:00 - 42:30 affairs so regulations that put in place they mean  well they want to protect the patients but they   provide tremendous hindrance to the development to  to the progress so um how that should change it's   really up to them but that's where so basically  the regulations need to change to be conducive   to a um kind of quick development without  compromising safety it's I know it's a to order
• 42:30 - 43:00 but um if we do not make progress there we will  be lagging behind and um probably even more so and   that actually affects patient safety and patients  outcome uh to start with so that's a that's a very   very big question that The Regulators have to  be facing nowadays yeah if anything I think that   the risks are are pretty well flagged in terms of  uh antennas are up people are concerned you know
• 43:00 - 43:30 what how how might this negatively impact things  maybe even to the too much so so that we we lose   out right I was listening exactly you kind of  throwing in the baby with a buff up water you   know that's right there's there's like um some  therapy tools that are AI based like um for   behavioral therapy like what you would see with  counseling that um you know completely I AI back   and could provide you know services to people  that can't afford to go speak to a therapist
• 43:30 - 44:00 and this is just like analogy or kind of like  a parallel type offering but I could see that   inside of healthcare too I heard of optometrist  recently where there was a AI back tool that could   do the job almost as good as an optometrist  and the optometrist themselves said well it's   it's about like 97% as good as the human um so  they didn't see real value in it but it could be   used to reach groups that could never go I doctor  right and so how do we understand those risks um
• 44:00 - 44:30 but not also be so scared of them that we we just  allow people yeah and also and also those outdated   processes make the make the health care and lab  specifically very more way more expensive than   it could be it's another thing it adds the cost  yes sometimes tremendous cost Jun your thoughts   potential limitations and risks associated with  this what how should we be thinking about it   should we be thinking about backing down kind of  the standard that we hold it to so we can reach
• 44:30 - 45:00 more people for the greater good um holding  the line make tightening it up what are your   thoughts in terms of how we should be thinking  about Ai and Healthcare yeah uh as Vali said   uh Regulators just by nature of their job uh are  going to be hopelessly outdated in terms of how   they think about AI um they're already somewhat  outdated in terms of how they think about Cloud   Technologies uh and modern software in general  or Hardware in general um but again it's like   Bal said it's not it's not because they're stupid  it's because they have good intentions that that
• 45:00 - 45:30 they're conservative about it um I would say that  some of the things that are inherent to AI make   it really uh risky uh in terms of use cases and  clinical Diagnostics uh the biggest one is that   AI tends to be a black box uh meaning that uh you  get you put in one input it's not always going to   give you the same output um and so so what that  means is that when something goes wrong in AI
• 45:30 - 46:00 applications um now you don't really know why it  goes wrong just like if you have a self-checking   car takes a left turn out of out of the blue you  don't really know why until you do a postmortem uh   and so that's one of the the limitations and risk  associated relying heavily on AI um obviously like   like I said earlier um when all you have is a AI  Hammer uh everything might look like a nail uh so   you have to be careful with that um but I think  the as long as folks adopt AI tools um where it's
• 46:00 - 46:30 less risky um you know like you don't want to be  messing around with AI uh all the time for uh test   resulting and whatnot um but if you're doing it  for something a little less little little bit less   risky I think um it it der risks the entire thing  a little bit um and obviously right now um there's   a lot of uh Innovation happening in in the field  uh right right now and so a lot of the limitations
• 46:30 - 47:00 that are current U might go away in a couple years  uh and so I think it's it's really important to   continue iterating uh on how our understanding  of AI goes and um yeah I think right now though   that there is risk in there's risk in kind of  understanding having to understand that the AI is   a black box and when when it goes wrong you don't  know why yeah yeah as we clinically validate tests   in the lab ldts in particular we're we're running  you know uh large batches of samples to make sure
• 47:00 - 47:30 that we get the outcome that we expect to get  consistently under different standards the the   sample's been sitting on the shelf for x amount  of time whatever it is um do we consistently   get the result that we're planning and I think  that um as you implement AI into your laboratory   workflows especially the clinical diagnostic  side at a minimum it should be part of that   validation pipeline right and we might might  not fully understand why we're not getting the
• 47:30 - 48:00 result but at least we can we can test that um  with a a significant end to to say do we as we   put things in one end to the pipe do they come out  as expected on the other end um is a is a minimal   level of uh of validation that I think we need to  do inside the cleab environment um so moving this   is just kind of looking ahead um clinical lab  versus anatomic lab um where do we think that   it's going to have greater impact anyone want  to opine on which of these is going to change
• 48:00 - 48:30 first right the answer well obviously answer  is both right as we talked about answer is   answer is both in different ways and there's  a different kind of hurdles to cross for the   lab for clinical lab I I think that's more  attainable actually uh because it's in the   lab's hand I mean really if lbs catch up with  the technology progress in and kind of improve   the processes they can really even today they  can they can take advantage of of of latest
• 48:30 - 49:00 AI um so by simplifying simplifying the process  making them more efficient and all all those good   things getting better paid um but for pathology  it's a big big promise uh but it's a longer uh   Longer Road uh to success because that's where  it needs to be cly approved and and accepted and   uh weaved into the pathologist training and kind  of uh workflows and those people don't change very
• 49:00 - 49:30 easily I mean I worked with Pathologists I mean  they've been trained for 10 20 years and you know   they they stick to the ways um so that will  be a much kind of Longer Road um to to to to   to the benefits but again potential potentially  tremendous tremendous U tremendous benefits for   both of those sure um yeah I think the in my mind  at least the the imagery associated with ap seems
• 49:30 - 50:00 to be something that we previously hadn't been  able to really use use our systems to to benefit   and the idea of being able to look at millions of  images um be smar because of but that's why it's   also kind of coming back to that risks AI has  an inheritant problem very inheritant problem   it's trained on existing data so it can provide  the data for you it can come up with with even
• 50:00 - 50:30 generative right based on existing data but by  definition it's not going to give you answers   if the answers have been known before so that just  it's just the way and we have to understand it so   that's a that's a major risk right if that's I'm  using a simple example if you are um if you're   are kind of training a child and AI is basically  child without content if you're training Ai and
• 50:30 - 51:00 saying this is a picture of a dog this is picture  of a cat this is picture of a dog this is picture   of a cat child will learn that very well but then  you show a picture of a tiger and child will say   what it's a cat because it hasn't it hasn't seen  tiger before so that's so that something we have   to understand about fundamentally right that's  what that's that's where we have to we have to be   very very careful about it uh in terms of getting  getting the benefits yeah as a lighthouse we do
• 51:00 - 51:30 a lot of recruiting and I think believe it was  Google you guys can fact check me if I'm wrong   but um produced a resume screening tool that  would replace a recruiter right it's going to   look and it's going to tell who's qualified and  it's going to be able to just utilize uh AI to   source resumés and make sure that people qualified  and then as they deployed it they realized that   it was fairly sexist and racist because it felt  like Executives should be maybe white males set
• 51:30 - 52:00 it had looked at was that was what was out there  right that was was in the roles and so as it tried   to find lookalikes um there was that bias and so  that they caught it they unplugged it they didn't   roll it out as they saw it but a great example of  maybe it doesn't never seen cat or tiger before   and so it doesn't know that exactly it is so  um I'm G to move us over to the the Q&A from   uh participants to see if we can get a couple  of those in we're coming up on the hour so we   get about seven minutes left I'd love to tackle a  couple questions I see multiple in here that are
• 52:00 - 52:30 intrigued by the pre-analytical looks like you  know that that comment jiun about how how do we   use AI to maybe uh help in the pre-analytical  processes is something that people are finding   intriguing can you talk about that anymore or  give an application of what that might look   like um to use AI on pre yeah yeah uh so I think  uh one of the easy kind of uh there's a lot of   lwh hanging fruit is is what I would say because  we always say at least in the revenue cycle space
• 52:30 - 53:00 people say garbage in garbage out uh and so a  lot of the bad data frankly comes from Lis uh   in the lab space and so uh if you can optimize  the data coming in correct and and correct for   the missing data or incorrect data um that can  really help and so what I mean by that is is   um there's multiple sources of of data coming in  to to the laboratory via Lis um whether it's EMR   whether it's a portal of some sort whether it's a  manual requisitions um there's the kind of obvious
• 53:00 - 53:30 uh applications for AI or or are um demographics  demographics cleanup um and documentation uh and   and when you have documentation you can when you  scan in let's say a requisition form or scan in   like an insurance card um there's tools out there  nowadays where you can scan the insurance card and   then it will take the text uh from the insurance  card and then transcribe it into form and then it   saves data um I I've asked around and and almost  nobody's adopted this in in in actuality um just
• 53:30 - 54:00 because people don't know about it and uh when  people are told about it they pretend like it's   black magic uh and I'm like well that's that's  AI for you right there um this we call it OCR or   optical character recognition um there's also  you know things like uh demographics checking   for example that's that's not when you get that  when you have a person's name and it's misspelled   um sometimes you could run an AI algorithm or  something like that to maybe spell it correctly or
• 54:00 - 54:30 autofill addresses or autofill information about  the person that may not be possible by human being   um there's also on the other side of things uh  post analytical side of things too there's a lot   of applications there but yeah I think I think  the ones I just mentioned are probably the most   obvious use cases for AI um and that's applicable  today uh not 5 years from now you can do it today   uh and so those are the ones I'll I'll bring up  excellent um jump to another question um here's
• 54:30 - 55:00 kind of a specific case but maybe we can broaden  it to um to have more application um Brian conl   is saying hey when we use our T can systems common  problem is either the tip can't reach the liquid   because the liquid's too low or goes too far  down and hits the bottom of the plate can AI   help is this a is this a application for AI to be  able to come in and determine how far the tip is
• 55:00 - 55:30 going into aload out of a sample can it be add  Dynamic ability to come of some of our existing systems I think the answer is I mean in some  cases it can but I think in this this problem   that you have you might want to talk to the tan  person get that result um no I think it is it   is representative of something that we kind of  work with systems that tend to be pretty well
• 55:30 - 56:00 coded in a finite way static way like this is  what it does every time and there probably is a   an evolution into it being more Dynamic at some  point to be able to look at what it's dealing   with and adapt probably better than older system  have been but probably probably a littleit ways   out um going through here think about billing  applications uh how practical affordable are   these tools for like a smaller lab right it seems  like AI That's great might help Lab Core Quest or
• 56:00 - 56:30 a big Health System but what about what about my  small lab is this going to be something that's in   the nor possibilities for me to reach yeah I can  I can definitely speak to that I mean we able to   pack even today we able to package it as a kind  of plug andplay solution for everybody La big   or small we basically what we do we can we can tap  into the current flow of your normal document that   you exchange with the payers so that's actually  one area that's standardized as a federal Hippa   law that requires everybody to send they claim  insurance information and payment information
• 56:30 - 57:00 exactly the same way which makes it very very  efficient for us we can just kind of tap into   that tap into that flow and it's really Plug and  Play Big Small doesn't matter there is like pretty   much zero uh zero effort to do it and here we go  you can you can have ai question is can you take   advantage of do you have a work process to sort of  pick up up and and you know follow up with it and   actually uh do something with it so that being  actionable that's the question the fact that
• 57:00 - 57:30 we can do it today absolutely yes for everybody  that's not that's not a sort of technical issue   today thank you uh I will um go ahead and  close us out here I want to thank ABI Vali   Jun appreciate your time your thoughts uh your  inputs on this thank you all for attending um we   we really appreciate it we will have an upcoming  webinar our next one I think we're going to do   sort of an ad hoc one whenever the FDA  ldt announcement drops we're going to be
• 57:30 - 58:00 uh following that a day or two later digest  that information as quick as possible and   bring you some of the best legal Minds in the  business on what that means for your lab so   we're watching that closely keep an eye out for  um notices on that but thank you guys all for   attending uh thank you to our our guest speakers  really appreciate it thank you very much John