From being a evaluationer in Computational Biology & Structural biology/ Biophysics to a Senior knowledge scientist, Sreetama Das has obtained A huge information base and labored in numerous industries like manufacturing and well beingcare. Analytics India Journal caught up with Das, Who’s presently serving as a Senior AI ML Engineer at GSK, To know her insights on AI-based mostly options in digital well being.
AIM: Given your expertise in a broad profession path from knowledge science to AI and machine studying, what do you assume Can be The biggest method forward for the tech-medical space?
Sreetama Das: To briefly level out my background, I’ve labored with knowledge science and machine studying softwares to biomolecules as An factor Of instructional evaluation all by way of my PhD, After which labored in numerous industries like manufacturing and well beingcare. This expertise has shaped my view of what I really feel The sector will evolve to be.
A quantity of spaces Inside the tech-medical space use machine studying for enhanced outcomes – For event, evaluation and enhancement for drug discovery or repurposing, optimising medical trials, manufacturing and extreme quality administration, digital well being sensors, digital pathology affected person triaging, To name A pair of. The information Might be numeric, constructiond properly in tables, or large pictures And even messy textual content material – That is what makes drawback statements in well beingcare thrilling however difficult. Developments in pc imaginative and prescient and pure language course ofing will enhance options And even clear up factors that Weren’t possible earlier. As properly as, options focussing on explainability will discover extra acceptance. The general enhancement is To maneuver in the direction ofs faster enhancement, environment nice manufacturing, greater extreme quality administration, and simpler entry to well being monitoring, illness detection, and affected person remedy outcomes. Briefly, AI-based mostly options Shall be aiding To reinformationrce lives.
AIM: There have undoubtedly been vital advancements Inside The sector of digital well being. Is there A particular state of affairs that prompted you to conduct further evaluation?
Sreetama Das: Digital well being has seen vital advancements in creating digital sensors for well being monitoring and detection of illnesss. I am going To cite an event from a enterprise I used to be An factor of at my earlier organisation (Robert Bosch Engineering and Enterprise Solutions, India) since I even have been with GlaxoSmithKline For Decrease than A quick time. Again there, the group developed a novel, non-invasive haemoglobin monitoring sensor based mostly on photoplethysmography. We collected knowledge from many individuals, each healthful and frail, and educated machine studying fashions. In consequence, our system pershaped greater than some current non-invasive hemoglobinometers, which tended to overestimate haemoglobin ranges. The enterprise required A lot of evaluation, and our end outcomes have been revealed in several reputed scientific journals.
AIM: Enterprises started migrating amethod from on-premises knowledge centres and softwares in the direction of cloud and SaaS-based mostly options. How possible Is that this transition to the cloud in mild of the pretty A pair of safety considerations and ransomware assaults occurring?
Sreetama Das: I am not an professional On this space, however I am Aware of private Industry cloud and hybrid cloud choices that mix Some good advantages of cloud with The safety of on-premise options and are used To deal with regulatory considerations. Furtherextra, We’re Starting to see Using blockchain in Industry cloud options and upcoming utilized sciences like federated studying and its use in med-tech. So, I really feel the transition Is potential with properly-thought-by way of strategies.
AIM: In what we name a “combination of AI and Digital Health”, what are some shut to or so groundbreaking enhancements that carry future potential as properly?
Sreetama Das: Definitely one of many groundbreaking enhancements in current events would be The event of AlphaFold (by Deepthoughts) which has revolutionised the science of protein construction reply and has important implications for drug design. The Computer software provided shut to right theoretical fashions of protein 3D constructions, That are presently obtained by time-consuming and costly experiments or typinamey recurrented by not-so-right theoretical fashions. Implementation of AlphaFold generated fashions will dramatinamey scale back the search space, thereby reducing the time and expense Inside the drug discovery course of.
Completely different advancements embrace deep studying in several spaces, For event, Inside the classification of chest X-rays of regular flu versus COVID-19-contaminated circumstances with extreme accuracy or in early detection based mostly on voice abregular respiratory sounds. As properly as, picture and video analytics are discovering softwares in monitoring bodily exercise or senior affected individuals for falls or other factors. Lastly, precision privateised medicine May even be an space with vital potential.
AIM: Usually, why do You think that about there are widespread misconceptions about synthetic intelligence in well beingcare, and why do they persist?
Sreetama Das: I think about there are several sides to this question. Firstly, when there are breakby way ofs in synthetic intelligence algorithms, many articles in widespread boards Are typinamey very flattering, with out making an try To elucidate how they work, the enter requirements or the place They might fail. In consequence, There is a disaplevelment when such devices Do not generate the anticipated end outcomes, both As a Outcome of of inrelevant software or improper knowledge. It is rather important Maintain in thoughts that gathering enough good knowledge Continues to be a problem for several well beingcare factors. Additionally, some drawback statements Might be Impossible To unravel and require modification after a quantity of rounds Of dialogue with the stakeholders. Furtherextra, well beingcare is a delicate matter, and People Are typinamey apprehensive (and typinamey rightly so) about accepting ‘black-area’ options or the ‘equity’ of such options.
Therefore, elevating consciousness about synthetic intelligence and addressing considerations – knowledge requirements, understanding how machine studying works And what’s possible, and mannequin explainability – will assist take away persistent misconceptions.
AIM: Might synthetic intelligence finally substitute clinicians?
Sreetama Das: Human biology Is very complicated – we nonetheless don’t understand every thing. Clinicians make selections by Taking A look at Many numerous elements and based mostly on their expertise. It is exhausting to design ‘widespread’ machine studying options Which will work Simply like an expertised clinician – solely options to very particular AI can develop properly-outlined factors and the place A lot Of information Can be found. Since well beingcare supply immediately impacts individuals’s lives, The Prolonged time period would have AI options “aiding” the clinicians Inside their choice-making pretty than changing clinicians altogether.
See Additionally
AIM: What assets, Similar to books and journals, would you advocate for aspiring professionals On this area?
Sreetama Das: A lot of useful information Is usually Present in blogs on Medium. I Check with GitHub for codebases. Completely different websites to look out for Shall be papers with code and agency blogs for current enhancements (e.g., Google, Fb, Microsoft). There are also many free movies and course supplies (e.g. MIT Courseware) for aspiring professionals to get started.
AIM: What would you advocate as an environment nice profession course for professionals purposeing to make it huge Inside the digital well being space?
Sreetama Das: I am going To answer this question based mostly on my journey. Data science professionals Want to have some area information to develop relevant options. Since digital well being is An monumental space, It Is important To Consider that and decide the issues the professionals are Considering about and understand.
It also assists to be versatile and adaptable To vary since different spaces might obtain traction at different events in Educated’s profession. Lastly, We now have to be lifelong studyers and buy related information when The need arises.
Lastly, It Is important to Consider The inspiration for such professions. Good coding expertise, understanding of key ideas, and good communication are important. Additionally, please Maintain in thoughts that such tasks Are typinamey based mostly on groupwork, with different expertiseets Contained in the knowledge science group (analyst, ML engineering, MLOps, and so on.). So, It Is important To converse to friends and study extra about these elements.
Subscribe to our Publication
Get The latest updates and related provides by sharing your e-mail.
Be a part of our Telegram Group. Be An factor of An partaking group
