AI for improved diagnostics and patient outcomes
US-based healthcare experts recognise the potential to leverage the collaborative momentum built over 2020, converting it into long term gains across the sector, developing technologies and communication networks that will ultimately improve patient outcomes...
Digital Diagnostics - Forging a New Path: Healthcare AI Done the Right Way
Digital Diagnostics is the leading pioneer in autonomous artificial intelligence. Founded in 2010 by a physician/scientist and computer engineer, Diagnostic Science has created a unique, patented biomarker-based approach to build algorithms to “think” like a physician so computers can make clinical decisions without human intervention. These algorithms are integrated into easy-to use systems that enable clinics to remove the diagnosis burden of common diseases from specialists. The FDA cleared the company’s first diagnostic system for diabetic retinopathy and diabetic macular edema in 2018, which is rapidly gaining adoption among the largest and most prestigious health systems in the U.S and globally.
Shifting from experimental research to prediction verification
Brandon Allgood is Chairperson of the Board at AAIH and VP, Data Science & AI, at Valo Health. Here he envisions AI driving shifts in experimental motives and outcomes and, ultimately, the potential to remove the need for animal-based studies for human health. “Some of the bigger CROs are really starting to think about machine learning and embracing it instead of opposing it. We are seeing a time when we move from doing experiments for results to a time when we’re actually doing experiments to verify prediction.” Allgood notes that, while the difference is subtle, this means that when carrying out experiments the probability of getting predictable data and successful results is very high, minimising the number of experiments that need to be carried out. So, he suggests, CROs and laboratories that base their work on volume might wish to start thinking about quality over volume as in-house models get better and better.
Beyond 2021, the AAIH envisions their work with the FDA will lead to a time when in silico modelling will evolve into system ‘whole human’ systems rather than simply in vitro or in vivo models. This will both enable improved predictions and reduce, and ultimately remove, the need for animal studies. “We know today that animal models are not good at predicting outcomes or predicting how a compound or biologic is going to interact with the human body. But as we start to use machine learning we’re going to get to a time where we’re going to be doing far fewer animal studies, and a time when we don’t do animal studies anymore,” he said, offering his own prediction that in silico models of humans will become better than animal models of humans. The Alliance for Artificial Intelligence in Healthcare (AAIH) is a coalition of technology developers, pharmaceutical companies, and research organisations who have expressed the common goal of realising the potential for AI and machine learning in healthcare to significantly improve quality of care, but who also recognise the need to address substantial industry challenges.
Amplifying end-to-end collaboration in science
Rafael Rosengarten, CEO of Genialis, believes we can leverage pandemic-driven collaboration approaches to continue to develop collaboration across drug development silos to support end-to-end collaboration and close the drug discovery and development feedback loop…
“2020 and COVID brought a collaborative approach to science in unprecedented ways. In the second half of 2021, when we return to more in-person-type interactions, the over index on collaboration will be a good thing.”
When asked for his thoughts on the outcomes of pandemic-driven collaboration, Rosengarten said; “…people miss hugging other people and the physical contact we used to share. And so, I think that folks are going to be really eager to continue to amplify this collaborative approach to science.”
Rosengarten notes that in addressing the global health crisis created by the COVID-19 pandemic, it was each to become deaf and blind to the progress made by science and biology throughout 2020.
“2020 has been a banner year for the application of AI and healthcare for public health reasons and otherwise,” he said, explaining that he’s most excited about now starting to think about how we build links across the current silos to connect workgroup functions in drug discovery and development. He hopes for the development of systems and frameworks to link the breakthrough technologies in generative chemistry that uncover new molecules, how they function and what they target, with the translational medicine at the clinic end, that takes clinical trials for these biomarkers into the market. He suggests that these solutions will most likely rely on AI.
“In 2021, it's about connective tissue. It's about connecting people, connecting data.” He says. The application areas required to make this happen are ripe. We should now be able to connect to one another and close the collaborative feedback loop.
AI-driven omics for clinical diagnostics
Dr Klaus Lindpaintner MD, is Chief Medical Officer and Chief Scientific Officer at InterVenn BioSciences. He believes we are currently witnessing an explosion in the application of omics to clinical diagnostics and that laboratories will increasingly focus on protein-based markers…
The development of medically applicable predictive tools is being driven by significant progress in wet-lab technology, suggests Lindpaintner.
Specifically, for instance, mass spectrometry vitrectomy – evaluation of biomarkers in protein rich vitreous biopsy samples. “That has really propelled proteomics forward from where we used to just look at genomics,” he said. “What we’ve been witnessing is the advent of what people refer to as liquid biopsies. Mainly - the ability to really understand, treat and monitor cancer based on blood samples rather than what used to be necessarily tissue samples, which are much more invasive and difficult to get.”
As we move into 2021, Lindpaintner believes more and more laboratories and companies, while continuing to look at DNA-based markers, will increasingly focus on protein-based markers.
“The amazing power of artificial intelligence allows us to really look at the data and process the amazing amount of data that are being generated,” he says, noting that AI-driven data collection and evaluation is both scalable and ultimately applicable to clinical practice, to the benefit of the patient. This has both fertilised how experts think and energised the clinical diagnostics field in a dramatic fashion. “…the pandemic has clearly highlighted the importance of biotechnology - the importance of thinking very actively about how we can improve medicine as rapidly as possible.”
"AI trends allow us to mine very very large sets of real-world clinical data. To extract information that is otherwise very, very difficult to get at. And that clearly fertilises clinical research and ultimately allows us to devise the testing algorithms as well as, down the line, novel therapeutics that heretofore just were not possible.”
Just as current AI-driven tools can recognise faces, future applications will recognise molecular structures in a way that is vastly more powerful and faster than was ever possible before. Given the complexity of biology, Lindpaintner believes the advance of artificial intelligence, machine learning and neural networks is essential to enable us to move forward at an increasing pace.