AI for Health - A promising future in healthcare
The online edition of “AI for Health” conference took place on Nov 10th 2020. The international participation was important and highlighted the intense activity in this area today.
Artificial Intelligence (AI) is a profound trend, where deep learning networks have invaded the data analysis space as of 2012.
AI in healthcare has expended from imaging analysis in oncology, such as Computed Tomography scans ‘review for early lung cancer (re)detection, to faster end-to-end pharmaceuticals faster discoveries with simulation, analysis and selection of candidate molecules. The American Food and Drug Administration now reviews en silico models and AI based simulations of “virtual clinical trials” to accept a clinical study design approach as reported by Nova. These strategies can start with a limited amount of data if properly combined to appropriate expert input.
With Natural Language Processing, AI is going one step further in early / innovative diagnosis. Most powerful impact of AI in health is expected to arise from analysis of real-world evidence, necessarily representing big data. Optina are working on retina imaging and analysis to uncover mental health pathologies such as Alzheimer. One of the goals of AI use is also to reach a better adequation between a tailored population and a specific treatment. Several actors are working in those directions (example Savana, AFX). France is leading an innovative centralization of medical big data with the Health Data Hub since 2019, the working groups are creating recommendations, collaboration means and partnerships.
The tendency is to use AI as an additional tool, to support physicians and medical staff where it can provide better results than the human opinion alone. Practitioners can then free their precious time for more important and complex tasks. AI now directly or indirectly plays a role at all clinical steps:
- Health behavior – personalized food complement pill for the general population
- Diagnosis - such as detecting a fracture on x-rays and detection of new biomarkers and risk factors based on retinal or face images, or promotion of self-diagnosis
- Treatment planning –thoracic images are sent to the Cloud and analyzed via AI and a second review doctor to support local pre-surgical planning and someday augmented reality visualization will be used in the operating theatre
- Treatment delivery –in case of glucose monitoring associated to insulin delivery tailored to patient specific need in diabetes
- Patient management, chronic sickness monitoring – to engage the patient via chatbots, to detect and provide early warning in case of worsening condition in epilepsy, cancer relapse, drug prescription, etc…
Additionally, the medical community is hoping that AI will provide a chance for everyone over the planet to get the right diagnosis and treatment plans thanks to mutualization of the medical data, combined to inputs from the best world experts and immediate feedback to the local caregivers. Typically, in Australia, Canada or the USA, professional caregivers may be located close to the patient but remote from university hospitals and appropriate specialists. AI has to potential to shorten the distance between patients and specialists. Non-for-profit foundations are active in promoting such development for cardiovascular treatments, which are now scarcely available in large portions of Africa.
The industry highlights that some AI development steps are required in the coming years:
- Focus on education, with the creation of dual pHd programs in data science and computer engineering. Collaboration between a variety of specialists is needed to use AI in a valuable way: data scientist, software developer, IT engineer, subject matter expert, regulatory specialist, etc. Professional “translators” between the different specialties are needed.
- Clarification of the regulatory acceptance framework at national and international levels – on the way with EIT Health and HAS recommendations
- All stakeholders are sensitive to the confidentiality issues related to sharing medical/personal data. The actors claim strict conformance to GDPR, HIPAA and work on cybersecurity. Building trust and confidence will be key to ensure welcoming of the new technologies in healthcare.
With this promising future of AI in healthcare, we look forward to meeting all actors live next year.