admmm

This forum focused on translating AI-based innovations in medicine into the commercial world and the clinic. The first speaker, Qian Cao, spoke about developing AI-enabled software as a medical device and the FDA Centre for Devices and Radiological Health’s perspective on AI/ML devices. Our second speaker, Eli Gibson, spoke about the complexities of translating AI/ML ideas from a manufacturer’s perspective and showcase examples. The presentations were followed by questions and answers about the talks, and discussion focused on lessons learned, opportunities to collaborate, and sharing of data.

The October forum focused on using machine learning to predict disease progression. Su-In Lee, described the use of text and phenotypic data in medical records for predicting patients’ clinical course, and potential uses for medical management. The presentation was followed by questions and answers about the talk, and discussion focused on lessons learned, opportunities to collaborate, and sharing of data.

The MaLMIC fall forum, hosted by Dr. Lincoln Stein, focused on human sample and data resources for machine learning research. Dr. Dianne Chadwick described the Ontario Tumour Bank (OTB)’s repository of human cancer biospecimens and de-identified clinical data. Following this, Dr. Philip Awadalla provided an overview of the Ontario Health Study (OHS) and described research from his lab using simulation-based and machine learning tools to analyze evolution in non-recombining systems and the use of liquid biopsies to identify genetic and epigenetic markers indicative of cancer development. The presentations were followed by questions and answers about the talks, and discussion focused on lessons learned, opportunities to collaborate, and the availability of data and biosamples from OTB and the OHS.

Artificial Intelligence (AI) applied to medical images has the potential to transform patient care, especially in radiation oncology where radiological imaging is ubiquitous and treatment planning is time consuming.
AI models are data hungry but medical data are often insufficient for many applications. Building foundational models and fine-tuning them for specific medical tasks with only a few data points represents a new avenue of research with high translational potential. However, there is a lack of large compendia of radiological data that are sufficiently curated for building these foundational models.
In this session, the presenters described current work on implementing Med-ImageNet, a compendium of radiological images that have been standardised to develop deep learning models, such radiomics predictors or the foundational MedSAM model.

For the May forum on machine learning in computational pathology, the first presentation by April Khademi, a researcher, was about the past, present and future of digital pathology and AI. The second presentation by Phedias Diamandis, a pathologist, was about approaches to solve implementation challenges of AI in the real world. The talks were followed by discussion focused on lessons learned, opportunities to collaborate, and sharing of resources.

For the April forum on machine learning in precision health, the first presentation by Ross Mitchell, a Professor of Medicine and AI scientist at the University of Alberta, was about AI and Precision Health. The second presentation by Jacob Jaremko a radiologist, entrepreneur and AI scientist, was about AI augmented ultrasound imaging. The talks were followed by discussion focused on lessons learned, opportunities to collaborate, and sharing of resources.