ML-CDS 2022: Multimodal Learning and Fusion Across Scales for Clinical Decision Support

Agenda


8:00-8:10 (SGT time) Welcome notes

8:10-8:50 (SGT time) Invited talk: "Multimodal AI in Medicine: Potential Areas of Application" by Dr. John Chang, Banner Health

Abstract: In 2016, many renowned scientists predicted that AI will soon replace radiologists and pathologists because of the routine, repetitive nature of diagnosis. Unbeknownst to these renowned scientists is the fact that diagnosis is a combination of pattern matching (which AI does extremely well), information integration (which AI can do well if properly trained), and intuition (which AI fails at). AI has excelled at well-defined tasks such as scheduling, imaging protocol selection, image reconstruction, organ segmentation, lesion tracking, and information gathering. These AI applications have certainly improved the efficiency of clinical practice. In order for further AI integration into clinical practice, the algorithms must make use of multiple modes of input (history, laboratory work, imaging, and pathology) and potentially produce multiple modes of output (clinical notes, prognosis, and treatment prediction). While the smart AI (no pun intended) application may seem to replace the physician, the physician and patient together will ultimately determine whether the presented choices are suitable for the patient based on his/her health and economic status or perhaps explore alternative such as clinical trials. Ultimately, this is humanity at its best: exploring the uncertainties of human disease with empathy.

8:50-10:20 (SGT time) Paper Presentation (30 minutes/talk)

Visually Aware Metadata-guided Supervision for Improved Skin Lesion Classification using Deep Learning
Anshul Pundhir, Ananya Agarwal, Saurabh Dadhich, and Balasubramanian Raman

Predicting Osteoarthritis of the Temporomandibular Joint Using Random Forest with Privileged Information
Elisa Warner, Najla Al-Turkestani, Jonas Bianchi, Marcela Gurgel, Lucia Cevidanes, and Arvind Rao

Hybrid Network Based on Cross-modal Feature Fusion for Diagnosis of Alzheimer's Disease
Zifeng Qiu, Peng Yang, Tianfu Wang, and Baiying Lei

10:20-11:00 (SGT time) Invited Talk: TBD

11:00-11:30 (SGT time) Panel Discussion



Keynote Speaker

Italian Trulli

John Chang, Radiologist, Banner Institute

Dr. Chang received his BS degree in EECS from UC Berkeley, MS and PhD degrees in ECE from UI Urbana-Champaign, and MD from UI College of Medicine at Urbana. After his obtaining his degrees, he then completed his radiology training at Stanford University. After a 6 month stint as clinical instructor at Stanford, Dr. Chang joined Banner Health in 2011, which partnered with the MD Anderson Cancer Network in late 2011. Dr. Chang has overseen the growth of body imaging since 2012 for Banner MD Anderson Cancer Center in Gilbert, AZ. He has also worked with collaborators at Arizona State University on developing MR spectroscopy for assessing cancer metabolism and nanoparticle solutions for radiation detection. His research also focused on decreasing radiology errors for improving patient outcome, which led to his research on artificial intelligence as potential second observer. He envisions that artificial intelligence will best serve as assistants to physicians rather than replacing them and that physicians will perform better in outcome and efficiency with AI assistance than without. In his keynote address, he will explore potential areas where AI assistance can be of great help to the practicing clinicians in the general medical clinic and for improving patient clinical outcome.