Collaborative Research Project 9
Principal investigators: Samuel Armato (University of Chicago), Lubomir Hadjiski, (University of Michigan), and Karen Drukker (University of Chicago)
Radiomics & machine intelligence for disease detection and diagnosis on chest radiographs and thoracic CTs.
Updated January 19, 2024
Members
Sam Armato, PhD (lead), University of Chicago, Karen Drukker, PhD, University of Chicago, Lubomir Hadjiiski, PhD University of Michigan, Mena Shenouda, University of Chicago, Emily Townley, AAPM
This collaborative research project is currently working to help organize and host MIDRC-sponsored data science Challenges on the detection, diagnosis, and prognosis of lung disease. It coordinates cross-MIDRC efforts involving all 3 organizations (AAPM, RSNA, and ACR) in this effort. The aim of these Challenges is to ultimately benefit the community at large through investigation and combination of machine intelligence algorithms' output for a better understanding of the imaging manifestations of, and impact of, various lung diseases. Rigorous evaluation will facilitate translation to clinical practice which will enable the urgent goals of MIDRC and could lead to innovative new technology. CRP9 and the MIDRC Grand Challenges Working Group successfully organized MIDRC’s first challenge in 2022, the COVIDx challenge, which was a diagnosis challenge based on portable chest X-ray radiographs. Code/algorithms associated with this challenge are available on GitHub and a manuscript is in preparation.
Moreover, CRP-9 is also actively developing its own machine learning/AI algorithms for the detection of pneumonia on chest radiographs and chest CT. Results have been presented at several conferences and manuscripts are in preparation.
Current plans include
Develop machine intelligence for assisting in detection and diagnosis of disease on other modalities such as MRI and ultrasound images, especially in other organ systems such as the brain and cardiac systems.
Host data science challenges on the detection and diagnosis of various disease types on MRI and ultrasound images, especially of the brain and cardiac systems.
Rigorously evaluate and rapidly translate these findings to clinical practice to enable the urgent goals of MIRDC.