MD, PhD, FSIIM
Radiologist and researcher working in medical imaging, artificial intelligence, and clinical workflow automation.
Research and implementation work in radiology artificial intelligence, workflow automation, and the evaluation of clinical AI systems.
Bradley J. Erickson, MD, PhD, is a Professor of Radiology at Mayo Clinic, Director of the Mayo Clinic AI Lab, and CEO of FlowSigma. He holds MD and PhD degrees from Mayo Medical and Graduate School and has conducted research in medical imaging informatics and artificial intelligence for over three decades, during which he has mentored more than 80 trainees.
His work spans clinical practice, research, and implementation, including direction of Mayo Clinic's transition to filmless and paperless operations and the development of deep learning methods for diagnostic imaging.
Awards received for contributions to medical imaging informatics and artificial intelligence.
Author of more than 200 peer-reviewed publications on medical imaging AI, bias mitigation, and clinical informatics. Selected publications are listed below.
The objective of this work is to reduce the distance between AI research and clinical implementation. A substantial proportion of promising AI tools fail at deployment, and in most cases the limiting factor is not algorithm performance but the absence of adequate attention to clinical workflow, calibrated clinician trust, and patient safety.
MDSynapse.org is intended to support clinicians in evaluating these systems critically: identifying what a reported result does and does not establish, requiring transparency regarding training data and stratified performance, and distinguishing tools that improve patient outcomes from those that improve benchmark metrics.
At FlowSigma, the corresponding work is infrastructural: embedded workflows, automated quality control, and systems designed around observed clinical practice rather than an idealized process model. The objective is not to substitute for clinical judgment but to provide reliable tools that reduce the administrative and verification burden placed on it.
LLMs are used in creating content. My process is that I identify interesting topics or papers, I create an outline, I use 2 LLMs to draft content and a 3rd LLM to combine the drafts, and then I hand edit to produce the final product.
Inquiries regarding medical AI research, workflow automation, and collaboration are welcome.