AI/ML-Enabled Longitudinal Tumor Tracking Radiology Viewer

Summary

Dr. Smith is a Professor and Vice-Chair of Innovation in Radiology at the University of Alabama in Birmingham. He was looking for a software-development team to help him create a workflow-focused longitudinal tumor tracking SaMD. 18 months after project kickoff I had an FDA cleared product and our first clinical installation. We iterated on the initial design, developed the software, handled the FDA 510(k) clearance (K202229) in 2020, and have continued building on the initial foundation. We’ve also helped them grow an in-house engineering team that we work alongside, and helped set up their initial QMS (which passed its first FDA audit).

Screenshot

About Dr. Andrew Smith

Dr. Smith is a Professor and Vice-Chair of Innovation in Radiology at the University of Alabama in Birmingham. He is also Co-Director of AI in Radiology, Chief of Abdominal CT, the Director of Clinical AI for the Heersink Institute for Biomedical Innovation, and Founder, Chief Medical Officer , and Chairman of the Board of AI Metrics Inc., a medical technology startup focused on augmented intelligence imaging solutions for radiologists. Smith is the inventor on 20 issued patents and has led FDA 510(k) clearance on 2 radiology technologies.

Challenges

• I was concerned about their rates—as they were higher than hiring an offshore firm or hiring my own engineers. However, investors are surprised at how much we have been able to accomplish with just $3M of funding. Their rates are well justified considering how much time and effort I saved on hiring costs, fringe benefits, and maintaining an effective engineering + regulatory culture.
• The team had never created a QMS before. However, we recently passed an FDA inspection with no major findings.
• I was concerned about vendor lock-in. However, they not only helped me hire my own software engineers, they also trained them and treat them as one of their own.
• I was concerned about not hiring a full time CTO. In retrospect, Innolitics filled in this role and helped me out a lot since I did not have to worry about making payroll during a funding pause.

Timeline

• January 2019. Strategy. We met with Andrew Smith in person to identify user needs and create UI mockups.
• October 2019. Prototype. We developed a prototype and deployed it at UAB for real world testing and user feedback. Our tech stack consisted of React, Django, and Tensorflow.
• November 2019. Trade Show #1. Innolitics manned the first booth at RSNA.
• March 2020. Trade Show #2. Innolitics manned booth at Society of Abdominal Radiology.
• May 20, 2020. Clinical Study #1. We assisted the publication of a research paper using the software we created.
• November 2, 2020. Seed round investment. Our prototype and expertise helped Andrew secure funding.
• July 2020. Version 1.0. We took user feedback and market signals to build the first version possible to get FDA cleared.
• September 9, 2020. 510(k) Submission. We prepared and submitted the 510(k) in 3 months. We did everything from presubmission meeting, writing+executing V&V protocols, IFU, Risk Management, Traceability, Cybersecurity, and FDA correspondence.
• December 22, 2020. 510(k) Cleared. FDA takes at least 13 weeks to clear, we were cleared in just 15. Clearance letter here.
• 2022. SaMD Team and Culture. We helped Andrew hire engineers and trained them on the software. We treated AI Metrics engineers as our own and fostered a collaborative SaMD culture.
• 2023. FDA Audit. FDA came knocking and the QMS that we helped set up passed with minimal findings.