making strong clinical predictions about DCIS patient outcomes
In the United States, 60,000 women per year are diagnosed with ductal carcinoma in situ (DCIS). DCIS is cancer in which the cells that line the breast’s milk ducts have become cancerous, but they have not spread to the surrounding breast tissue. There are both aggressive and slowly progressing forms of DCIS, but current technologies cannot identify which form of DCIS is present in the tissue sample. This means physicians working with DCIS patients cannot predict cancer recurrences or non-recurrences during the early stages of the disease, so all patients are treated as if they have invasive disease. This leads to overdiagnosis, overtreatment, and high financial costs without a clear benefit.
University of Michigan’s Howard Petty, Ph.D., has developed a technology that uses an innovative biomarker kit, imaging methods, and a machine learning database to determine if the patient matches recurrent or non-recurrent classifications of DCIS.
“By predicting the potential for cancer recurrences, our biomarker kit and diagnostic methods benefit both physicians and patients in the development of an evidence-based cancer management plan,” explains Petty. “Our company analyzes tissue samples provided by the pathologists, and we are looking at ways to streamline this intake process, including the use of cloud-based software-as-a-service diagnostic tests for pathology departments with advanced imaging abilities.”
Significant Need
Current technologies cannot identify which pre-invasive DCIS lesions are slowly progressing and which will lead to life-threatening disease. Additionally, some diagnostic kits on the market are woefully insufficient and are generally not prescribed, indicating a need for a more precise and readily available diagnostic.
Compelling Science
This procedure uses biomarkers to identify images of elements contributing to disease instead of identifying the disease. It combines high information content and high signal-to-noise ratio images with machine learning to sort through images using the latest computer algorithms to perform binary classifications of DCIS images from patients who will or will not experience a cancer recurrence. Each time the process is successfully repeated on a new biomarker, another step in the mechanism of cancer recurrence is revealed. The machine learning reduced false negatives.
Competitive Advantage
The biomarker test provides clinically actionable findings of high accuracy and helps make strong clinical predictions about DCIS patient outcomes and cancer recurrences, unlike the kits currently on the market that have low signal-to-noise ratios, errors when combining multiple biomarkers, and the lack of spatial expression information.
Overall Commercialization
- Intellectual Property: Provisional patent has been filed
- Commercialization Strategy: Start-up company
- Regulatory Pathway: FDA pre-market approval
- Engage Investors: Obtain venture funding and FDA approval, while follow-on funding will be STTR, SBIR, and VC funding.
- Product Launch Strategy: Initially address unmet need in the ability to predict cancer recurrences or non-recurrences during early stages of DCIS of the breast. Could extend to other cancers.
Milestones
- Increase N to publish findings and develop data for FDA improve model
- Test past data to identify the best stromal biomarker
- Assemble stromal dataset for solid DCIS
- Test new model for solid DCIS using holdout data
- Publish findings
- Improve computer methods; reduce cost