Breast cancer is the most commonly diagnosed type of cancer globally, according to the World Health Organization, which has estimated that the case count could climb from about 2.3 million new diagnoses in 2020 to more than 3 million in 2040.
To address that growing prevalence, many artificial intelligence developers are building algorithms capable of detecting breast cancer as early as possible, when it’s still highly treatable. Many have already received FDA nods for their work, including, most recently, MedCognetics.
The Dallas-based startup announced the clearance of its QmTriage AI platform this week, which sets up the breast cancer screening software for a U.S. rollout.
The software takes in standard 2D mammograms, and its machine learning algorithms analyze the images for potential signs of cancer. If any are found, the system labels them and automatically flags the scan for closer review by a human radiologist.
MedCognetics developed the AI in partnership with the University of Texas Southwestern (UTSW) Medical Center, which also holds an equity stake in the startup. The algorithms were trained using deidentified mammography scans from UTSW patients. MedCognetics paid specific attention to including a diverse group of patients in that training set, CEO Debasish Nag said in this week’s announcement, ensuring that the AI could be used to accurately screen for cancer across patients of all races and ethnicities.
In study results submitted (PDF) to the FDA for clearance of the software, MedCognetics concluded that QmTriage achieved sensitivity of 87% and specificity of 89% in identifying signs of breast cancer in a test set of 800 mammograms, about half of which had been proven positive for breast cancer with a follow-up biopsy.
“Our software’s high detection accuracy enables reduced time for review by radiologists,” Nag said, adding that with the FDA’s OK for the breast cancer-focused AI under its belt, MedCognetics will now “work toward expanding to other realms of cancer.”
QmTriage is far from the only AI-powered option available to radiologists to speed up the process of screening for breast cancer. The first such system to be cleared by the FDA was Qlarity Imaging’s QuantX, which got the go-ahead in 2017 and uses machine learning AI to look for irregularities in breast scans.
Meanwhile, QmTriage’s application for 510(k) clearance compared it to the precedent set by Zebra Medical’s HealthMammo. That tool was cleared in 2020 and works similarly to MedCognetics and Qlarity’s offerings, using its own AI to automatically flag suspicious lesions in 2D mammograms. RadNet, iCad, Koios Medical, Paige and even Google’s DeepMind subsidiary have also thrown their hats in the ring, with all developing new AI models that look for signs of the disease in X-rays, ultrasounds, pathology slides and more.