July 27, 2017 • DNA
The U.S. Food and Drug Administration cleared for marketing a diagnostics system aided by machine learning algorithms to help radiologists detect breast cancer. The system, known as QuantX, is developed by Quantitative Insights Inc., a spin-off enterprise from University of Chicago. The company says QuantX is the first computer-aided diagnostics system that uses machine learning to evaluate breast abnormalities.
MRI of breast (National Cancer Institute)
QuantX, says the company is a decision-support system for radiologists evaluating magnetic resonance images (MRIs) of breast tissue suspected of cancer. The system has software with analytics that employ machine learning algorithms to evaluate the patient’s images and medical data with a database of abnormalities from documented clinical pathology findings, producing an index value called a QI score. The company says turnaround times for diagnostics are improved, and results of the analysis add to the confidence of the radiologists’ assessments.
FDA cleared QuantX for known breast cancer cases needing additional analysis or assessment, such as those considered at high risk. For complex cases, with multiple abnormalities, says FDA, QuantX can be used to evaluate each abnormality individually. The system can also be used as an image viewer to assess multiple images, including images from technologies other than MRIs, including ultrasound and mammography. However, QuantX is not cleared for primary interpretation of digital mammography images.
QuantX’s clearance used FDA’s de novo classification pathway for new devices rated of low to moderate risk. FDA rates QuantX as a Class 2 device, of moderate risk requiring a reasonable assurance of safety and effectiveness.
The agency’s clearance is based in part on results from a clinical study of Quant X by Quantitative Insights among 19 experienced radiologists reading MRI images of breast tissue and distinguishing between cancerous and non-cancerous samples. The study shows QuantX improves the performance of radiologists in interpreting the images, compared to conventional commercially-available tools. Among the findings, the number of false negatives — missed cases of cancer — was reduced by 39 percent among QuantX users.
FDA’s clearance allows Quantitative Insights to begin marketing QuantX, subject to additional documentation provided to FDA on the algorithms, data sets, and software testing routines, as well as preparation of labels, including warning statements.
Quantitative Insights was founded in 2010 to commercialize the research of Maryellen Giger, professor of radiology and medical physics, and colleagues at University of Chicago. Giger is considered a pioneer in computer-assisted diagnostics, including quantitative image analysis and data-mining in breast imaging. She is currently a scientific advisor to the company.
Originally posted in Science & Enterprise