14-layer CNN accurately predicts breast cancer molecular subtype

A 14-layer convolutional neural network (CNN) trained on MRI and pathology data accurately predicted the molecular subtype of breast cancers, according to a Jan. 31 study published in the Journal of Digital Imaging. The method may help personalize treatment plans for the disease.

Trained on available breast MRIs and immunohistochemical (IHC) staining pathology data of 216 patients with known breast cancer, the CNN predicted breast cancer subtype with 70 percent accuracy. A balanced holdout set of 40 patients was used as the testing set.

“Despite advantages with the use of IHC surrogates, the range of agreement between its use in predicting breast cancer subtype and explicit genetic testing is between 41 and 100 percent,” the authors wrote. “Given the wide spectrum of prognosis and indicated treatment strategies based on tumor subtype, a need exists for more accurate diagnosis to aid in an individualized treatment plan.”

Many prior studies, which were cited in the current research, are limited by a reliance on semi-automated feature extraction, argued first author Richard Ha, MD, with Columbia University Medical Center’s Department of Radiology in New York, and colleagues. This limits the function of such methods because they depend on accurate human extraction of crucial features, the authors added.

“In contrast, CNN algorithms are trained in a manner that allows automatic extraction of features from the input feed that are crucial to the defined problem domain,” Ha et al., wrote. “This process improves its ability to study the input features in an end-to-end manner, using complex, stacked layers to predict a desired output. Therefore, CNN feature extraction is not a variable with each new MRI and thus results are consistent.”

The authors noted a few limitations of their study, including their use of an IHC surrogate to define molecular subtype instead of genetic analysis. According to the team, defining molecular subtypes with genetic analysis is both costly and technically challenging.

However, they concluded using a CNN plus MRI data instead of gene expression profiling may be superior for personalized breast cancer risk stratification. With the addition of more data, their model will likely improve.

""

Matt joined Chicago’s TriMed team in 2018 covering all areas of health imaging after two years reporting on the hospital field. He holds a bachelor’s in English from UIC, and enjoys a good cup of coffee and an interesting documentary.

Around the web

The new technology shows early potential to make a significant impact on imaging workflows and patient care. 

Richard Heller III, MD, RSNA board member and senior VP of policy at Radiology Partners, offers an overview of policies in Congress that are directly impacting imaging.
 

The two companies aim to improve patient access to high-quality MRI scans by combining their artificial intelligence capabilities.