A team of U.S. researchers accurately trained a deep learning convolutional neural network (CNN) to classify three common dermatopathology diagnoses, according to recent research published in the Journal of Pathology Informatics.
“The combined effect of the introduction of deep learning, the democratization of powerful graphics processing unit (GPU) computing capacity, and increasing acceptance and use of digital pathology has created an unprecedented opportunity to explore the power of deep learning,” wrote lead author Thomas George Olsen, with Wright State University School of Medicine in Ohio, and colleagues.
With that in mind, Olsen et al. broke down their research into three consecutive studies. In the first, they scanned 200 previously diagnosed nodular basal cell carcinoma (BCC) slides and 100 distractor glass slides into a whole slide imaging (WSI) system.
Researchers in the second study then scanned 125 previously diagnosed dermal nevi glass slides along with 100 detractors. Finally, in the third study, Olsen and colleagues entered 125 diagnosed seborrheic keratoses slides with 100 additional distractors. Each training set consisted of unannotated whole slide images with five common neoplastic and inflammatory distractor diagnoses, according to the study.
The artificial intelligence platform accurately classified more than 99 percent (123 of 124) of BCCs, more than 99 percent (113 of 114) of dermal nevi and 100 percent of seborrheic keratoses.
“Results from this study provide proof of concept that can serve as a framework for refinement and expansion of algorithmic development for common diagnoses in a dermatopathology laboratory,” Olsen et al. wrote.
Implementing an algorithm, such as the one described in this study, into digital pathology workflows could not only improve patient care, the authors argued, but also reduce healthcare costs.
“Integration of refined and fully developed computer algorithms into digital pathology workflow could facilitate efficient triaging of cases and aid in diagnosis, resulting in significant cost savings to the health-care system,” the researchers concluded.