AI proves capable in analyzing kidney biopsy images
A research team with Boston University of Medicine has demonstrated artificial intelligence (AI) can analyze kidney biopsy images more accurately than traditional human methods.
The researchers published their findings online Jan. 10 in Kidney International Reports.
Renal biopsy samples were collected from 171 patients treated at Boston Medical Center. The images were analyzed by the team’s AI model—a convolutional neural network (CNN)—as well as human nephropathologists.
The CNN model was able to better predict kidney analysis compared to the specialists in six key classification tasks—most notably in recognizing the stage of chronic kidney disease.
“We have outlined the development of a CNN that at least matches the performance of a skilled nephropathologist’s ability to quantify the extent of kidney fibrosis,” wrote Vijaya B. Kilachalama with the University of Boston School of Medicine and colleagues.
The promise demonstrated by machine learning in this study may clear a path for the CNN model to be used in clinical workflow and in treating similar diseases in other organs, authors of the study noted.
“This framework can also be adapted to other organ specific pathologies focused on evaluating fibrosis, as well as image datasets developed using other histological staining protocols,” Kilachalama et al. wrote.
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.