(HealthDay News) — An artificial intelligence program outperforms pathologists for differentiating ductal carcinoma in situ (DCIS) from atypia, according to a study published online Aug. 9 in JAMA Network Open.

Ezgi Mercan, Ph.D., from the University of Washington in Seattle, and colleagues used 240 breast biopsies from the Breast Cancer Surveillance Consortium registries that varied by breast density, diagnosis, patient age, and biopsy type. The biopsies were categorized by three expert pathologists as benign, atypia, DCIS, and invasive cancer. The performance of automated analysis of two image features (tissue distribution feature and structure feature) from high-resolution digital slide images was compared to independent interpretations from 87 practicing U.S. pathologists.

The researchers found that the accuracy of machine learning tissue distribution features, structure features, and pathologists for classification of invasive cancer versus noninvasive cancer was 0.94, 0.91, and 0.98, respectively. The accuracy for classification of atypia and DCIS versus benign tissue was 0.70, 0.70, and 0.81, respectively, and the accuracy for classification of DCIS versus atypia was 0.83, 0.85, and 0.80, respectively. For the invasive versus noninvasive classification, the sensitivity of both machine learning features was lower than that of the pathologists (tissue distribution feature, 0.70; structure feature, 0.49; pathologists, 0.84), but it was higher for the classification of atypia and DCIS versus benign cases (tissue distribution feature, 0.79; structure feature, 0.85; pathologists, 0.72) and the classification of DCIS versus atypia (tissue distribution feature, 0.88; structure feature, 0.89; pathologists, 0.70). The specificity of the machine learning feature classification was similar to that of the pathologists for the DCIS versus atypia classification (tissue distribution feature, 0.78; structure feature, 0.80; pathologists, 0.82).

Continue Reading

Related Articles

“The findings suggest that machine learning methods are potentially suitable as diagnostic support systems in differentiating challenging preinvasive lesions of the breast,” the authors write.

Abstract/Full Text