WEDNESDAY, May 3, 2023 -- Cooperation of dermatologists with a convolutional neural network (CNN) improves diagnostic performance for suspect melanocytic lesions, according to a study published online May 3 in JAMA Dermatology.

Julia K. Winkler, M.D., from the University of Heidelberg in Germany, and colleagues conducted a prospective diagnostic study in which dermatologists performed skin cancer screenings using naked-eye examination and dermoscopy. Suspect melanocytic lesions were graded by the probability of malignancy. Dermoscopic images of suspect lesions were then examined using a market-approved CNN. Reference diagnoses were based on histopathologic examination in 54.8 percent of lesions or on clinical follow-up and expert consensus in the case of nonexcised lesions.

Twenty-two dermatologists detected 228 suspect melanocytic lesions in 188 patients. The researchers found that when dermatologists additionally integrated CNN results into decision-making, diagnostic sensitivity and specificity improved significantly (mean sensitivity improved from 84.2 to 100.0 percent; mean specificity from 72.1 to 83.7 percent; mean accuracy from 74.1 to 86.4 percent; mean receiver operator characteristic area under the curve from 0.895 to 0.968). For classifying melanocytic lesions, CNN alone achieved comparable sensitivity, higher specificity, and higher diagnostic accuracy than dermatologists alone. Furthermore, unnecessary excisions of benign nevi were reduced by 19.2 percent with dermatologist cooperation with CNN. The most diagnostic improvement was seen for dermatologists with less dermoscopy experience cooperating with the CNN compared with more experienced dermatologists.

"These results indicate that a broader application of this human with machine approach, particularly in nonspecialized institutions, could be beneficial to clinicians and patients," the authors write.

Two authors disclosed financial ties to the pharmaceutical and medical device industries.