Pigmented skin lesion biopsies after computer-aided multispectral digital skin lesion analysis

Abstract

Background: The incidence of melanoma has been rising over the past century. With 37% of patients presenting to their primary care physician with at least 1 skin problem, primary care physicians and other nondermatologist practitioners have substantial opportunity to make an impact at the forefront of the disease process. New diagnostic aids have been developed to augment physician analysis of suspicious pigmented skin lesions (PSLs). Objective: To determine the effects of computer-aided multispectral digital skin lesion analysis (MSDSLA) on dermatologists’ and nondermatologist clinicians’ decisions to biopsy suspicious PSLs after clinical and dermatoscopic evaluation. Methods: Participants were shown 6 images of PSLs. For each PSL, participants were asked 3 times if they would biopsy the lesion: first after reviewing a clinical image of the PSL, again after reviewing a high-resolution dermatoscopic image, and again after reviewing MSDSLA probability findings. An answer was right if a melanoma or high-risk lesion was selected for biopsy or a low-risk lesion was not selected for biopsy. An answer was wrong if a melanoma or high-risk lesion was not selected for biopsy or a low-risk lesion was selected for biopsy. Clinicians’ decisions to biopsy were evaluated using χ2 analysis for proportions. Results: Data were analyzed from a total of 212 participants, 177 of whom were dermatologists. Overall, sensitivity of clinical image review was 63%; dermatoscopic image review, 5%; and MSDSLA, 83%. Specificity of clinical image review was 59%; dermatoscopic image review, 40%; and MSDSLA, 76%. Biopsy decision accuracy was 61% after review of clinical images, 52% after review of dermatoscopic images, and 80% after review of MSDSLA findings. The number of lesions participants indicated that they would biopsy increased significantly, from 52% after reviewing clinical images to 63% after reviewing dermatoscopic images (P