AI Device Enhances Skin Cancer Diagnosis in Primary Care

Medscape

a day ago

An artificial intelligence (AI)–enabled handheld elastic scattering spectroscopy (ESS) device improved the sensitivity for skin cancer diagnosis and management by primary care physicians (PCPs), with better overall diagnostic accuracy.
METHODOLOGY:
ESS is a sampling technique that distinguishes between benign and malignant tissue without surgical biopsy and the ESS device is cleared by the US Food and Drug Administration for use by nondermatologist expert physicians. An AI-enabled handheld ESS device was developed to aid PCPs in the management of suspicious skin lesions.
This companion study of the DERM-SUCCESS study conducted across 22 primary care sites in the United States evaluated the impact of the device on PCPs' diagnostic performance in the detection and management of skin cancer.
The study involved 108 PCPs (65.7% men) who assessed 50 skin lesion cases (25 benign and 25 malignant) from 50 patients (median age, 59 years; 100% White individuals) in two phases: First, with visual assessment without using the handheld device, followed by an assessment using the device, separated by a 2-hour break. For each lesion, they provided a diagnosis, management decision, and level of confidence in their assessment.
The validated device emitted light pulses over specific areas of a lesion to analyze lesion structure and classified lesions as 'Investigate Further' or 'Monitor,' with a 1-10 score reflecting the degree of similarity to malignant lesions.
The coprimary endpoints were (1) the referral sensitivity of physicians, a measurement of the ability to appropriately refer malignant cases for further evaluation (management sensitivity) and (2) the referral sensitivity and specificity of physicians who knew the device results. Biopsy results served as the reference standard for both.
TAKEAWAY:
Management sensitivity was higher for physicians after using the AI-enabled device compared with when they did not have the device results (91.4% vs 82%; P = .0027). PCPs' diagnostic sensitivity also increased from 71.1% without the use of the AI device to 81.7% with the use of the AI device ( P = .0085).
= .0027). PCPs' diagnostic sensitivity also increased from 71.1% without the use of the AI device to 81.7% with the use of the AI device ( = .0085). Device-aided management and diagnostic specificity decreased, though not significantly. The use of the device led to 11.8% more benign lesions being incorrectly referred but also led to 9.4% more malignant lesions being correctly referred.
In physician management decisions, the proportion of high-confidence assessments increased from 36.8% without the use of the device to 53.4% after the use of the device, and the observed area under the curve performance improved from 0.708 to 0.762.
PCPs using the device were better at detecting malignant lesions than by chance (odds ratio, 6.8; P < .0001). Most PCPs reported benefits from the device, including improved confidence in clinical assessments and management decisions.
IN PRACTICE:
'The findings demonstrate that utilization of the ESS device output may improve physician performance in the management of suspicious lesions and referral of skin cancer to ensure timely diagnosis,' the authors of the study wrote.
SOURCE:
This study was led by Laura K. Ferris, of the Department of Dermatology at The University of North Carolina at Chapel Hill. It was published online on May 30, 2025, in Journal of Primary Care & Community Health .
LIMITATIONS:
The device could not directly affect how PCPs managed lesions owing to the study design because the device was experimental. Clinical information was provided, but the PCPs could not perform hands-on evaluations, which did not reflect real clinical practice but aligned with telemedicine. The study included only White patients, which limited the results for the assessment of different skin types.
DISCLOSURES:
This study was funded by DermaSensor, Inc. The authors declared no relevant conflicts of interest.