AI Device Effectively Identifies Skin Lesions in Study

In January, after the DermaSensor became the first artificial intelligence (AI)-enabled device cleared by the US Food and Drug Administration (FDA) to evaluate lesions suggestive of skin cancer, the potential of the device to bridge a gap in access to dermatologists drew significant attention.

In a letter published in npj Digital Medicine, for example, Stephen Gilbert of the Else Kröner Fresenius Center for Digital Health at Dresden University of Technology, Dresden, Germany, and coauthors from Harvard University, Boston, postulated that the use of the DermaSensor, a handheld device that uses elastic scattering spectroscopy (ESS) to identify skin lesions concerning skin cancer, “could strengthen the diagnostic abilities of PCPs [primary care physicians] who normally refer such cases, and could therefore meaningfully address access limitations in dermatology, which affect more than a third of patients with an average wait time of 35 days.”

Since FDA clearance of DermaSensor was announced in January 2024, results from a single-center comparative effectiveness study have been published, demonstrating a sensitivity of 90% and a specificity of 60.7%.

“These results suggest that the use of the handheld ESS device by PCCs [primary care clinicians] can significantly improve diagnostic and management sensitivity for cancerous skin lesions,” senior author Nathalie C. Zeitouni, MD, a Phoenix-based Mohs and reconstructive surgeon, and colleagues wrote in the study, which was published in May in the Journal of the American Board of Family Medicine. “Furthermore, the accuracy of the device in correctly identifying benign but suspicious skin lesions indicates that its usage could decrease needless dermatology referrals and reduce patient concerns for malignancy.”

Developed by DermaSensor, the DermaSensor device operates on battery power, uses spectroscopy and algorithms to evaluate skin lesions for potential cancer in a matter of seconds, and is indicated for evaluating skin lesions suggestive of melanoma, basal cell carcinoma, and/or squamous cell carcinoma in patients aged 40 or over to help healthcare providers determine whether to refer a patient to a dermatologist, according to the FDA’s January 2024 announcement.

After the device completes the scan of a lesion, a result of “investigate further” (positive result) suggests further evaluation through a referral to a dermatologist, while “monitor” (negative result) suggests that there is no immediate need for a referral to a dermatologist.

In a pivotal trial of the device that evaluated 224 high-risk lesions at 18 primary care study sites in the United States and four in Australia, the device had an overall sensitivity of 95.5% for detecting malignancy.

For the current study, three PCCs in a rural setting — a board-certified family physician, a board-certified internist, and a board-certified nurse practitioner — evaluated skin lesions reported by patients as concerning and scanned them with the DermaSensor. The comparison for performance of the PCCs and the device were pathology results when available for review by a panel of three dermatologists who examined high-resolution dermatoscopic and clinical images. Evaluation of results included sensitivity, specificity, negative predictive value (NPV), and positive predictive value.

The final analysis included 178 lesions and 155 patients. Zeitouni and colleagues reported that the DermaSensor had a sensitivity of 90.0% and a specificity of 60.7% compared with the PCCs’ standard-of-care management sensitivity of 40% and a specificity of 84.8% without the device. For pigmented lesions, the device demonstrated a specificity of 76.9%, showing strong concordance with the panel of dermatologists. Also, the NPV with the device was 98.9%, indicating its reliability in negative “monitor” results correctly being diagnosed as benign.

According to Vishal A. Patel, MD, a Mohs micrographic surgeon who is director of cutaneous oncology at the George Washington University Cancer Center, Washington, DC, who was asked to comment on the study, there is limited evidence that asymptomatic skin cancer screening provides meaningful impact on the morbidity and mortality related to skin cancer. “As a result, the United States Preventive Services Task Force does not recommend widespread asymptomatic skin cancer screening for the general population,” Patel said.

“What has been shown to potentially help is for primary care providers to be trained to risk stratify patients and/or perform initial triage of potentially suspicious lesions and then refer for evaluation by a dermatologist,” he added. “This can help address the ongoing shortage of dermatologists for the US population and the lack of access for patients, especially in rural areas.”

The DermaSensor device “can potentially help bridge and empower primary care physicians to perform a first-pass triage to better identify those lesions that do not require specialist evaluation compared to those that might,” he said.

However, Patel advised caution in the use of diagnostic support tools that use machine learning and AI to improve medical care and raised two main concerns. “First, the algorithms developed do not discriminate between in situ and invasive lesions or low-risk and high-risk histologies, [and] the impact of this on referrals and the overall skin cancer epidemic is unclear,” Patel said. Second, he added, “by lowering the bar and threshold of identification of potentially low-risk lesions that may not require aggressive treatment and therapy, we may inadvertently lead to overdiagnosis and overtreatment, especially in vulnerable patient populations, such as the elderly. Some of these lesions could be preventively treated with topical or preventative therapies but may now be pushed towards biopsy if a device is advocating specialist evaluation.”

The study was funded with a grant from DermaSensor. Zeitouni and Miguel Tepedino, MD, were paid investigators for the study. Patel reported having no relevant disclosures.