Dermatologists, Patients Divided on Augmented Intelligence for Melanoma Screening

Published on: 

This analysis explored views on safety, general levels of trust, and preferences on examination methods, with the goal being a greater awareness of opinion on AI incorporation.

Dermatologists differ in their views on use of artificial or augmented intelligence for skin cancer screenings, according to new findings, with patients preferring 3D-total-body photography (TBP) and clinicians trusting their own decision-making during screenings.1

These study results were the product of a new study conducted to assess the subjective experiences of dermatologists and their patients following screening for skin cancer. The screenings involved the implementation of commercially-available 2D/3D TBP systems and convolutional neural network (CNN)-classification.

The study’s investigators acknowledged prior research on the diagnostic accuracy of such innovations, though they noted that patient and clinician acceptance remained less well-known and that real-world evidence had been lacking prior to this study.2 This research was led by Elisabeth Victoria Goessinger, from the department of dermatology at the University Hospital Basel in Switzerland.

“Our prospective data are unique, because patients and physicians undergo all three diagnostic methods (human vs. artificial vs. augmented intelligence) and immediately share their comparative experiences,” Goessinger and colleagues wrote.

Background and Methods

The investigators conducted their research at the University Hospital of Basel, using a prospective, single-center observational cohort study design. Their research was done from January - December 2021, with the team recruiting 205 subjects and 8 dermatologists for their work.

The research team’s work was considered part of the ongoing ‘Melanoma detection with 3D Vectra in Switzerland’ (MELVEC) research done at the hospital. The goal of MELVEC was to evaluate the augmented intelligence’s efficacy with 2D/3D TBP systems and CNN classification of the patients’ dermoscopic images, with the team seeking to identify skin cancer among patients considered to be at elevated risk.

Subjects were given screenings for melanoma that were carried out by the recruited dermatologists, and this was later followed by 2D and 3D TBP. Suspicious lesions on participants’ skin, especially melanocytic lesions which measured ≥3 mm, were digitally imaged through the use of dermoscopes.

Investigators then evaluated the lesion images through the use of CNNs. The determinations made by clinicians on the excisions were then based upon increased CNN risk scores as defined by MELVEC, clinicians’ suspicions of skin cancer, and/or indications of melanoma risk due to the results of augmented intelligence.

The investigators then had both the subjects and the dermatologists take surveys to record their feedback on the procedures. These surveys involved an evaluation of the subjective safety score (SSS) with a range from 0 - 10, assessing subjects’ views on safety at the different examination procedures.


The research team reported, overall, that 95.5% of subjects were found to view artificial intelligence as enhancing diagnostic accuracy. The team added that 83.4% were shown to have preferred AI-based screenings of melanoma over evaluations implementing AI only or dermatologists only (3D-TBP: 61.3%; 2D-TBP: 22.1%).

In their assessment of participants’ views on safety, the investigators found that AI-assisted assessments had scores which were far higher than AI exclusively (mean-SSS (mSSS): 9.5 versus 7.7, P < .0001). This system was also shown to be ranked slightly higher than dermatologist-only assessments (mSSS: 9.5 versus 9.1, P = .001).

Majority percentages of clinicians (3D-TBP: 90.2%; 2D-TBP: 96.1%) were identified by the research team as having expressed high confidence in AI evaluation results. They also found that 68.3% of the study’s dermatologists noted that there had been improved diagnostic accuracy when implementing AI-assistance, especially for beginners (61.8%) as opposed to experts (20.9%).

Despite dermatologists’ confidence in the use of AI, only 1.5% were shown to have favored a benign CNN-classification above personal suspicion of malignancy in risk evaluations.

“Our findings underline the importance of (augmented intelligence) in melanoma recognition, in the sense that (augmented intelligence) should serve as a supportive tool for physicians to provide best possible patient care,” they wrote. “Based on our results, we advocate to perform (augmented intelligence) examinations for high-risk patients in routine practice and expect that centres offering 3D TBP will have a competitive advantage in the future.”


  1. Goessinger EV, Niederfeilner JC, Cerminara S, et al. Patient and dermatologists' perspectives on augmented intelligence for melanoma screening: A prospective study. J Eur Acad Dermatol Venereol. Published online February 27, 2024. doi:10.1111/jdv.19905.
  2. Nelson CA, Pérez-Chada LM, Creadore A, Li SJ, Lo K, Manjaly P, et al. Patient perspectives on the use of artificial intelligence for skin cancer screening: a qualitative study. JAMA Dermatol. 2020; 156(5): 501–512.