RESUMO
There has been growing scientific interest in the research field of deep learning techniques applied to skin cancer diagnosis in the last decade. Though encouraging data have been globally reported, several discrepancies have been observed in terms of study methodology, result presentations and validation in clinical settings. The present review aimed to screen the scientific literature on the application of DL techniques to dermoscopic melanoma/nevi differential diagnosis and extrapolate those original studies adequately by reporting on a DL model, comparing them among clinicians and/or another DL architecture. The second aim was to examine those studies together according to a standard set of statistical measures, and the third was to provide dermatologists with a comprehensive explanation and definition of the most used artificial intelligence (AI) terms to better/further understand the scientific literature on this topic and, in parallel, to be updated on the newest applications in the medical dermatologic field, along with a historical perspective. After screening nearly 2000 records, a subset of 54 was selected. Comparing the 20 studies reporting on convolutional neural network (CNN)/deep convolutional neural network (DCNN) models, we have a scenario of highly performant DL algorithms, especially in terms of low false positive results, with average values of accuracy (83.99%), sensitivity (77.74%), and specificity (80.61%). Looking at the comparison with diagnoses by clinicians (13 studies), the main difference relies on the specificity values, with a +15.63% increase for the CNN/DCNN models (average specificity of 84.87%) compared to humans (average specificity of 64.24%) with a 14,85% gap in average accuracy; the sensitivity values were comparable (79.77% for DL and 79.78% for humans). To obtain higher diagnostic accuracy and feasibility in clinical practice, rather than in experimental retrospective settings, future DL models should be based on a large dataset integrating dermoscopic images with relevant clinical and anamnestic data that is prospectively tested and adequately compared with physicians.
RESUMO
BACKGROUND: Due to progressive ageing of the population, the incidence of facial lentigo maligna (LM) of the face is increasing. Many benign simulators of LM and LMM, known as atypical pigmented facial lesions (aPFLs-pigmented actinic keratosis, solar lentigo, seborrheic keratosis, seborrheic-lichenoid keratosis, atypical nevus) may be found on photodamaged skin. This generates many diagnostic issues and increases the number of biopsies, with a subsequent impact on aesthetic outcome and health insurance costs. OBJECTIVES: Our aim was to develop a risk-scoring classifier-based algorithm to estimate the probability of an aPFL being malignant. A second aim was to compare its diagnostic accuracy with that of dermoscopists so as to define the advantages of using the model in patient management. MATERIALS AND METHODS: A total of 154 dermatologists analysed 1111 aPFLs and their management in a teledermatology setting: They performed pattern analysis, gave an intuitive clinical diagnosis and proposed lesion management options (follow-up/reflectance confocal microscopy/biopsy). Each case was composed of a dermoscopic and/or clinical picture plus metadata (histology, age, sex, location, diameter). The risk-scoring classifier was developed and tested on this dataset and then validated on 86 additional aPFLs. RESULTS: The facial Integrated Dermoscopic Score (iDScore) model consisted of seven dermoscopic variables and three objective parameters (diameter ≥ 8 mm, age ≥ 70 years, male sex); the score ranged from 0 to 16. In the testing set, the facial iDScore-aided diagnosis was more accurate (AUC = 0.79 [IC 95% 0.757-0.843]) than the intuitive diagnosis proposed by dermatologists (average of 43.5%). In the management study, the score model reduced the number of benign lesions sent for biopsies by 41.5% and increased the number of LM/LMM cases sent for reflectance confocal microscopy or biopsy instead of follow-up by 66%. CONCLUSIONS: The facial iDScore can be proposed as a feasible tool for managing patients with aPFLs.