Risk Classification and Prediction of Histopathological Subtypes in Basal Cell Carcinoma Using a CNN-Based Artificial Intelligence Model on Dermoscopic Images

Basal cell carcinoma (BCC) is the most common skin malignancy and comprises histopathological subtypes with different biological behaviors, recurrence risks, and treatment implications. Accurate identification of high-risk and low-risk subtypes is important for clinical decision-making. Dermoscopy improves diagnostic accuracy in BCC; however, prediction of histopathological risk categories based solely on dermoscopic findings remains challenging.

This retrospective observational study will use archived clinical and dermoscopic images, histopathology reports, and clinical records of patients with histopathologically confirmed BCC. All data will be anonymized before analysis. Images containing identifiable patient information will be excluded.

A convolutional neural network (CNN)-based artificial intelligence model will be developed using clinical and dermoscopic images. Images will undergo preprocessing, including standardization of image size, normalization procedures, and removal of potentially identifiable information. The dataset will be divided into training, validation, and test sets while maintaining separation at the patient level to avoid data leakage.

The primary outcome is the diagnostic performance of the CNN model for classification of BCC into low-risk and high-risk histopathological groups. Secondary outcomes include prediction of histopathological subtypes and comparison of model performance with dermatologist assessments. Histopathological diagnosis will serve as the reference standard.

Model performance will be evaluated using accuracy, sensitivity, specificity, precision, recall, F1 score, and area under the receiver operating characteristic curve (ROC-AUC). Comparisons between the artificial intelligence model and physician assessments will be performed using appropriate statistical methods. Interobserver agreement may also be assessed when applicable.