Early diagnosis of skin cancer significantly improves patient prognosis and enhances survival rates. However, traditional diagnostic methods may be subject to variability due to the involvement of multiple diagnosticians, such as dermatologists. This paper evaluates the performance of convolutional neural networks for skin cancer detection using single-lesion crops derived from the provided TBP dataset. Furthermore, whole-brain 3D TBP scans are utilized to develop an image analysis pipeline aimed at excluding irrelevant lesions, followed by the construction and evaluation of the CNN model. The proposed network architecture is designed to effectively differentiate between malignant and benign lesions. Model performance is assessed using key evaluation metrics, including accuracy, sensitivity, specificity, and the AUC-ROC curve. According to the results of this study, the CNN model developed herein can accurately distinguish between skin lesions and is promising to be incorporated into the computer-aided diagnosis system of skin cancer by dermatologists. In this paper, it will be necessary to examine the need for advancing deep learning in relation to 3D imaging, something that may completely redefine the current paradigm of dermatological diagnosis and greatly increase the speed with which skin cancer is detected.