Ying-Ying Ren ^1* Jian-Ping Mei ^2* Chun-Long Lu ^2* Xuan-Guang Ye ^1* Li-Hong Mei ^1* Gao YANG ^1*

• ¹ Jinshan Hospital, Fudan University, Shanghai, China
• ² Zhejiang University of Technology, Hangzhou, Zhejiang Province, China

Skin cancer is a significant global health concern, with actinic keratosis (AK) being a prevalent precancerous lesion resulting from long-term sun exposure, and seborrheic keratosis (SK) being a benign skin growth of keratinocytes. Distinguishing between AK and SK poses a substantial challenge for dermatologists due to their visual similarities. This study aims to investigate the efficacy of deep learning techniques in assisting dermatologists in accurately classifying AK and SK lesions. The authors conducted a study utilizing deep learning algorithms to analyze images of AK and SK lesions. The dataset consisted of a large number of annotated images for training and validation purposes. Performance metrics such as accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC) were employed to evaluate the deep learning model's performance. The results demonstrate promising performance of the deep learning model in accurately classifying AK and SK lesions. The model achieved high accuracy and sensitivity, indicating its potential as a valuable tool for dermatologists in clinical practice. This study highlights the utility of deep learning techniques in augmenting dermatologists' diagnostic capabilities for distinguishing between AK and SK lesions. The findings suggest that deep learning-based approaches have the potential to improve diagnostic accuracy, reduce misdiagnosis rates, and enhance patient outcomes in the management of skin cancer.