Abstract

Skin lesions are a common dermatological condition that can be benign or malignant. The purpose of this paper is to develop highly accurate system for early detection and characterization of skin lesions which will be crucial for effective treatment and prevention of further complications. To achieve this Photo-acoustic imaging which is a non-invasive imaging technique is used as laser light to generate ultrasound signals for imaging tissues. In this study, we investigated the use of artificial intelligence (AI)-enabled photo-acoustic imaging for the characterization analysis of skin lesions. We developed a deep learning algorithm such as neural network (NN) and Convolutional neural network (CNN) that can automatically classify skin lesions as benign or malignant based on the Photo-acoustic images. CNN found an accuracy of 98.5% for lesion classification and segmentation while NN found an accuracy of 93% for lesion classification and segmentation. The CNN achieved Sensitivity of 95% while NN achieved Sensitivity of 94%. The results demonstrate the potential of AI-enabled photo-acoustic imaging as a non-invasive tool for early detection and characterization of skin lesions.

Keywords:Artificial Intelligence; Classification; Skin lesion; CNN; NN