Researchers at the department of electronics and communication engineering in Punjabi University claimed to have developed a non-invasive method for the early detection of skin cancer using advanced dermoscopic imaging and machine learning techniques.
The study, titled ‘Detection of skin cancer in dermoscopic digital images’, explores the use of deep learning models and image processing techniques to accurately diagnose both benign and malignant skin lesions without the need for surgical biopsies. Four research papers of the study were recently published in SCI-indexed journals, and its findings have been presented at four national and international conferences.
Dr Shelly, a researcher in the department and the lead author of the study, said the method was designed to aid medical professionals by providing a safer and faster diagnosis of skin cancer at early stages. “Early intervention is critical in skin cancer cases, and our approach aims to reduce reliance on invasive procedures,” she said.
The research was conducted under the supervision of Dr Bal Krishan, who explained that the study integrated multiple stages of image analysis including pre-processing, segmentation, feature extraction, and classification.
Using publicly available medical image datasets - ISIC and PH2 - the model achieved high accuracy levels: 99.62% on ISIC and 99.98% on PH2, with an enhanced convolutional neural network framework. The study also introduced a hybrid feature extraction approach, combining shape, colour, and texture analysis to improve pattern detection in dermoscopic images. The use of the Cuckoo Search algorithm for feature selection further reduced computational complexity while increasing classification accuracy.
{{/usCountry}}Using publicly available medical image datasets - ISIC and PH2 - the model achieved high accuracy levels: 99.62% on ISIC and 99.98% on PH2, with an enhanced convolutional neural network framework. The study also introduced a hybrid feature extraction approach, combining shape, colour, and texture analysis to improve pattern detection in dermoscopic images. The use of the Cuckoo Search algorithm for feature selection further reduced computational complexity while increasing classification accuracy.
{{/usCountry}}Dr Shelly noted that the system was trained to distinguish between benign lesions, such as moles and cysts, and malignant forms that require immediate medical attention. Vice-chancellor Dr Jagdeep Singh said that contributions such as these are important in developing tools that can assist healthcare systems and improve public health outcomes.