Machine Learning–Driven Recognition of Basal Cell Carcinoma via Multimodal Ultrasound and Optical Imaging
Isabella A. Serebryakova1, Yuriy I. Surkov1, Elina A. Genina1, Yana K. Kuzinova3, Olga M. Konopatskova3, Valery V. Tuchin1,2,4
1 Saratov State University, Saratov, Russia
2 Tomsk State University, Tomsk, Russia
3 Saratov State Medical University, Saratov, Russia
4 FRC “Saratov Scientific Centre of the Russian Academy of Sciences”, Saratov, Russia
Abstract
Noninvasive diagnosis of skin conditions remains a key challenge in contemporary medical practice. In this study, we propose a novel multimodal framework that combines data from Optical Coherence Tomography (OCT), Diffuse Reflectance Spectroscopy (DRS), and High-Frequency Ultrasound (HFUS) to improve the detection of skin lesions. The approach leverages advanced image processing, spectral feature extraction, texture analysis, and machine learning methods—specifically gradient boosting and classifier fusion techniques—to interpret the multimodal input. Our integrated model achieves a high classification accuracy of 96% in distinguishing benign from malignant skin neoplasms, showcasing its strong potential for clinical application. The findings underscore the value of combining complementary imaging modalities through intelligent data fusion, significantly enhancing the accuracy and robustness of skin cancer diagnosis.
This study was supported by the grant from the Russian Sciece Foundation № 23-14-00287
Speaker
Isabella A. Serebryakova
Saratov State University, Saratov, Russia
Russia
Upload Report
Discussion
Ask question
