Utilization of Hybrid Topological Data Analysis and Deep Learning for Basal Cell Carcinoma Diagnosis
According to a study published in the Journal of Imaging Informatics in Medicine, telangiectasia features improve basal cell carcinoma (BCC) diagnosis, and topological data analysis (TDA) techniques hold the potential to improve deep learning (DL) performance.
Researchers introduced a novel approach by combining TDA and DL through ensemble learning to develop a hybrid TDA-DL BCC diagnostic model. Utilizing persistence homology, a TDA technique, topological features are extracted from segmented telangiectasia and skin lesions. DL features are generated by fine-tuning a pretrained EfficientNet-B5 model. The integration of these techniques results in a sophisticated diagnostic model.
The hybrid TDA-DL model achieved 97.4% accuracy and an AUC of 0.995 on a holdout test comprising 395 skin lesions for BCC diagnosis. This study highlights the significance of telangiectasia features in improving BCC diagnosis and demonstrates the potential of TDA techniques to enhance DL performance.
Overall, the integration of TDA and DL methodologies showcases a promising approach for advancing skin cancer diagnosis, particularly in accurately identifying BCC lesions. The study underscores the importance of leveraging multiple computational techniques to maximize diagnostic accuracy and efficacy in clinical settings.
Reference
Maurya A, Stanley RJ, Lama N, et al. Hybrid topological data analysis and deep learning for basal cell carcinoma diagnosis. J Imaging Inform Med. Published online January 12, 2024. doi:10.1007/s10278-023-00924-8