Masksemble-Aided Cross-ViT for Uncertainty Estimation in Skin Cancer Diagnosis
*Corresponding Author: Aniket Guchhait, Department of Computer Science and Engineering, Ramkrishna Mahato Government Engineering College, Bakura, West Bengal, India, India, aniketguchhait_cse_2025@rkmgec.ac.inReceived Date: Sep 19, 2024 / Published Date: Jun 11, 2025
Citation: Guchhait A, Barman A, Roy SK (2025) Masksemble-Aided Cross-ViT for Uncertainty Estimation in Skin Cancer Diagnosis. J Oncol Res Treat 10:310.DOI: 10.4172/aot.1000310
Copyright: © 0 . This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
| Peer-reviewed Full Article 
Abstract
In this work, we investigate a Masksemble-aided Cross ViT model to measure the uncertainty of feature representations for cancer identification. We propose a Cross ViT with a special Masksemble block in order to create discriminative image features. The Masksemble layer estimates the uncertainty of a given dermatoscopy image that plays a crucial role in cancer identification, and then it is passed to the Cross ViT network for the classification task. The comprehensive results show that our method outperforms CNN models and vision transformers. The model will detect skin cancer by differentiating the cancerous cells (malignant) from the non-cancerous ones (benign). The prediction of the model is measured by performance metrics such as precision, recall, F1-score, and average accuracy along with class-wise accuracy, which shows the effectiveness of the proposed method. In addition to being verified for binary classification, the suggested model is also tested for many classes using the HAM-10000 dataset, demonstrating the system’s effectiveness in multiple-classification scenarios.
