SinoXivPredicting Brain Tumours Through Transfer Learning and Optimized Hyperparameters Using the Modified Whale’s Algorithm
Abstract
Brain tumours cause severe morbidity. Needs to be diagnosed in the minimum time and with the Highest accuracy. It takes a long time to learn Deep learning Models that can predict the type and severity of brain tumours. Using values for some selected Hyperparameters often leads to non-convergence, and accuracy cannot be guaranteed. Full-domain-focused adaptive Transfer Learning using models such as ResNet18 reduces learning time and archives 100\% accuracy. The learning time can be reduced while maintaining 100\% accuracy by selecting appropriate, relevant, and ranked hyperparameters and their values using the Modified Whales algorithm, which focuses on the specified search space. The proposed method reduced learning and prediction time by 66\% (from 3 to 1 second) while achieving 100\% accuracy, exceeding the accuracy of other methods reported in the literature.
Citation Information
@article{sivanipinnaboina2026,
title={Predicting Brain Tumours Through Transfer Learning and Optimized Hyperparameters Using the Modified Whale’s Algorithm},
author={Sivani Pinnaboina and Venkata Sowmya Kambhampati and Kodanda Rama Sastry Jammalamadaka and Sasi Bhanu Jammalamadaka},
journal={Scientific Reports},
year={2026},
doi={https://doi.org/10.21203/rs.3.rs-9365075/v1}
}
