@article {10.3844/jcssp.2026.813.825,
article_type = {journal},
title = {Hybrid Deep Learning for Kidney Stone Detection in CT Scans With Noise Reduction and Feature Enhancement},
author = {Dimlo, U. M. Fernandes and Banoth, Sreenu and Bihade, Priti and Mjery, Yahia and Kumar, K. Jayaram and R, Umanesan and Kumar, A. Saran and Bhoopathy, V.},
volume = {22},
number = {3},
year = {2026},
month = {Mar},
pages = {813-825},
doi = {10.3844/jcssp.2026.813.825},
url = {https://thescipub.com/abstract/jcssp.2026.813.825},
abstract = {The identification of kidney stones using CT scans is an essential but difficult effort in medical diagnostics, frequently obstructed by imaging noise and the complexity of manual interpretation. Although conventional methods are widely used, they suffer from inaccuracies and inefficiencies, necessitating the development of automated diagnostic solutions. This study tackles these issues by presenting HDCNRNet, a hybrid deep learning network explicitly developed for automated kidney stone identification. The suggested approach incorporates Convolutional Neural Networks (CNNs) with sophisticated noise reduction methodologies and improved feature extraction modules to boost the diagnostic precision and dependability of kidney stone identification. HDCNRNet surpasses current models by attaining exceptional performance measures, including an accuracy of 97.8±0.3, sensitivity of 97.8, specificity of 99.2 precision of 98.0, and an F1-score of 97.6% ± 0.4. These findings demonstrate a significant improvement over baseline models such as ResNet-50 and VGG16, highlighting the model's superior ability to identify kidney stones while minimizing false positives and negatives. The use of excellent noise reduction techniques and feature enhancement components guarantees the model's efficacy despite fluctuating and noisy CT scan data. This research advances medical imaging by providing a scalable, efficient, and highly accurate AI-based method for kidney stone identification, readily integrable into clinical workflows. The results indicate that HDCNRNet can substantially boost diagnostic outcomes, lower the workload of radiologists, and improve patient care by providing more reliable and quicker diagnoses.},
journal = {Journal of Computer Science},
publisher = {Science Publications}
}