EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

DEEP LEARNING-BASED CLINICAL DECISION SUPPORT SYSTEM FOR AUTOMATED KIDNEY STONE DETECTION IN CT IMAGES USING CUSTOM CNN AND TRANSFER LEARNING

Amjad khan, Muhammad Javed, Nasir Gul, Saif Ullah Noor, Reyan, Ashraf Ullah

Kashf Journal of Multidisciplinary Research · 20.06.2026 · 3 (06) · 23-47

DOI: 10.71146/kjmr952 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Kashf Journal of Multidisciplinary Research, online erschienen am 20.06.2026. Verfasst von Amjad khan und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Kidney stone disease (nephrolithiasis) is a prevalent urological condition that is arguably the most prevalent one and also has a high burden of disease for the healthcare system. Even today, non-contrast computed tomography (NCCT) remains the gold standard imaging technique for stone detection, although the manual reading of computed tomography (CT) scans may be time-consuming and observer-dependent, especially for high-volume clinical practice. In this work, we develop an automated deep learning system for CT image detection of kidney stones from axial images. Two CNN-based systems were designed and tested: one with a specifically designed CNN architecture optimized for computation efficiency (cnnD) and another (cnnT) with a transfer learning approach using an architecture pre-trained on ImageNet. There were a total of 3,154 annotated CT images, which were used for training (2,522 images), validation (316 images), and testing (316 images). Pre-processing and normalization of standard images and data augmentation were used to enhance the model's robustness and generalisation. The custom CNN model's accuracy of 94.3%, precision of 0.95, recall of 0.94, and F1-score of 0.94 on the independent test set were validated in the experiment. Therefore, transfer learning for medical image analysis was demonstrated to be effective when the fine-tuned VGG16 model classified the images with a higher accuracy (96.5%). The custom CNN had less computation needed and also had quicker inference time as compared to VGG16, even though the latter achieved better predictive performance and thus could be applied in real clinical settings. The results suggest that the deep learning methods are useful for detecting kidney stones with high accuracy and efficiency from CT scans, and potentially for assisting radiologists in their clinical diagnosis and decisions. Overall, the proposed framework emphasizes the potential of using AI throughout the diagnostic process, and its goal is to improve the efficiency, uniformity, and accessibility of the diagnosis.

Zitieren

APA

khan, A., Javed, M., Gul, N., Noor, S. U., Reyan, & Ullah, A. (2026). DEEP LEARNING-BASED CLINICAL DECISION SUPPORT SYSTEM FOR AUTOMATED KIDNEY STONE DETECTION IN CT IMAGES USING CUSTOM CNN AND TRANSFER LEARNING. Kashf Journal of Multidisciplinary Research, 3(06), 23-47. https://doi.org/10.71146/kjmr952

Vancouver

khan A, Javed M, Gul N, Noor SU, Reyan, Ullah A. DEEP LEARNING-BASED CLINICAL DECISION SUPPORT SYSTEM FOR AUTOMATED KIDNEY STONE DETECTION IN CT IMAGES USING CUSTOM CNN AND TRANSFER LEARNING. Kashf Journal of Multidisciplinary Research. 2026;3(06):23-47. doi:10.71146/kjmr952

BibTeX
@article{khan2026,
  author = {khan, Amjad and Javed, Muhammad and Gul, Nasir and Noor, Saif Ullah and Reyan and Ullah, Ashraf},
  title = {DEEP LEARNING-BASED CLINICAL DECISION SUPPORT SYSTEM FOR AUTOMATED KIDNEY STONE DETECTION IN CT IMAGES USING CUSTOM CNN AND TRANSFER LEARNING},
  journal = {Kashf Journal of Multidisciplinary Research},
  year = {2026},
  volume = {3},
  number = {06},
  pages = {23--47},
  doi = {10.71146/kjmr952}
}
Kostenlos starten →