A Clinically Interpretable Deep CNN Framework for Early Chronic Kidney Disease Prediction Using Grad-CAM-Based Explainable AI

arXiv ID: 2512.09244v1

Published: 2025-12-10

Authors: Anas Bin Ayub, Nilima Sultana Niha, Md. Zahurul Haque

Categories: cs.CV, cs.AI

Relevance Score: 1.00 / 1.00

View on arXiv Download PDF

Summary

This paper introduces a deep Convolutional Neural Network (CNN) framework for early Chronic Kidney Disease (CKD) prediction from CT kidney images, incorporating SMOTE for class balancing and Grad-CAM for explainability. Trained on a diverse dataset encompassing normal, cyst, stone, and tumor cases, the model achieved a remarkable 100% classification accuracy, demonstrating significant potential for improving clinical diagnostics and early intervention strategies in nephrology.

Medical Relevance

Early and accurate detection of Chronic Kidney Disease (CKD) is paramount for preventing disease progression to end-stage renal failure, reducing associated morbidity and mortality, and enabling timely clinical management to preserve renal function and improve patient quality of life.

AI Health Application

The paper describes the development of a deep Convolutional Neural Network (CNN) framework, enhanced with explainable AI (Grad-CAM), for the early and accurate prediction/detection of Chronic Kidney Disease from CT kidney images. This AI application serves as a diagnostic aid to clinicians for timely identification of CKD, supporting early medical intervention and improved patient outcomes.

Key Points

Addresses Chronic Kidney Disease (CKD) as a major global medical burden requiring efficient early diagnostic approaches.
Proposes a deep Convolutional Neural Network (CNN) specifically for early CKD detection using CT kidney images.
Integrates Synthetic Minority Over-sampling Technique (SMOTE) for robust class balancing, crucial for medical datasets often prone to imbalance.
Utilizes Gradient-weighted Class Activation Mapping (Grad-CAM) to provide explainable AI (XAI), enhancing the clinical interpretability of predictions.
Evaluated on the CT KIDNEY DATASET, consisting of 12,446 CT images categorized into normal, cyst, stone, and tumor cases.
Achieved a 100% classification accuracy in the early detection of chronic kidney disease, highlighting highly effective diagnostic capabilities.
Demonstrates strong potential to address critical clinical diagnostic challenges and enhance early medical intervention strategies for kidney health.

Methodology

The study employed a deep Convolutional Neural Network (CNN) architecture trained on the CT KIDNEY DATASET, which comprises 12,446 CT images encompassing four categories: cyst (3,709), normal (5,077), stone (1,377), and tumor (2,283) cases. To mitigate potential issues arising from class imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was applied. Furthermore, Gradient-weighted Class Activation Mapping (Grad-CAM) was integrated to provide explainable AI capabilities, generating visual heatmaps to highlight the regions within CT images most salient for the model's predictions, thereby enhancing clinical interpretability.

Key Findings

The proposed deep CNN model achieved a 100% classification accuracy in the early detection of chronic kidney disease, demonstrated by its performance in classifying specific kidney conditions (cyst, normal, stone, tumor) from CT images.

Clinical Impact

This highly accurate, AI-driven diagnostic tool for early CKD detection from CT scans holds significant promise for clinical practice. It could enable faster, more precise identification of various kidney pathologies, leading to earlier and more effective medical interventions, potentially delaying disease progression, improving patient management, and ultimately enhancing long-term health outcomes for individuals at risk of or suffering from CKD.

Limitations

The abstract does not explicitly state any limitations of the study.

Future Directions

The abstract does not explicitly suggest future research directions.

Medical Domains

Nephrology Radiology Diagnostic Imaging Artificial Intelligence in Medicine Urology

Keywords

Chronic Kidney Disease Deep Learning Convolutional Neural Network CT Images Explainable AI Grad-CAM SMOTE Early Detection

Abstract

Chronic Kidney Disease (CKD) constitutes a major global medical burden, marked by the gradual deterioration of renal function, which results in the impaired clearance of metabolic waste and disturbances in systemic fluid homeostasis. Owing to its substantial contribution to worldwide morbidity and mortality, the development of reliable and efficient diagnostic approaches is critically important to facilitate early detection and prompt clinical management. This study presents a deep convolutional neural network (CNN) for early CKD detection from CT kidney images, complemented by class balancing using Synthetic Minority Over-sampling Technique (SMOTE) and interpretability via Gradient-weighted Class Activation Mapping (Grad-CAM). The model was trained and evaluated on the CT KIDNEY DATASET, which contains 12,446 CT images, including 3,709 cyst, 5,077 normal, 1,377 stone, and 2,283 tumor cases. The proposed deep CNN achieved a remarkable classification performance, attaining 100% accuracy in the early detection of chronic kidney disease (CKD). This significant advancement demonstrates strong potential for addressing critical clinical diagnostic challenges and enhancing early medical intervention strategies.