Artificial Intelligence-Based Classification of CT Images Using a Hybrid SpinalZFNet.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Nephrology

Background:

• Kidney diseases, including chronic kidney disease and kidney failure, arise from various factors like diet and medical conditions.
• Accurate and timely diagnosis of renal disorders is crucial and benefits from integrating patient data with computed tomography (CT) images.
• Deep Neural Networks (DNNs) offer high accuracy in medical tasks but face computational challenges in real-time applications.

Purpose of the Study:

• To introduce SpinalZFNet, a novel hybrid deep learning model for accurate kidney disease classification using CT images.
• To enhance feature analysis and reduce computational overhead in kidney disease detection.
• To classify kidney diseases into normal, tumor, cyst, and stone categories.

Main Methods:

• CT images were pre-processed using a median filter and segmented with Efficient Neural Network (ENet).
• Image augmentation and feature extraction were performed on the pre-processed CT images.
• The proposed SpinalZFNet model, integrating Spinal Network and Zeiler Fergus Network, was used for classification.

Main Results:

• SpinalZFNet achieved high performance metrics: 99.9% sensitivity, 99.5% specificity, 99.6% precision, 99.8% accuracy, and 99.7% F1-Score.
• The hybrid model demonstrated superior performance compared to other evaluated models.
• The approach effectively reduced computational overhead while maintaining high classification accuracy.

Conclusions:

• SpinalZFNet offers a highly accurate and computationally efficient method for classifying kidney diseases from CT images.
• The integration of SpinalNet and ZFNet architectures provides enhanced feature extraction for improved diagnostic capabilities.
• This deep learning approach holds significant potential for timely and accurate diagnosis of various renal disorders.