Design and validation of renal stone detection using multi-architecture feature extraction with deep sequential...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Nephrology

Background:

• Kidney stone disease (renal calculi) is a growing public health concern.
• Accurate detection of kidney stones is crucial for effective treatment and management.
• Computed tomography (CT) is a key imaging modality for kidney stone diagnosis.

Purpose of the Study:

• To develop a reliable and efficient deep learning system for accurate kidney stone identification from CT images.
• To introduce the Feature Integration and Sequential Attention Framework for Kidney Stone Detection (FISAF-KSD).

Main Methods:

• Image pre-processing and augmentation were applied to CT scans.
• Feature extraction was performed using a fusion of EfficientNetV2L, InceptionV3, and ResNet-101 deep learning models.
• A bidirectional gated recurrent unit network with an attention mechanism (BiGRU-AM) was utilized for classification.

Main Results:

• The FISAF-KSD methodology achieved high performance metrics, including an accuracy of 98.75%.
• The proposed framework demonstrated superior performance compared to existing kidney stone detection methods.
• The system effectively identified kidney stones on an Axial CT imaging dataset.

Conclusions:

• The FISAF-KSD approach offers a robust and accurate solution for kidney stone detection using deep learning.
• This framework has the potential to enhance the diagnostic process for renal calculi.
• The study highlights the efficacy of integrating multiple deep learning models and attention mechanisms for medical image analysis.