A two-stage deep learning framework for kidney disease detection using modified specular-free imaging and...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Nephrology

Background:

• Kidney diseases pose a growing public health challenge with increasing incidence.
• Accurate and early diagnosis of kidney pathologies is crucial for effective patient management.
• Existing diagnostic methods may face limitations in image quality and classification accuracy.

Purpose of the Study:

• To develop and validate a novel two-stage diagnostic model for enhanced detection of kidney diseases.
• To improve the visual quality of renal images using an advanced enhancement technique.
• To achieve high accuracy in classifying various renal pathologies.

Main Methods:

• Implementation of a Modified Specular-Free (MSF) technique for adaptive renal image enhancement.
• Classification of enhanced images using the EfficientNet-B2 deep learning architecture.
• Comparative analysis against established pre-trained models (VGG16, ResNet50, DenseNet, EfficientNet variants).

Main Results:

• The proposed model achieved a diagnostic accuracy of 98.27%.
• The model demonstrated superior performance compared to all benchmarked pre-trained models.
• Effective differentiation was achieved between normal renal conditions and pathologies like tumors, stones, and cysts.

Conclusions:

• The integrated MSF technique and EfficientNet-B2 model offer a powerful approach for accurate kidney disease diagnosis.
• This research highlights the potential of advanced image enhancement coupled with deep learning for medical imaging.
• The developed model presents a scalable solution for improving diagnostic accuracy in complex medical imaging scenarios.