A deep learning-based precision volume calculation approach for kidney and tumor segmentation on computed tomography...

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Nephrology and Urology

Background:

• Traditional interpretation of kidney computed tomography (CT) images relies on physician experience, which is time-consuming and prone to inconsistencies.
• The increasing volume of CT scans necessitates automated solutions for efficient and accurate kidney disease diagnosis.
• Existing neural network models often focus on architecture modifications, overlooking the critical role of data pre-processing.

Purpose of the Study:

• To systematically investigate and present essential pre-processing methods for medical images prior to neural network analysis.
• To evaluate the impact of proposed pre-processing techniques on the accuracy of kidney and tumor detection in CT images.
• To demonstrate the clinical applicability and cost-effectiveness of optimized deep learning models for kidney imaging.

Main Methods:

• Development and application of novel data pre-processing techniques tailored for medical imaging.
• Implementation of deep learning models, specifically neural networks, for automated image analysis.
• Systematic comparison of model performance with and without the proposed pre-processing steps.

Main Results:

• Significant improvements in accuracy rates for kidney segmentation and tumor detection were observed post pre-processing.
• The Dice score for kidney segmentation improved from 0.9436 to 0.9648.
• Tumor detection accuracy reached a Dice score of 0.7294, demonstrating effectiveness across various tumor types.

Conclusions:

• Data pre-processing is a critical factor in enhancing the performance of neural network models for medical image analysis.
• The proposed methods enable accurate kidney volume calculation and tumor detection with reduced computational resources.
• The developed approach offers a cost-effective and efficient solution for clinical applications in nephrology and oncology.