Algorithm for Accurate CT Kidney Localization and Marking by Dual-polar Coordinates Multi-image Fusion

This work aims to address the challenges posed by the large number of kidney CT images, which reduce diagnostic efficiency. A multi-image fusion CT kidney localization and identification algorithm based on a dual-polar coordinate system is proposed.

First, the spine position is determined from the kidney CT image, and a polar coordinate system is established accordingly. The kidney's polar coordinates are then derived based on the spine's polar coordinate system. This two-stage polar representation-where the spine is first localized using one polar coordinate system and the kidney is subsequently described relative to it-is termed "dual-polar coordinates." In addition, an adaptive multi-image identification algorithm is proposed that leverages changes in renal polar radius across multiple images-before and after adaptation-to identify kidney images accurately. Finally, when data volume is limited, this algorithm can be extended to automatically annotate the kidney region, effectively reducing label noise and improving diagnostic efficiency.

Experimental results demonstrate the high accuracy of the proposed algorithm in spine and kidney localization, recognition, and annotation. It performs particularly well on kidney images with cysts, correctly identifying both normal and cystic renal tissues and providing complete kidney outlines.

By simplifying kidney localization via the dual-polar coordinate system and fusing information from multiple images, the algorithm effectively handles morphological variations caused by cysts. It reduces physicians' workload and improves diagnostic robustness, offering a novel and efficient tool for medical imaging diagnosis.

The proposed algorithm significantly enhances the accuracy and efficiency of kidney CT image analysis, especially for cyst-containing kidneys, and provides strong support for clinical diagnosis and treatment of renal diseases.