CYSTSCAN-PKD: a comprehensive pipeline for automatic cyst segmentation and counting on µCT scans from PKD animal...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Biology

Background:

• Autosomal dominant polycystic kidney disease (ADPKD) is a genetic disorder characterized by progressive kidney cyst formation, leading to enlarged kidneys and impaired function.
• The PCK rat serves as a crucial preclinical model for ADPKD research, necessitating precise quantification of total kidney volume (TKV), total cyst volume (TCV), and cyst count.
• Accurate and reproducible phenotyping is essential for evaluating therapeutic interventions in ADPKD models.

Purpose of the Study:

• To develop and validate an automated segmentation pipeline for quantifying kidney and cyst volumes from micro-CT scans of excised PCK rat kidneys.
• To implement and assess an automated cyst counting algorithm for enhanced precision in preclinical ADPKD research.
• To establish a fast, accurate, and reproducible method for morphological biomarker quantification in ADPKD research.

Main Methods:

• An automated segmentation pipeline utilizing a 3D U-Net ensemble within the nnU-Net framework was developed, employing dual-channel input (raw µCT and Sobel-filtered images).
• Segmentation models were trained on varying subsets of Dataset 1 and evaluated on internal and independent external datasets (Dataset 2) to assess generalizability.
• Automated cyst counting was performed using a morphological algorithm optimized with a genetic algorithm, based on 3D distance transform and peak detection.

Main Results:

• The segmentation pipeline achieved high accuracy, with Dice Similarity Coefficients exceeding 0.99 for kidney and 0.98 for cysts on the training dataset and demonstrating comparable performance on the external dataset.
• The automated cyst counting method exhibited low variability, with operator-algorithm agreement surpassing inter-operator agreement across all evaluated metrics.
• The developed pipeline significantly reduced processing time compared to manual methods, offering substantial gains in research efficiency.

Conclusions:

• The automated segmentation and cyst counting pipeline provides a fast, accurate, and reproducible solution for quantifying key morphological biomarkers in preclinical ADPKD research.
• This computational approach enhances the reliability of phenotyping in PCK rats, facilitating more effective evaluation of novel therapeutic strategies.
• The pipeline's cross-scanner generalizability supports its broad applicability in diverse research settings for polycystic kidney disease studies.