Characterizing the ADPKD-IFT140 Phenotypic Signature With Deep Learning and Advanced Imaging Biomarkers.

Area of Science:

• Nephrology
• Genetics
• Medical Imaging

Background:

• Autosomal dominant polycystic kidney disease (ADPKD) is a common genetic disorder.
• IFT140 variants represent the third most frequent cause of ADPKD.
• Understanding the unique characteristics of ADPKD-IFT140 is crucial for diagnosis and management.

Purpose of the Study:

• To characterize the clinical presentation, disease progression, and imaging phenotype of ADPKD-IFT140.
• To develop and validate predictive models for distinguishing ADPKD-IFT140 from other ADPKD subtypes.
• To assess the utility of deep-learning and practical models in identifying ADPKD-IFT140.

Main Methods:

• Retrospective cohort study of patients with ADPKD variants in IFT140, PKD1NT, or PKD2.
• Development and validation of a deep-learning model and a practical model for ADPKD subtype prediction.
• Models were trained and validated using cyst characteristics, total kidney volume (TKV), and estimated glomerular filtration rate (eGFR) data.

Main Results:

• ADPKD-IFT140 patients had fewer, larger cysts and no liver cysts compared to PKD1NT and PKD2 groups.
• eGFR decline was significantly slower in ADPKD-IFT140 patients.
• The deep-learning model achieved high accuracy (AUC 0.949), and a practical TKV ratio demonstrated good sensitivity (72.2%) in identifying ADPKD-IFT140.

Conclusions:

• ADPKD-IFT140 exhibits a distinct clinical and imaging phenotype.
• Predictive models, including a deep-learning approach and a practical volume-based method, can effectively distinguish ADPKD-IFT140.
• These models offer valuable tools for clinical settings to identify and manage ADPKD-IFT140 patients.