Automated Segmentation Method Enables Fast TKV Quantification

A COHORT STUDY found that an automated segmentation method can enable fast, accurate, and reproducible total kidney volume (TKV) quantification from unenhanced CT images in patients with autosomal dominant polycystic kidney disease (ADPKD).

Automated Deep Learning-Based Method for TKV

The development dataset comprised unenhanced CT scans from 236 patients, including 111 patients with ADPKD. This dataset was divided into a training cohort (150 patients; 56 with ADPKD) and an internal test cohort (86 patients; 55 with ADPKD). The independent validation dataset included 70 patients with ADPKD.

Non-ADPKD cases included a broad spectrum of renal morphologies, such as normal kidneys as well as kidneys with renal cell carcinoma.

Examinations were performed using multi-detector CT scanners (≥ 16 detector rows) without contrast enhance­ment, and the reference standard was generated using semi-automated segmentation followed by slice-by-slice manual correction.

Two experienced radiologists, blinded to the automated segmenta­tion results and clinical information, manually reviewed and corrected the kidney boundaries on the original 2-mm CT images.

The trained model automatically generated three-dimensional kidney segmentation masks, from which right kidney volume (RKV), left kidney volume (LKV), and total kidney volume (TKV) were calculated.

Fast TKV Quantification

Automated segmentation demonstrated excellent spatial agreement with the reference standard. Mean dice similarity coefficient (DSC) values were 0.95 ± 0.01, 0.96 ± 0.01, and 0.96 ± 0.01 for RKV, LKV, and TKV, respectively.

Researchers observed minimal bias and narrow 95% limits of agreement across the measurement range. For TKV, the mean bias was −0.14%, with a precision of 2.09%.

Mean absolute TKV growth was 126.07 ± 141.61 mL using the reference standard and 125.35 ± 140.89 mL, and percent­age TKV growth was 6.85 ± 6.12% and 6.89 ± 6.04%, respectively.

Automated segmentation required 5 ± 2.3 s per examination, compared with 30 ± 12.1 min for generation of the reference standard.

Future of TKV Quantification 

Researchers concluded that automated segmentation enables fast, reproducible, and clinically feasible TKV quantification from routinely acquired unenhanced CT examinations and may facilitate broader implementation of automated volu­metric assessment in patients with ADPKD and may offer a clinically feasible alternative to manual volumetric analysis.

Multicentre studies involving different scanners, imaging protocols, and patient populations could confirm validate these findings in broader clinical settings.

Reference:

Hu Y et al. A rapid automated segmentation method for total kidney volume measurement on unenhanced computed tomography in autosomal dominant polycystic kidney disease. Abdom Radiol.2026. DOI:org/10.1007/s00261-026-05799-1.