Abstract: SA-PO005
Fully Automated Segmentation of Kidneys and Exophytic Cysts Using Deep Convolutional Neural Networks in Subjects With Autosomal Dominant Polycystic Kidney Disease
Session Information
- Bioengineering
November 05, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
Abstract Time: 10:00 AM - 12:00 PM
Category: Bioengineering
- 300 Bioengineering
Authors
- Kim, Youngwoo, Korea Institute of Industrial Technology, Cheonan, Korea (the Republic of)
- Bu, Seon-ah, Korea Institute of Industrial Technology, Cheonan, Korea (the Republic of)
- Bae, Kyongtae Ty, The University of Hong Kong Li Ka Shing Faculty of Medicine, Hong Kong, Hong Kong
Group or Team Name
- KITECH
Background
In diagnosing autosomal dominant polycystic kidney disease (ADPKD), total kidney volume (TKV) is a critical imaging biomarker. Nevertheless, manual computation of TKV, particularly with the exclusion of exophytic cysts, is laborious and time consuming. We developed a fully automated segmentation method for TKV using a deep learning network to selectively segment kidney regions while excluding exophytic cysts.
Methods
We used abdominal T2-weighted magnetic resonance images from 210 ADPKD subjects who were divided into two groups: 157 to train the network and 53 to test. With a 3D U-Net architecture using dataset fingerprints, the network was trained by K-fold cross-validation in that 80% out of 157 cases were for training and the remaining 20% cases for validation. To assess the performance of the automated segmentation method in reference to the manual method, three metrics were analyzed: Dice similarity coefficient (DSC), intra-class correlation coefficient (ICC), and Bland-Altman analysis.
Results
Excellent geometric concordance was achieved between the automated and manual reference methods (DSC: 0.962±0.018, on average) on the test datasets with widely varying volumes of kidneys (1058.5±706.8 ml, range: 178.9–2776.0 ml) and exophytic cysts (549.0±559.1 ml, range: 113.4–2497.6 ml). The ICC was 0.9994 (P < 0.001; CI: 0.9991–0.9996) with a minimum bias of -2.424 ml (95% limits of agreement: -49.80, 44.95).
Conclusion
A fully automated segmentation method that we developed measures TKV with exclusion of exophytic cysts as accurately as the level of a human expert. This technique will be useful in clinical studies that require automated computation of TKV to evaluate progression and treatment response of ADPKD.
Funding
- Government Support – Non-U.S.