Automatically Detecting Pancreatic Cysts in Autosomal Dominant Polycystic Kidney Disease on MRI Using Deep Learning

• Published in: Tomography (Ann Arbor) 2024-07, Vol.10 (7), p.1148-1158
• Main Authors: Wang, Sophie J, Hu, Zhongxiu, Li, Collin, He, Xinzi, Zhu, Chenglin, Wang, Yin, Sattar, Usama, Bazojoo, Vahid, He, Hui Yi Ng, Blumenfeld, Jon D, Prince, Martin R
• Format: Article
• Language: English
• Subjects: ADPKD; Adult; Aged; Deep Learning; Diseases; Female; Genetic aspects; Humans; Image Interpretation, Computer-Assisted - methods; imaging biomarker; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Male; Middle Aged; MRI; pancreas; Pancreas - diagnostic imaging; Pancreas - pathology; pancreatic cyst; Pancreatic Cyst - diagnostic imaging; Pancreatic Cyst - pathology; Polycystic kidney disease; Polycystic Kidney, Autosomal Dominant - complications; Polycystic Kidney, Autosomal Dominant - diagnostic imaging; Polycystic Kidney, Autosomal Dominant - pathology; Reproducibility of Results
• ISSN: 2379-139X; 2379-1381; 2379-139X
• DOI: 10.3390/tomography10070087

Pancreatic cysts in autosomal dominant polycystic kidney disease (ADPKD) correlate with PKD2 mutations, which have a different phenotype than PKD1 mutations. However, pancreatic cysts are commonly overlooked by radiologists. Here, we automate the detection of pancreatic cysts on abdominal MRI in ADPKD. Eight nnU-Net-based segmentation models with 2D or 3D configuration and various loss functions were trained on positive-only or positive-and-negative datasets, comprising axial and coronal T2-weighted MR images from 254 scans on 146 ADPKD patients with pancreatic cysts labeled independently by two radiologists. Model performance was evaluated on test subjects unseen in training, comprising 40 internal, 40 external, and 23 test-retest reproducibility ADPKD patients. Two radiologists agreed on 52% of cysts labeled on training data, and 33%/25% on internal/external test datasets. The 2D model with a loss of combined dice similarity coefficient and cross-entropy trained with the dataset with both positive and negative cases produced an optimal dice score of 0.7 ± 0.5/0.8 ± 0.4 at the voxel level on internal/external validation and was thus used as the best-performing model. In the test-retest, the optimal model showed superior reproducibility (83% agreement between scan A and B) in segmenting pancreatic cysts compared to six expert observers (77% agreement). In the internal/external validation, the optimal model showed high specificity of 94%/100% but limited sensitivity of 20%/24%. Labeling pancreatic cysts on T2 images of the abdomen in patients with ADPKD is challenging, deep learning can help the automated detection of pancreatic cysts, and further image quality improvement is warranted.