Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cyst-X: A Federated AI System Outperforms Clinical Guidelines to Detect Pancreatic Cancer Precursors and Reduce Unnecessary Surgery

Hongyi Pan, Gorkem Durak, Elif Keles, Deniz Seyithanoglu, Zheyuan Zhang, Alpay Medetalibeyoglu, Halil Ertugrul Aktas, Andrea Mia Bejar, Ziliang Hong, Yavuz Taktak, Gulbiz Dagoglu Kartal, Mehmet Sukru Erturk, Timurhan Cebeci, Maria Jaramillo Gonzalez, Yury Velichko, Lili Zhao, Emil Agarunov, Federica Proietto Salanitri, Concetto Spampinato, Pallavi Tiwari, Ziyue Xu, Sachin Jambawalikar, Ivo G. Schoots, Marco J. Bruno, Chenchan Huang, Candice W. Bolan, Tamas Gonda, Frank H. Miller, Rajesh N. Keswani, Michael B. Wallace, Ulas Bagci

TL;DR

Cyst-X introduces a federated AI framework for IPMN risk prediction that outperforms Kyoto guidelines and expert radiologists on a large, multi-center MRI dataset. It combines PanSegNet-based pancreas segmentation with DenseNet-121 classification and radiomics to achieve $AUC=0.82$ and markedly higher sensitivity for high-risk IPMNs, while enabling privacy-preserving federated training with near-centralized performance for classification. The study provides a comprehensive methods pipeline, including 3D radiomics, robust segmentation, and analysis across MRI modalities, and publicly releases the dataset and models to accelerate research. Collectively, the work offers a path toward improved early pancreatic cancer detection and reduced unnecessary surgeries, with practical implications for multi-institutional collaboration under data-privacy constraints.

Abstract

Pancreatic cancer is projected to be the second-deadliest cancer by 2030, making early detection critical. Intraductal papillary mucinous neoplasms (IPMNs), key cancer precursors, present a clinical dilemma, as current guidelines struggle to stratify malignancy risk, leading to unnecessary surgeries or missed diagnoses. Here, we developed Cyst-X, an AI framework for IPMN risk prediction trained on a unique, multi-center dataset of 1,461 MRI scans from 764 patients. Cyst-X achieves significantly higher accuracy (AUC = 0.82) than both the established Kyoto guidelines (AUC = 0.75) and expert radiologists, particularly in correct identification of high-risk lesions. Clinically, this translates to a 20% increase in cancer detection sensitivity (87.8% vs. 64.1%) for high-risk lesions. We demonstrate that this performance is maintained in a federated learning setting, allowing for collaborative model training without compromising patient privacy. To accelerate research in early pancreatic cancer detection, we publicly release the Cyst-X dataset and models, providing the first large-scale, multi-center MRI resource for pancreatic cyst analysis.

Cyst-X: A Federated AI System Outperforms Clinical Guidelines to Detect Pancreatic Cancer Precursors and Reduce Unnecessary Surgery

TL;DR

Cyst-X introduces a federated AI framework for IPMN risk prediction that outperforms Kyoto guidelines and expert radiologists on a large, multi-center MRI dataset. It combines PanSegNet-based pancreas segmentation with DenseNet-121 classification and radiomics to achieve and markedly higher sensitivity for high-risk IPMNs, while enabling privacy-preserving federated training with near-centralized performance for classification. The study provides a comprehensive methods pipeline, including 3D radiomics, robust segmentation, and analysis across MRI modalities, and publicly releases the dataset and models to accelerate research. Collectively, the work offers a path toward improved early pancreatic cancer detection and reduced unnecessary surgeries, with practical implications for multi-institutional collaboration under data-privacy constraints.

Abstract

Pancreatic cancer is projected to be the second-deadliest cancer by 2030, making early detection critical. Intraductal papillary mucinous neoplasms (IPMNs), key cancer precursors, present a clinical dilemma, as current guidelines struggle to stratify malignancy risk, leading to unnecessary surgeries or missed diagnoses. Here, we developed Cyst-X, an AI framework for IPMN risk prediction trained on a unique, multi-center dataset of 1,461 MRI scans from 764 patients. Cyst-X achieves significantly higher accuracy (AUC = 0.82) than both the established Kyoto guidelines (AUC = 0.75) and expert radiologists, particularly in correct identification of high-risk lesions. Clinically, this translates to a 20% increase in cancer detection sensitivity (87.8% vs. 64.1%) for high-risk lesions. We demonstrate that this performance is maintained in a federated learning setting, allowing for collaborative model training without compromising patient privacy. To accelerate research in early pancreatic cancer detection, we publicly release the Cyst-X dataset and models, providing the first large-scale, multi-center MRI resource for pancreatic cyst analysis.

Paper Structure

This paper contains 24 sections, 7 equations, 7 figures, 13 tables.

Table of Contents

  1. Main
  2. Results
  3. Superior performance of PanSegNet for pancreas segmentation
  4. Cyst-X classification models outperform clinical guidelines and expert radiologists
  5. Importance of accurate pancreas segmentation in classification
  6. Comparison with expert radiologists
  7. Privacy-preserving federated learning maintains high performance
  8. UMAPs of quality indicators on T1W and T2W modalities
  9. Discussion
  10. Methods
  11. Construction and characteristics of the Cyst-X dataset
  12. Image preprocessing and segmentation
  13. Radiologist visual scoring
  14. Radiomics approach for IPMN classification
  15. DenseNet-121 for IPMN classification
  16. ...and 9 more sections

Figures (7)

  • Figure 1: Cyst-X Pipeline. (A) T1W and T2W MRI scans were collected from seven medical centers. (B) PanSegNet segments the pancreas: MRI scans are input into a pre-trained PanSegNet model to extract pancreas regions. To reduce computational cost, linear self-attention is used at the bottleneck (highlighted in yellow) instead of standard self-attention. (C) Radiomics analysis and deep learning classification assess the malignancy risk of IPMNs: In the radiomics pipeline, scans undergo resizing, bias field correction, and intensity normalization, followed by textural features extraction and a random forest classification model. In the deep learning pipeline, DenseNet-121 serves as the backbone for malignancy risk prediction.
  • Figure 2: Cyst-X Segmentation Results. Annotation masks are displayed in red. Model outputs are shown in green and yellow, with green indicating the overlap between the annotation masks and the model outputs.
  • Figure 3: t-SNE Visualization. 0: No risk. 1: Low risk. 2: High risk.
  • Figure 4: ROC curves for Multi-Center IPMN MRI binary classification Using DenseNet-121.
  • Figure 5: MRQy analysis. Row 1: T1W modality MRQy of each center. Row 2: T1W modality MRQy of thickness range. Row 3: T2W modality MRQy of each center. Row 4: T2W modality MRQy of thickness range.
  • ...and 2 more figures