PENGWIN 2026: Peripelvic Fracture Segmentation and Reduction Planning Challenge

Peripelvic fractures are typically caused by high-energy trauma and are among the most severe injuries, with disability rates exceeding 50% and mortality rates over 13%, making them the deadliest among all complex fractures. The intricate anatomy of the pelvis and the surrounding soft tissues render surgical interventions particularly challenging. In recent years, there has been a shift from open reduction to image-guided manual or robot-assisted fracture reduction surgeries, which have demonstrated improved surgical outcomes. Consequently, an intelligent surgical planning pipeline is crucial. By integrating automated segmentation with reduction planning, the comprehensive workflow significantly enhances the efficiency, objectivity, and precision of surgical decision-making compared to traditional labor-intensive and expertise-demanding manual approaches.

Due to the diverse shapes and positions of bone fragments and the complex fracture surfaces resulting from bone collisions, segmenting peripelvic fracture fragments is technically challenging. Furthermore, determining the optimal geometric restoration of these multiple spatially separated fragments represents a high-dimensional, non-convex matching problem, often serving as a major bottleneck in surgical planning. Although deep learning has driven significant progress in general segmentation and reconstruction tasks within computer vision, applying these techniques directly to surgical workflows is still fraught with unique challenges. This is primarily due to the lack of algorithms specifically tailored for these complex orthopedic tasks and the scarcity of extensive image datasets with comprehensive annotations and ground-truth reduction poses.

In 2026, the PENGWIN Challenge will consist of three tasks aimed at evaluating and advancing state-of-the-art techniques: developing fully automated segmentation methods (Task 1), interactive segmentation methods (Task 2), and fracture reduction planning methods (Task 3) for 3D CT of peripelvic fractures. Our dataset has been significantly expanded to include CT scans from 500 clinical patients (incorporating 350 new cases) scheduled for peripelvic reduction surgery, collected from seven institutions using various scanning devices, representing a wide range of patient populations and fracture types. To support the training of data-driven reduction models (Task 3), we additionally provide 16,000 simulated fracture cases generated from 400 healthy peripelvic CTs using advanced fracture simulation algorithms. To ensure accurate performance evaluation, multiple orthopedic medical experts conducted comprehensive annotations and cross-checks on the training, validation, and testing datasets. The primary objective of the PENGWIN Challenge is to foster innovative research in intelligent surgical planning while establishing foundational resources for future peripelvic-related studies and interdisciplinary collaborations.