Prediction of Frozen Region Growth in Kidney Cryoablation Intervention Using a 3D Flow-Matching Model

2025-03-06

AI Summary

To address the challenge of dynamically predicting ice-ball growth during renal tumor cryoablation, this paper proposes an end-to-end differentiable 3D flow-matching model—the first to formulate ice-ball evolution as continuous deformation field learning, eliminating reliance on physics-based simulation or diffusion models. The method integrates intraoperative CT guidance, deep deformation learning, and joint segmentation to enable anatomy-adaptive, data-driven, simultaneous prediction of ice-ball volume expansion and morphological evolution. Evaluated on clinical data, it achieves an IoU of 0.61 and a Dice coefficient of 0.75—significantly outperforming conventional physics- and diffusion-based approaches. The framework supports real-time, high-accuracy intraoperative navigation, establishing a new paradigm for safe ablation and preservation of surrounding healthy tissue.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

This study presents a 3D flow-matching model designed to predict the progression of the frozen region (iceball) during kidney cryoablation. Precise intraoperative guidance is critical in cryoablation to ensure complete tumor eradication while preserving adjacent healthy tissue. However, conventional methods, typically based on physics driven or diffusion based simulations, are computationally demanding and often struggle to represent complex anatomical structures accurately. To address these limitations, our approach leverages intraoperative CT imaging to inform the model. The proposed 3D flow matching model is trained to learn a continuous deformation field that maps early-stage CT scans to future predictions. This transformation not only estimates the volumetric expansion of the iceball but also generates corresponding segmentation masks, effectively capturing spatial and morphological changes over time. Quantitative analysis highlights the model robustness, demonstrating strong agreement between predictions and ground-truth segmentations. The model achieves an Intersection over Union (IoU) score of 0.61 and a Dice coefficient of 0.75. By integrating real time CT imaging with advanced deep learning techniques, this approach has the potential to enhance intraoperative guidance in kidney cryoablation, improving procedural outcomes and advancing the field of minimally invasive surgery.

Problem

• Predict frozen region growth in kidney cryoablation.
• Overcome computational demands of conventional methods.
• Enhance intraoperative guidance using real-time CT imaging.

Innovation

• 3D flow-matching model predicts iceball growth
• Leverages intraoperative CT imaging for accuracy
• Deep learning captures spatial and morphological changes