Learning Multi-Modal Volumetric Prostate Registration with Weak Inter-Subject Spatial Correspondence

TL;DR

This paper presents a weakly-supervised CNN approach for multi-modal prostate registration using segmentation masks, improving accuracy and efficiency over traditional methods, and incorporating prior prostate location information for enhanced performance.

Contribution

The study introduces a novel weakly-supervised learning framework combining segmentation similarity measures and prior location input for prostate registration, reducing reliance on high-quality ground truth data.

Findings

Achieved registration accuracy comparable to state-of-the-art methods.

Demonstrated the effectiveness of combining mDSC and SDM similarity measures.

Significantly improved registration performance with prior prostate location input.

Abstract

Recent studies demonstrated the eligibility of convolutional neural networks (CNNs) for solving the image registration problem. CNNs enable faster transformation estimation and greater generalization capability needed for better support during medical interventions. Conventional fully-supervised training requires a lot of high-quality ground truth data such as voxel-to-voxel transformations, which typically are attained in a too tedious and error-prone manner. In our work, we use weakly-supervised learning, which optimizes the model indirectly only via segmentation masks that are a more accessible ground truth than the deformation fields. Concerning the weak supervision, we investigate two segmentation similarity measures: multiscale Dice similarity coefficient (mDSC) and the similarity between segmentation-derived signed distance maps (SDMs). We show that the combination of mDSC and SDM similarity measures results in a more accurate and natural transformation pattern together with a stronger gradient coverage. Furthermore, we introduce an auxiliary input to the neural network for the prior information about the prostate location in the MR sequence, which mostly is available preoperatively. This approach significantly outperforms the standard two-input models. With weakly labelled MR-TRUS prostate data, we showed registration quality comparable to the state-of-the-art deep learning-based method.