Deep Learning-based Anatomy-Aware Morph Model for Registration of Prostate Whole-Mount Histopathology to MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Pathology

Background:

• Accurate registration of presurgical prostate MRI and whole-mount histopathology (WMHP) images is crucial for precise cancer localization.
• Existing methods may face challenges in handling anatomical variations and image distortions.

Purpose of the Study:

• To develop and evaluate a novel deep learning-based approach for registering presurgical prostate MRI and WMHP images.
• To improve the accuracy and reliability of multimodal image registration in prostate cancer assessment.

Main Methods:

• A retrospective study utilizing ex vivo MRI as a reference for in vivo MRI and WMHP registration.
• Development of an Anatomy-Aware Morph model, a hybrid attention and convolutional neural network, for multimodality registration.
• Implementation of a pipeline to correct for distortion and motion in prostate specimens.

Main Results:

• The Anatomy-Aware Morph model achieved a Dice Similarity Coefficient (DSC) of 0.95 ± 0.06 and a Hausdorff distance of 1.84 mm ± 0.38.
• Significantly reduced target registration errors from 3.93 mm ± 0.80 to 1.18 mm ± 0.28 (P < .001) after registration.
• Outperformed the state-of-the-art VoxelMorph method in multimodality prostate image registration (P < .0001).

Conclusions:

• The developed deep learning method successfully aligns presurgical prostate MR and histopathology images.
• This facilitates automated mapping of prostate cancer from WMHP to MRI, enhancing surgical planning.
• The approach demonstrates superior performance compared to existing methods for prostate image registration.