Deep Representation Learning for Unsupervised Clustering of Myocardial Fiber Trajectories in Cardiac Diffusion Tensor Imaging

2025-04-02

AI Summary

Cardiac diffusion tensor imaging (DTI) yields myocardial fiber tracts that are inherently ambiguous, highly interwoven, and lack ground-truth anatomical labels—posing a fundamental challenge for unsupervised fiber clustering. Method: We propose the first end-to-end unsupervised fiber clustering framework, built upon a novel BiLSTM-Transformer hybrid autoencoder that jointly captures local sequential dynamics and global geometric structure of fiber trajectories, augmented with point-wise anatomical priors; clustering is performed robustly using DBSCAN. Results: Applied to real cardiac DTI data, our method achieves stable, unsupervised identification of 33–62 anatomically meaningful fiber clusters—the first such result without supervision—demonstrating significantly improved inter-cluster separability and structural interpretability. This framework establishes a generalizable tool for quantitative microstructural analysis, surgical navigation, and assessment of disease-induced myocardial remodeling.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Segmentation

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

Understanding the complex myocardial architecture is critical for diagnosing and treating heart disease. However, existing methods often struggle to accurately capture this intricate structure from Diffusion Tensor Imaging (DTI) data, particularly due to the lack of ground truth labels and the ambiguous, intertwined nature of fiber trajectories. We present a novel deep learning framework for unsupervised clustering of myocardial fibers, providing a data-driven approach to identifying distinct fiber bundles. We uniquely combine a Bidirectional Long Short-Term Memory network to capture local sequential information along fibers, with a Transformer autoencoder to learn global shape features, with pointwise incorporation of essential anatomical context. Clustering these representations using a density-based algorithm identifies 33 to 62 robust clusters, successfully capturing the subtle distinctions in fiber trajectories with varying levels of granularity. Our framework offers a new, flexible, and quantitative way to analyze myocardial structure, achieving a level of delineation that, to our knowledge, has not been previously achieved, with potential applications in improving surgical planning, characterizing disease-related remodeling, and ultimately, advancing personalized cardiac care.

Problem

• Unsupervised clustering of myocardial fiber trajectories in cardiac DTI
• Lack of ground truth labels for myocardial fiber structure
• Ambiguous intertwined nature of fiber trajectories in DTI data

Innovation

• Bidirectional LSTM captures local fiber sequences
• Transformer autoencoder learns global shape features
• Density-based clustering identifies robust fiber bundles