MRI Plane Orientation Detection using a Context-Aware 2.5D Model

2025-11-17

AI Summary

MRI 2D slices frequently lack anatomical plane (axial/coronal/sagittal) metadata, causing domain shift across datasets, reduced analytical robustness, and degraded diagnostic performance. To address this, we propose a context-aware 2.5D convolutional neural network that mitigates single-slice ambiguity by fusing multi-sequence 3D contextual information from adjacent slices, and incorporates an uncertainty-guided gating mechanism to dynamically integrate reliable metadata. Our method achieves 99.49% accuracy in plane classification—reducing error rate by 60% versus standard 2D CNNs—and improves brain tumor detection accuracy from 97.0% to 98.0%, with a 33.3% reduction in false positives. An interactive web interface enables clinical deployment. This work is the first to jointly leverage uncertainty modeling and 2.5D architecture for anatomical plane identification, significantly enhancing the generalizability and reliability of AI systems in medical imaging.

Technology Category

• Computer Vision: Medical and Biological Imaging
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Humans can easily identify anatomical planes (axial, coronal, and sagittal) on a 2D MRI slice, but automated systems struggle with this task. Missing plane orientation metadata can complicate analysis, increase domain shift when merging heterogeneous datasets, and reduce accuracy of diagnostic classifiers. This study develops a classifier that accurately generates plane orientation metadata. We adopt a 2.5D context-aware model that leverages multi-slice information to avoid ambiguity from isolated slices and enable robust feature learning. We train the 2.5D model on both 3D slice sequences and static 2D images. While our 2D reference model achieves 98.74% accuracy, our 2.5D method raises this to 99.49%, reducing errors by 60%, highlighting the importance of 2.5D context. We validate the utility of our generated metadata in a brain tumor detection task. A gated strategy selectively uses metadata-enhanced predictions based on uncertainty scores, boosting accuracy from 97.0% with an image-only model to 98.0%, reducing misdiagnoses by 33.3%. We integrate our plane orientation model into an interactive web application and provide it open-source.

Problem

• Automated detection of MRI anatomical plane orientation from 2D slices
• Addressing missing metadata that complicates analysis and reduces classifier accuracy
• Developing context-aware model using multi-slice information to resolve slice ambiguity

Innovation

• 2.5D context-aware model using multi-slice information
• Training combines 3D slice sequences and static 2D images
• Gated strategy uses metadata-enhanced predictions with uncertainty scores