Multimodality Fusion Method Based on Multiview Subspace Clustering for Pulmonary Embolism Diagnosis

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Clinical Diagnostics

Background:

• Pulmonary embolism (PE) is a critical condition requiring prompt diagnosis and treatment.
• Computed tomographic pulmonary angiography (CTPA) is the standard for PE diagnosis.
• Integrating CTPA with Electronic Medical Record (EMR) data shows promise for computer-aided PE detection.

Purpose of the Study:

• To propose and evaluate a novel multimodality fusion method for enhanced PE diagnosis.
• To leverage multi-view subspace clustering guided feature selection (MSCUFS) for optimal feature extraction and fusion.
• To compare the performance of the proposed method against traditional single-modality and fusion strategies.

Main Methods:

• Developed a multimodality fusion method (MSCUFS) integrating CTPA and EMR data.
• Employed multi-view subspace clustering for guided feature selection and fusion.
• Utilized machine learning models with varying fusion strategies for PE classification.

Main Results:

• The proposed MSCUFS method achieved a high Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.947.
• The joint fusion strategy with MSCUFS demonstrated superior performance compared to early and late fusion models.
• Multimodality analysis confirmed the effectiveness of the proposed approach over single-modality methods.

Conclusions:

• The MSCUFS-based multimodality fusion method offers a significant advancement in computer-aided diagnosis of pulmonary embolism.
• This approach effectively integrates imaging and clinical data for improved diagnostic accuracy.
• The findings highlight the potential of advanced feature selection and fusion techniques in medical diagnostics.