Meningioma Analysis and Diagnosis using Limited Labeled Samples

TL;DR

It is observed that the weighted fusion of spatial-frequency domain features significantly influences meningioma classification performance, and a feature fusion architecture with adaptive weights of different frequency band information and spatial domain information is proposed for few-shot meningioma learning.

Abstract

The biological behavior and treatment response of meningiomas depend on their grade, making an accurate diagnosis essential for treatment planning and prognosis assessment. We observed that the weighted fusion of spatial-frequency domain features significantly influences meningioma classification performance. Notably, the contribution of specific frequency bands obtained by discrete wavelet transform varies considerably across different images. A feature fusion architecture with adaptive weights of different frequency band information and spatial domain information is proposed for few-shot meningioma learning. To verify the effectiveness of the proposed method, a new MRI dataset of meningiomas is introduced. The experimental results demonstrate the superiority of the proposed method compared with existing state-of-the-art methods in three datasets. The code will be available at: https://github.com/ICL-SUST/AMSF-Net

Figures (7)

• Figure 1: Timeline of the development of meningioma diagnosis and grading techniques. (a) Multimodal-based methods; (b) Unimodal-based methods.
• Figure 2: Workflow of data collection and standardization. (a) Participant recruitment and clinical data acquisition. (b) MRI data preprocessing pipeline from DICOM to standardized dataset.
• Figure 3: Overall architecture of the proposed adaptive multi-scale spatial-frequency network, which integrates the adaptive multi-scale feature fusion module, the adaptive cross-attention based spatial–frequency feature fusion module, the similary module.
• Figure 4: (a) Structure of the adaptive multi-scale feature feature fusion module. (b) Directional gating for adaptive fusion of LH, HL, and HH components. (c) Scale gating for multi-scale high-frequency feature fusion.
• Figure 5: The architecture of the proposed ACA-SFF module which integrates dual-branch cross-attention block.