Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

SFNet: A Spatial-Frequency Domain Deep Learning Network for Efficient Alzheimer's Disease Diagnosis

Created by
  • Haebom

Author

Xinyue Yang, Meiliang Liu, Yunfang Xu, Xiaoxiao Yang, Zhengye Si, Zijin Li, Zhiwen Zhao

Outline

In this paper, we propose a novel deep learning framework, Spatio-Frequency Network (SFNet), based on 3D MRI for early diagnosis of Alzheimer's disease (AD). SFNet improves the accuracy of AD diagnosis by simultaneously utilizing information in both spatial and frequency domains. Unlike previous studies that utilized only one of the spatial or frequency domains or were limited to 2D MRI, SFNet is the first end-to-end deep learning model that utilizes both spatial and frequency information of 3D MRI. It extracts local spatial features through an enhanced Dense Convolutional Network, captures global frequency-domain representations through a global frequency module, and improves spatial feature extraction through a multi-scale attention module. Experimental results using the ADNI dataset show that SFNet achieves higher accuracy (95.1%) and reduces computational costs compared to existing methods.

Takeaways, Limitations

•
Takeaways:
◦
Improving the accuracy of Alzheimer's disease diagnosis by simultaneously utilizing spatial and frequency domain information of 3D MRI.
◦
Achieves higher accuracy (95.1%) and lower computational cost than existing methods.
◦
Presenting an efficient AD diagnosis model through an end-to-end deep learning framework.
◦
Improved spatial feature extraction via multi-scale attention module.
•
Limitations:
◦
Only performance evaluations on the ADNI dataset are presented, requiring further study on generalizability.
◦
Applicability to other MRI datasets or other neurodegenerative diseases needs to be verified.
◦
Further research is needed on the interpretability of the model.
◦
Despite the high accuracy of 95.1%, further validation is needed for practical clinical application.
👍
Made with Slashpage