Related Experiment Video
Updated: May 25, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
938
Improving ALS detection and cognitive impairment stratification with attention-enhanced deep learning models.
Yuqing Xia1, Jenna M Gregory2, Fergal M Waldron2
1School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, UK.
Scientific Reports
|February 27, 2025
Summary
Researchers developed Miniset-DenseSENet, an AI model for early amyotrophic lateral sclerosis (ALS) detection. This tool accurately identifies ALS subtypes using brain images, potentially improving patient diagnosis and treatment strategies.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Medical Imaging
Background:
- Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease with significant diagnostic challenges.
- Early detection of ALS is hindered by disease complexity and overlapping symptoms with other neurological disorders.
- Cognitive impairment, particularly frontotemporal dementia, is a common comorbidity in ALS.
Purpose of the Study:
- To develop and validate an artificial intelligence model for accurate classification of ALS patients and controls using neuroimaging data.
- To investigate the efficacy of combining DenseNet121 with a Squeeze-and-Excitation attention mechanism for enhanced diagnostic performance.
- To differentiate between ALS patients with and without cognitive impairment (ALS-frontotemporal dementia).
Main Methods:
- Development of Miniset-DenseSENet, a convolutional neural network integrating DenseNet121 and Squeeze-and-Excitation attention.
- Utilized a dataset of 190 autopsy brain images from the Gregory Laboratory.
- Employed transfer learning techniques for model training and evaluation.
Main Results:
- Miniset-DenseSENet achieved an overall accuracy of 97.37% in distinguishing controls, ALS patients without cognitive impairment, and ALS patients with cognitive impairment.
- The model demonstrated superior performance compared to other transfer learning models, with a sensitivity of 1 and specificity of 0.95.
- Successfully addressed the challenge of differentiating overlapping neurodegenerative disorders characterized by TDP-43 proteinopathy.
Conclusions:
- Integrating transfer learning and attention mechanisms in neuroimaging analysis significantly enhances diagnostic accuracy for ALS.
- The developed model shows potential for earlier ALS detection and improved patient stratification.
- This AI-driven approach can aid clinical decision-making and support the development of personalized therapeutic strategies for ALS.
