Integrating Transfer Learning and Attention Mechanisms for Accurate ALS Diagnosis and Cognitive Impairment Detection

Amyotrophic lateral sclerosis (ALS) is a fatal neurological disease characterized by motor deterioration and cognitive decline, leading to respiratory failure. Early diagnosis is crucial but challenging due to the undefined risk population and the complexity of sporadic ALS. In this study, we used a dataset of 190 autopsy brain images from the Gregory Laboratory at the University of Aberdeen to develop Miniset-DenseSENet, a convolutional neural network combining DenseNet121 with a Squeeze-and-Excitation (SE) attention mechanism. Our model not only distinguishes ALS patients from control groups with 97.37% accuracy but also detects cognitive impairments, which are increasingly recognized as a critical but underdiagnosed feature of ALS. Miniset-DenseSENet outperformed other transfer learning models, achieving a sensitivity of 1 and specificity of 0.95. These findings suggest that integrating transfer learning and attention mechanisms into neuroimaging analysis could enhance clinical diagnostic capabilities, enabling earlier and more accurate diagnosis of ALS and cognitive impair-ment. This approach has the potential to improve patient stratification, guide clinical decision-making, and inform the development of personalized therapeutic strategies.