CSEPC: Deep Learning Framework for Small Sample Multimodal Medical Image Data in Alzheimer’s Disease Prediction

Abstract

Background Alzheimer’s disease (AD) is a neurodegenera­tive disorder that has a significant impact on global healthcare, especially among the elderly population. The prediction of its progression is crucial for slowing down the disease's progression and subsequent intervention management. However, the challenge of small sample sizes remains a significant obstacle in predicting the progression of AD. Methods In this study, we propose a novel diagnostic algorithm network architecture named cross-scale equilibrium pyramid coupling (CSEPC). This model adopts the scale equilibrium theory and integrates it with modal coupling properties, taking into account the comprehensive features of multimodal data. This structure not only enhances the feature representation of intermodal and intramodal information from multimodal data but also significantly reduces the number of learning parameters, making it better suited for small-sample characteristics. Results Through our experimental tests, our proposed model performs comparably or even superior to those from previous studies in conversion prediction and AD diagnosis. Our model achieves an accuracy (ACC) of 85.67% and an area under the curve (AUC) of 0.98 in predicting the progression from mild cognitive impairment (MCI) to AD. To further validate its efficacy, we used our method to perform diagnostic tasks for different stages of AD. In these two distinct AD classification tasks, our approach also achieved leading performance. Conclusions In conclusion, the performance of our model in various tasks has demonstrated its significant potential in the field of small-sample multimodal medical imaging classification, especially in the application of predicting the progression of Alzheimer's disease. This advancement could significantly assist clinicians in effectively managing and intervening in the disease progression of patients with early-stage Alzheimer's disease.