Research on Alzheimer’s Disease Assisted Diagnosis Model Based on Deep Machine Learning for Corpus Cavernosum Segmentation and Plasma Biomarkers

Abstract

Early diagnosis of Alzheimer’s disease (AD) is key to improving prognosis, but existing methods have limitations. This article reviews the research on AD-assisted diagnosis based on deep learning sponge segmentation and plasma biomarker fusion. Firstly, elucidate the pathological mechanism and clinical characteristics of AD, and clarify the core value of the corpus cavernosum as an imaging biomarker and plasma biomarkers (such as A β and p-tau) as molecular markers. Next, analyze the technical foundation of deep learning in medical image segmentation and summarize its application progress in sponge segmentation. MRI is the main modality, and after preprocessing, models such as U-Net variants can achieve high-precision segmentation (Dice coefficient exceeding 0.85). At the same time, the application of deep learning in plasma biomarker screening, AD diagnosis, and other scenarios was reviewed, and the model AUC can reach 0.85~0.92. Multimodal fusion achieves macroscopic and microscopic pathological complementarity by integrating imaging and plasma data, significantly improving diagnostic efficiency. However, it faces challenges such as data heterogeneity, insufficient sample matching, and poor model interpretability. Finally, it is pointed out that the future needs to focus on the construction of standardized datasets, the development of lightweight fusion models, and clinical translation, in order to provide technical support for accurate diagnosis of AD.