There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
Principal Investigator | |
Principal Investigator's Name: | Yongmeng Zhang |
Institution: | Donghua University |
Department: | Computer Science and Technology |
Country: | |
Proposed Analysis: | Title:Research on Alzheimer's disease prediction method based on self attention and feature fusion. (Predict the deterioration of MCI patients) The following is the research content: Firstly, the expanded MRI and PET images are combined with the self attention mechanism for feature extraction, and then the MMSE and MRI features are fused, MOCA and pet features are fused. The fused two types of feature vectors are generated by adaptive prototyping, and finally the similarity score, that is, the classification result, is obtained. The reason why MRI and pet are used is that when the disease is in the stage of molecular level change in the early stage, the morphological structure of the lesion area has not been abnormal, and the diagnosis cannot be made clearly only by MRI, the lesion can be found through pet, three-dimensional images can be obtained, and quantitative analysis can be carried out to achieve early diagnosis. Shortcomings of PET: non multi sequence imaging, low resolution and limited image information. Therefore, the fused feature image combines the advantages of pet and MRI. It can more accurately locate the lesion area. The method of MMSE scale is simple and suitable for community grass-roots level, and the test frequency of MMSE in ADNI data set is very high, but some aspects are too simple, resulting in poor sensitivity to mild cognitive impairment. The sensitivity of MOCA scale to mild cognitive impairment is significantly higher than MMSE, but its test method is more complex and has certain requirements for the subject's education level. Considering the high resolution of MRI image and the low resolution of PET image, in order to prevent over fitting, MRI and MMSE fusion of weak features, pet and MOCA fusion of strong features are combined |
Additional Investigators |