Researchers developed a novel transfer learning-based machine learning model to diagnose and prognose mild cognitive impairment (MCI) due to Alzheimer’s disease (AD) with varying availability of imaging modalities from MRI, FDG-PET, and amyloid-PET. The model was discussed at the Radiological Society of North America 106th Scientific Assembly and Annual Meeting (RSNA 2020).
The study results demonstrate that this machine learning model can assist in early diagnosis of AD from multi-modality imaging, even when some modalities may be missing. Early detection is key to slowing AD progression. Neuroimaging holds great promise to facilitate early diagnosis and prognosis of AD in its early stage, but there are gaps in available neuroimaging data.
“Our research provides a clinical tool to assist physicians in diagnosis and prognosis of AD when the disease is still early, which has tremendous clinical benefits,” says study author Fleming Y. Lure, Ph.D.
The study included 214 MCI patients from the Alzheimer’s Disease Neuroimaging Initiative database: 97 MCI due to AD, 26 and 46 MCI converting to AD within 2 and 6 years, respectively. Patients were divided into four sub-cohorts based on the imaging available: MRI only; MRI & FDG-PET; MRI & amyloid-PET; and all three modalities.
The machine-learning model achieved much better accuracy than the competing model, using each cohort for the diagnosis and prognosis.
The study’s authors are: Jing Li, Teresa Wu, Kewei Chen, Ph.D., Dave Weidman, M.D., Yi Su, Ph.D., and Fleming Y. Lure, Ph.D.Back To Top
RSNA 2020: Machine Learning Model Aids Alzheimer’s Diagnosis, Prognosis. Appl Radiol.