The goal of this project is to predict instances of Alzheimer's disease in patients after they have completed an MRI scan. By predicting who has Alzheimer's, we are able to save people by putting them into earlier preventative care treatment before the disease can progress. Especially with Alzheimer's, early detection is critical in saving patients. Through the use of various machine learning algorithms such as K-Nearest Neighbors, Support Vector Machines, and Linear Regression, this project provides various methods of prediction.
Of the various algorithms and methods used in this project, the one that had the lowest Root Mean Square Error was the K-Nearest Neighbors Algorithm with a RMSE of 0.2839809171235324. The Support Vector Machine got a RSME of 0.2978416985906353 and the linear regression (both regular and linear) got a RSME of 0.30399743103099924. However, all the algorithms are relatively close in their error and since the error lies between 0.2-0.5 for all of them, the algorithms are reliable.
Each of the algorithm takes inputs from the user for various features, ranging from simple ones like age and gender to complex features that can only be attained after an MRI scan like the Estimated Total Intracranial Volume or eTIV. Based on these features, the algorithm uses a combination of advanced mathematical models and computational tools to predict a result. Each algorithm will output either a 1 or 0 based on user inputs. A 1 corresponds to a prediction of Alzheimer's disease and a 0 to a prediction that the patient will not have Alzheimer's disease.
Try out the Algorithm by Clicking Here: Alzheimer's Predictor
