Deep Neural Network for Alzheimer's disease detection
Author
Cacho Zurrunero, Raquel
Term
4. term
Publication year
2019
Abstract
Formålet med dette speciale er at udvikle et system, der viser, hvordan dyb læring kan bruges til at forbedre diagnosticeringen af Alzheimers sygdom. Dyb læring er en metode inden for kunstig intelligens, der lærer mønstre i data. Vi gennemgår litteraturen om sygdommen og de nyeste metoder i maskinlæring for at vurdere mulige løsninger og udarbejde et konceptuelt design. Den foreslåede løsning bygger på en Multilayer Perception (også kaldet Multilayer Perceptron), en type neuralt netværk, som beregner sandsynligheden for, at en patient har sygdommen, ud fra kliniske data. Løsningen implementeres og testes; i evalueringen opnås en nøjagtighed på 82,61% (andelen af korrekte forudsigelser).
This thesis develops a system that demonstrates how deep learning can help improve the diagnosis of Alzheimer’s disease. Deep learning is an AI method that learns patterns from data. We review the medical literature on the disease and recent machine learning techniques to assess possible solutions and produce a conceptual design. The proposed solution uses a Multilayer Perception (also known as a Multilayer Perceptron), a type of neural network, to estimate the probability that a patient has the disease based on clinical data. We then implement and test the solution; in our evaluation it achieves 82.61% accuracy (the share of correct predictions).
[This abstract was generated with the help of AI]
Documents