Classification of Alzheimer's disease using Ricci flow-based spherical parameterization and machine learning techniques

Magnetic Resonance Imaging (MRI) is an imaging tool employed to analyze brain structures, aiding in diagnosis and treatment planning. Alzheimer's disease (AD), a progressive neurodegenerative disorder leading to memory and cognitive function impairments, is the primary cause of dementia. Early...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Signal, image and video processing image and video processing, 2024-09, Vol.18 (10), p.6529-6545
Hauptverfasser: Khodaei, Masoumeh, Bidabad, Behroz, Shiri, Mohammad Ebrahim, Sedaghat, Maral Khadem, Amirifard, Hamed
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Magnetic Resonance Imaging (MRI) is an imaging tool employed to analyze brain structures, aiding in diagnosis and treatment planning. Alzheimer's disease (AD), a progressive neurodegenerative disorder leading to memory and cognitive function impairments, is the primary cause of dementia. Early detection of Mild Cognitive Impairment (MCI), a precursor to AD, is crucial for timely treatment. Diagnosis of atrophy, especially in the hippocampus, as a reliable biomarker for early diagnosis of Alzheimer's disease can be done using an MRI scan. This paper presents a new method for AD detection utilizing discrete surface Ricci flow theory, which creates Riemannian metrics on hippocampus surfaces with user-defined Gaussian curvatures. First, the surface of the hippocampus is extracted from the brain's subcortical surface. Then, Euclidean Ricci flow is applied to map this surface onto a sphere. Edge lengths of the triangular mesh are calculated, resulting in a feature vector. This vector is used as input for a classifier that distinguishes between brains affected by AD and healthy ones. The model is trained using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, and its accuracy, precision, sensitivity, and specificity performance are evaluated. Experimental results show an accuracy rate of over 90% in classifying AD and healthy hippocampus. The multiclass classification model achieves impressive performance metrics, with accuracy, precision, sensitivity, and specificity at 83, 80, 83, and 82%, respectively. These results are exceptionally satisfactory and outperform alternative methods.
ISSN:1863-1703
1863-1711
DOI:10.1007/s11760-024-03296-w