GA-MADRID: design and validation of a machine learning tool for the diagnosis of Alzheimer's disease and frontotemporal dementia using genetic algorithms.

Artificial Intelligence aids early diagnosis and development of new treatments, which is key to slow down the progress of the diseases, which to date have no cure. The patients' evaluation is carried out through diagnostic techniques such as clinical assessments neuroimaging techniques, which provide high-dimensionality data. In this work, a computational tool is presented that deals with the data provided by the clinical diagnostic techniques. This is a Python-based framework implemented with a modular design and fully extendable. It integrates (i) data processing and management of missing values and outliers; (ii) implementation of an evolutionary feature engineering approach, developed as a Python package, called PyWinEA using Mono-objective and Multi-objetive Genetic Algorithms (NSGAII); (iii) a module for designing predictive models based on a wide range of machine learning algorithms; (iv) a multiclass decision stage based on evolutionary grammars and Bayesian networks. Developed under the eXplainable Artificial Intelligence and open science perspective, this framework provides promising advances and opens the door to the understanding of neurodegenerative diseases from a data-centric point of view. In this work, we have successfully evaluated the potential of the framework for early and automated diagnosis with neuroimages and neurocognitive assessments from patients with Alzheimer's disease (AD) and frontotemporal dementia (FTD).

Diagnosis of Alzheimer's disease and behavioural variant frontotemporal dementia with machine learning-aided neuropsychological assessment using feature engineering and genetic algorithms.

BACKGROUND: Neuropsychological assessment is considered a valid tool in the diagnosis of neurodegenerative disorders. However, there is an important overlap in cognitive profiles between Alzheimer's disease (AD) and behavioural variant frontotemporal dementia (bvFTD), and the usefulness in diagnosis is uncertain. We aimed to develop machine learning-based models for the diagnosis using cognitive tests. METHODS: Three hundred and twenty-nine participants (170 AD, 72 bvFTD, 87 healthy control [HC]) were enrolled. Evolutionary algorithms, inspired by the process of natural selection, were applied for both mono-objective and multi-objective classification and feature selection. Classical algorithms (NativeBayes, Support Vector Machines, among others) were also used, and a meta-model strategy. RESULTS: Accuracies for the diagnosis of AD, bvFTD and the differential diagnosis between them were higher than 84%. Algorithms were able to significantly reduce the number of tests and scores needed. Free and Cued Selective Reminding Test, verbal fluency and Addenbrooke's Cognitive Examination were amongst the most meaningful tests. CONCLUSIONS: Our study found high levels of accuracy for diagnosis using exclusively neuropsychological tests, which supports the usefulness of cognitive assessment in diagnosis. Machine learning may have a role in improving the interpretation and test selection.

Genetic Algorithms for Optimized Diagnosis of Alzheimer's Disease and Frontotemporal Dementia Using Fluorodeoxyglucose Positron Emission Tomography Imaging.

Genetic algorithms have a proven capability to explore a large space of solutions, and deal with very large numbers of input features. We hypothesized that the application of these algorithms to 18F-Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) may help in diagnosis of Alzheimer's disease (AD) and Frontotemporal Dementia (FTD) by selecting the most meaningful features and automating diagnosis. We aimed to develop algorithms for the three main issues in the diagnosis: discrimination between patients with AD or FTD and healthy controls (HC), differential diagnosis between behavioral FTD (bvFTD) and AD, and differential diagnosis between primary progressive aphasia (PPA) variants. Genetic algorithms, customized with K-Nearest Neighbor and BayesNet Naives as the fitness function, were developed and compared with Principal Component Analysis (PCA). K-fold cross validation within the same sample and external validation with ADNI-3 samples were performed. External validation was performed for the algorithms distinguishing AD and HC. Our study supports the use of FDG-PET imaging, which allowed a very high accuracy rate for the diagnosis of AD, FTD, and related disorders. Genetic algorithms identified the most meaningful features with the minimum set of features, which may be relevant for automated assessment of brain FDG-PET images. Overall, our study contributes to the development of an automated, and optimized diagnosis of neurodegenerative disorders using brain metabolism.

Machine learning in the clinical and language characterisation of primary progressive aphasia variants.

INTRODUCTION: Primary progressive aphasia (PPA) is a clinical syndrome of neurodegenerative origin with 3 main variants: non-fluent, semantic, and logopenic. However, there is some controversy about the existence of additional subtypes. Our aim was to study the language and cognitive features associated with a new proposed classification for PPA. MATERIAL AND METHODS: Sixty-eight patients with PPA in early stages of the disease and 20 healthy controls were assessed with a comprehensive language and cognitive protocol. They were also evaluated with 18F-FDG positron emision tomography (PET). Patients were classified according to FDG PET regional metabolism, using our previously developed algorithm based on a hierarchical agglomerative cluster analysis with Ward's linkage method. Five variants were found, with both the non-fluent and logopenic variants being split into 2 subtypes. Machine learning techniques were used to predict each variant according to language assessment results. RESULTS: Non-fluent type 1 was associated with poorer performance in repetition of sentences and reading of irregular words than non-fluent type 2. Conversely, the second group showed a higher degree of apraxia of speech. Patients with logopenic variant type 1 performed more poorly on action naming than patients with logopenic type 2. Language assessments were predictive of PET-based subtypes in 86%-89% of cases using clustering analysis and principal components analysis. CONCLUSIONS: Our study supports the existence of 5 variants of PPA. These variants show some differences in language and FDG PET imaging characteristics. Machine learning algorithms using language test data were able to predict each of the 5 PPA variants with a relatively high degree of accuracy, and enable the possibility of automated, machine-aided diagnosis of PPA variants.

Clustering Analysis of FDG-PET Imaging in Primary Progressive Aphasia.

Background: Primary progressive aphasia (PPA) is a clinical syndrome characterized by the neurodegeneration of language brain systems. Three main clinical forms (non-fluent, semantic, and logopenic PPA) have been recognized, but applicability of the classification and the capacity to predict the underlying pathology is controversial. We aimed to study FDG-PET imaging data in a large consecutive case series of patients with PPA to cluster them into different subtypes according to regional brain metabolism. Methods: 122 FDG-PET imaging studies belonging to 91 PPA patients and 28 healthy controls were included. We developed a hierarchical agglomerative cluster analysis with Ward's linkage method, an unsupervised clustering algorithm. We conducted voxel-based brain mapping analysis to evaluate the patterns of hypometabolism of each identified cluster. Results: Cluster analysis confirmed the three current PPA variants, but the optimal number of clusters according to Davies-Bouldin index was 6 subtypes of PPA. This classification resulted from splitting non-fluent variant into three subtypes, while logopenic PPA was split into two subtypes. Voxel-brain mapping analysis displayed different patterns of hypometabolism for each PPA group. New subtypes also showed a different clinical course and were predictive of amyloid imaging results. Conclusion: Our study found that there are more than the three already recognized subtypes of PPA. These new subtypes were more predictive of clinical course and showed different neuroimaging patterns. Our results support the usefulness of FDG-PET in evaluating PPA, and the applicability of computational methods in the analysis of brain metabolism for improving the classification of neurodegenerative disorders.