Extensive Evaluation of Morphological Statistical Harmonization for Brain Age Prediction

Abstract

Characterizing both neurodevelopmental and aging brain structural trajectories is important for understanding normal biological processes and atypical patterns that are related to pathological phenomena. Initiatives to share open access morphological data contributed significantly to the advance in brain structure characterization. Indeed, such initiatives allow large brain morphology multi-site datasets to be shared, which increases the statistical sensitivity of the outcomes. However, using neuroimaging data from multi-site studies requires harmonizing data across the site to avoid bias. In this work we evaluated three different harmonization techniques on the Autism Brain Imaging Data Exchange (ABIDE) dataset for age prediction analysis in two groups of subjects (i.e., controls and autism spectrum disorder). We extracted the morphological features from T1-weighted images of a mixed cohort of 654 subjects acquired from 17 sites to predict the biological age of the subjects using three machine learning regression models. A machine learning framework was developed to quantify the effects of the different harmonization strategies on the final performance of the models and on the set of morphological features that are relevant to the age prediction problem in both the presence and absence of pathology. The results show that, even if two harmonization strategies exhibit similar accuracy of predictive models, a greater mismatch occurs between the sets of most age-related predictive regions for the Autism Spectrum Disorder (ASD) subjects. Thus, we propose to use a stability index to extract meaningful features for a robust clinical validation of the outcomes of multiple harmonization strategies.

Brain Age Prediction With Morphological Features Using Deep Neural Networks: Results From Predictive Analytic Competition 2019

“…For example, ( 3 ) obtained a MAE = 4 year by using a sample with subjects aged from 45 to 91, while ( 44 ) reported lower accuracy for the older group with MAE ranging between 1.57 (for the 8−18 age range) and 5.5 (for the oldest group in 65–96 age range) with neural networks using all the morphological descriptors. In our very recent works we obtained MAE = 2.2 with complex network modeling ( 7 ) and MAE = 2.5 with morphological features ( 21 ) on ABIDE dataset (6–40 years).…”

“…As in our previous work ( 21 ), we exploited ReCaS datacenter 2 to create a custom pipeline for preprocessing and analysis of T1 raw images ( 22 ). The ReCaS-Bari computing farm has been built by the ReCaS project 3 , funded by the Italian Research Ministry of Education, University and Research to the University of Bari and INFN (National Institute for Nuclear Physics) and offers a complete scientific high-throughput and high- performance computing environment to deal with common problems of large-scale neuroimaging processing.We integrated the software tool FreeSurfer 4 into a pipeline to extract the morphometric properties of both cortical and sub-cortical brain structures.…”

“…In this work, we present the results of the predictive model based on deep neural networks (DNN) proposed during the Predictive Analytic Challenge 2019 for brain age prediction of healthy individuals by using some morphological descriptors extracted from their raw MRI scans. Recently we have used a set of morphological features to describe the trajectories of neurodevelopment on a cohort of ABIDE database ( 20 ), proving the efficiency of this representation for brain age prediction in a limited age range ( 21 ). In this paper we propose a different architecture and a machine learning framework for a more in-depth comparison with other machine learning techniques commonly used in the literature.…”

Association between Structural Connectivity and Generalized Cognitive Spectrum in Alzheimer’s Disease

“…In fact, a feature selection algorithm might be sensitive with respect to changes in the training set, yielding subsets of features not representative of the overall population under investigation [ 54 ]. The assessment of the stability of the selected features over the rounds was thus carried out to select the list of more stable features with respect to small changes in the training sets taken from the whole sample distribution [ 55 , 56 ].…”

Explainable Deep Learning for Personalized Age Prediction With Brain Morphology

“…In this study, we exploited the same dataset used in our previous work (Lombardi et al, 2020b). In particular, we selected T = 378 T1-weighted MRI publicly available scans of a cohort of typically-developing individuals from the Autism Brain Imaging Data Exchange (ABIDE I) collected from 17 international sites.…”