Elucidating a Magnetic Resonance Imaging-Based Neuroanatomic Biomarker for Psychosis: Classification Analysis Using Probabilistic Brain Atlas and Machine Learning Algorithms

Abstract

Background No objective diagnostic biomarkers or laboratory tests have yet been developed for psychotic illness. Magnetic resonance imaging (MRI) studies consistently find significant abnormalities in multiple brain structures in psychotic patients relative to healthy control subjects, but these abnormalities show substantial overlap with anatomic variation that is in the normal range and therefore nondiagnostic. Recently, efforts have been made to discriminate psychotic patients from healthy individuals using machine-learning-based pattern classification methods on MRI data. Methods Three-dimensional cortical gray matter density (GMD) maps were generated for 36 patients with recent-onset psychosis and 36 sex- and age-matched control subjects using a cortical pattern matching method. Between-group differences in GMD were evaluated. Second, the sparse multinomial logistic regression classifier included in the Multivariate Pattern Analysis in Python machine-learning package was applied to the cortical GMD maps to discriminate psychotic patients from control subjects. Results Patients showed significantly lower GMD, particularly in prefrontal, cingulate, and lateral temporal brain regions. Pattern classification analysis achieved 86.1% accuracy in discriminating patients from controls using leave-one-out cross-validation. Conclusions These results suggest that even at the early stage of illness, psychotic patients present distinct patterns of regional cortical gray matter changes that can be discriminated from the normal pattern. These findings indicate that we can detect complex patterns of brain abnormality in early stages of psychotic illness, which has critical implications for early identification and intervention in individuals at ultra-high risk for developing psychosis/schizophrenia.

Brain Image Classification Based on Automated Morphometry and Penalised Linear Discriminant Analysis with Resampling

“…The efficiency of the algorithm is comparable or superior to the other state-of-the-art studies dealing with the classification of schizophrenia patients [5]- [18]. However, the classification performance is smaller than in [19], in which the COMPARE algorithm enabled classification of schizophrenia and healthy females with accuracy equal to 91.8% and the classification of diseased and healthy males with accuracy of 90.8%.…”

“…In [19] the COMPARE algorithm, classification of schizophrenia patients with very high classification accuracy (91.8% for female subjects and 90.8% for male subjects) was applied. Thus, the complex pipeline seems to enable classification with a higher efficiency than other commonly used methods that have reported classification accuracy between 70% and 90% [5]- [18].…”

“…Although pioneering works employing machine learning techniques have recently borne fruit in case of neurological diseases, this is extremely difficult in mental diseases. For instance, in schizophrenia -a disease with a complex neurobiology -the brain-imaging measurements in patients show considerable overlap with the normal range [5].…”

Recent developments in multivariate pattern analysis for functional MRI

“…Another alternative feature is coefficients from GLM using impulse response (IRF) basis functions [58,59] . In MVPA studies aiming to classify subjects into different groups, the choice of features is much wider, including but not limited to raw fMRI data [40] , grey matter density [70] , volumetric brain morphometry [71] , and functional connectivity measures [37,41] . The 'feature selection methods' are used to reduce the number of variables (dimensionality) to avoid over-fitting the data with limited training samples.…”

Multivariate Pattern Recognition for Diagnosis and Prognosis in Clinical Neuroimaging: State of the Art, Current Challenges and Future Trends

“…Second, predictive modeling at the level of the single subject is key to ultimately provide new neuroimaging markers with diagnostic value. Initially, whole-brain morphometry from structural MRI has been used to train models that can discriminate between healthy controls and patients, such as Alzheimer's disease and frontotemporal dementia (Kloppel et al 2008;Fan et al, 2008a, c;Davatzikos et al 2008), fragile-X syndrome (Hoeft et al 2008), psychosis (Davatzikos et al 2005;Fan et al 2008b), depression (Costafreda et al 2009), psychosis (Sun et al 2009), multiple sclerosis (Weygandt et al 2011) and so on. Advances in functional MRI, and more recently resting-state fMRI, have made it possible to study alterations in functional networks (Fox and Greicius 2010) without behavioral confounds (Bullmore 2012).…”