Can we accurately classify schizophrenia patients from healthy controls using magnetic resonance imaging and machine learning? A multi-method and multi-dataset study.

Machine learning is a powerful tool that has previously been used to classify schizophrenia (SZ) patients from healthy controls (HC) using magnetic resonance images. Each study, however, uses different datasets, classification algorithms, and validation techniques. Here, we perform a critical appraisal of the accuracy of machine learning methodologies used in SZ/HC classifications studies by comparing three machine learning algorithms (logistic regression [LR], support vector machines [SVMs], and linear discriminant analysis [LDA]) on three independent datasets (435 subjects total) using two tissue density estimates and cortical thickness (CT). Performance is assessed using 10-fold cross-validation, as well as a held-out validation set. Classification using CT outperformed tissue densities, but there was no clear effect of dataset. LR, SVMs, and LDA each yielded the highest accuracies for a different feature set and validation paradigm, but most accuracies were between 55 and 70%, well below previously reported values. The highest accuracy achieved was 73.5% using CT data and an SVM. Taken together, these results illustrate some of the obstacles to constructing effective disease classifiers, and suggest that tissue densities and CT may not be sufficiently sensitive for SZ/HC classification given current available methodologies and sample sizes.

A harmonized segmentation protocol for hippocampal and parahippocampal subregions: Why do we need one and what are the key goals?

The advent of high-resolution magnetic resonance imaging (MRI) has enabled in vivo research in a variety of populations and diseases on the structure and function of hippocampal subfields and subdivisions of the parahippocampal gyrus. Because of the many extant and highly discrepant segmentation protocols, comparing results across studies is difficult. To overcome this barrier, the Hippocampal Subfields Group was formed as an international collaboration with the aim of developing a harmonized protocol for manual segmentation of hippocampal and parahippocampal subregions on high-resolution MRI. In this commentary we discuss the goals for this protocol and the associated key challenges involved in its development. These include differences among existing anatomical reference materials, striking the right balance between reliability of measurements and anatomical validity, and the development of a versatile protocol that can be adopted for the study of populations varying in age and health. The commentary outlines these key challenges, as well as the proposed solution of each, with concrete examples from our working plan. Finally, with two examples, we illustrate how the harmonized protocol, once completed, is expected to impact the field by producing measurements that are quantitatively comparable across labs and by facilitating the synthesis of findings across different studies. © 2016 Wiley Periodicals, Inc.

Manual-Protocol Inspired Technique for Improving Automated MR Image Segmentation during Label Fusion.

Recent advances in multi-atlas based algorithms address many of the previous limitations in model-based and probabilistic segmentation methods. However, at the label fusion stage, a majority of algorithms focus primarily on optimizing weight-maps associated with the atlas library based on a theoretical objective function that approximates the segmentation error. In contrast, we propose a novel method-Autocorrecting Walks over Localized Markov Random Fields (AWoL-MRF)-that aims at mimicking the sequential process of manual segmentation, which is the gold-standard for virtually all the segmentation methods. AWoL-MRF begins with a set of candidate labels generated by a multi-atlas segmentation pipeline as an initial label distribution and refines low confidence regions based on a localized Markov random field (L-MRF) model using a novel sequential inference process (walks). We show that AWoL-MRF produces state-of-the-art results with superior accuracy and robustness with a small atlas library compared to existing methods. We validate the proposed approach by performing hippocampal segmentations on three independent datasets: (1) Alzheimer's Disease Neuroimaging Database (ADNI); (2) First Episode Psychosis patient cohort; and (3) A cohort of preterm neonates scanned early in life and at term-equivalent age. We assess the improvement in the performance qualitatively as well as quantitatively by comparing AWoL-MRF with majority vote, STAPLE, and Joint Label Fusion methods. AWoL-MRF reaches a maximum accuracy of 0.881 (dataset 1), 0.897 (dataset 2), and 0.807 (dataset 3) based on Dice similarity coefficient metric, offering significant performance improvements with a smaller atlas library (< 10) over compared methods. We also evaluate the diagnostic utility of AWoL-MRF by analyzing the volume differences per disease category in the ADNI1: Complete Screening dataset. We have made the source code for AWoL-MRF public at: https://github.com/CobraLab/AWoL-MRF.

Hippocampal (subfield) volume and shape in relation to cognitive performance across the adult lifespan.

Newer approaches to characterizing hippocampal morphology can provide novel insights regarding cognitive function across the lifespan. We comprehensively assessed the relationships among age, hippocampal morphology, and hippocampal-dependent cognitive function in 137 healthy individuals across the adult lifespan (18-86 years of age). They underwent MRI, cognitive assessments and genotyping for Apolipoprotein E status. We measured hippocampal subfield volumes using a new multiatlas segmentation tool (MAGeT-Brain) and assessed vertex-wise (inward and outward displacements) and global surface-based descriptions of hippocampus morphology. We examined the effects of age on hippocampal morphology, as well as the relationship among age, hippocampal morphology, and episodic and working memory performance. Age and volume were modestly correlated across hippocampal subfields. Significant patterns of inward and outward displacement in hippocampal head and tail were associated with age. The first principal shape component of the left hippocampus, characterized by a lengthening of the antero-posterior axis was prominently associated with working memory performance across the adult lifespan. In contrast, no significant relationships were found among subfield volumes and cognitive performance. Our findings demonstrate that hippocampal shape plays a unique and important role in hippocampal-dependent cognitive aging across the adult lifespan, meriting consideration as a biomarker in strategies targeting the delay of cognitive aging.

Quantitative comparison of 21 protocols for labeling hippocampal subfields and parahippocampal subregions in in vivo MRI: towards a harmonized segmentation protocol.

OBJECTIVE: An increasing number of human in vivo magnetic resonance imaging (MRI) studies have focused on examining the structure and function of the subfields of the hippocampal formation (the dentate gyrus, CA fields 1-3, and the subiculum) and subregions of the parahippocampal gyrus (entorhinal, perirhinal, and parahippocampal cortices). The ability to interpret the results of such studies and to relate them to each other would be improved if a common standard existed for labeling hippocampal subfields and parahippocampal subregions. Currently, research groups label different subsets of structures and use different rules, landmarks, and cues to define their anatomical extents. This paper characterizes, both qualitatively and quantitatively, the variability in the existing manual segmentation protocols for labeling hippocampal and parahippocampal substructures in MRI, with the goal of guiding subsequent work on developing a harmonized substructure segmentation protocol. METHOD: MRI scans of a single healthy adult human subject were acquired both at 3 T and 7 T. Representatives from 21 research groups applied their respective manual segmentation protocols to the MRI modalities of their choice. The resulting set of 21 segmentations was analyzed in a common anatomical space to quantify similarity and identify areas of agreement. RESULTS: The differences between the 21 protocols include the region within which segmentation is performed, the set of anatomical labels used, and the extents of specific anatomical labels. The greatest overall disagreement among the protocols is at the CA1/subiculum boundary, and disagreement across all structures is greatest in the anterior portion of the hippocampal formation relative to the body and tail. CONCLUSIONS: The combined examination of the 21 protocols in the same dataset suggests possible strategies towards developing a harmonized subfield segmentation protocol and facilitates comparison between published studies.

Automatic segmentation of the hippocampus for preterm neonates from early-in-life to term-equivalent age.

INTRODUCTION: The hippocampus, a medial temporal lobe structure central to learning and memory, is particularly vulnerable in preterm-born neonates. To date, segmentation of the hippocampus for preterm-born neonates has not yet been performed early-in-life (shortly after birth when clinically stable). The present study focuses on the development and validation of an automatic segmentation protocol that is based on the MAGeT-Brain (Multiple Automatically Generated Templates) algorithm to delineate the hippocampi of preterm neonates on their brain MRIs acquired at not only term-equivalent age but also early-in-life. METHODS: First, we present a three-step manual segmentation protocol to delineate the hippocampus for preterm neonates and apply this protocol on 22 early-in-life and 22 term images. These manual segmentations are considered the gold standard in assessing the automatic segmentations. MAGeT-Brain, automatic hippocampal segmentation pipeline, requires only a small number of input atlases and reduces the registration and resampling errors by employing an intermediate template library. We assess the segmentation accuracy of MAGeT-Brain in three validation studies, evaluate the hippocampal growth from early-in-life to term-equivalent age, and study the effect of preterm birth on the hippocampal volume. The first experiment thoroughly validates MAGeT-Brain segmentation in three sets of 10-fold Monte Carlo cross-validation (MCCV) analyses with 187 different groups of input atlases and templates. The second experiment segments the neonatal hippocampi on 168 early-in-life and 154 term images and evaluates the hippocampal growth rate of 125 infants from early-in-life to term-equivalent age. The third experiment analyzes the effect of gestational age (GA) at birth on the average hippocampal volume at early-in-life and term-equivalent age using linear regression. RESULTS: The final segmentations demonstrate that MAGeT-Brain consistently provides accurate segmentations in comparison to manually derived gold standards (mean Dice's Kappa > 0.79 and Euclidean distance <1.3 mm between centroids). Using this method, we demonstrate that the average volume of the hippocampus is significantly different (p < 0.0001) in early-in-life (621.8 mm(3)) and term-equivalent age (958.8 mm(3)). Using these differences, we generalize the hippocampal growth rate to 38.3 ± 11.7 mm(3)/week and 40.5 ± 12.9 mm(3)/week for the left and right hippocampi respectively. Not surprisingly, younger gestational age at birth is associated with smaller volumes of the hippocampi (p = 0.001). CONCLUSIONS: MAGeT-Brain is capable of segmenting hippocampi accurately in preterm neonates, even at early-in-life. Hippocampal asymmetry with a larger right side is demonstrated on early-in-life images, suggesting that this phenomenon has its onset in the 3rd trimester of gestation. Hippocampal volume assessed at the time of early-in-life and term-equivalent age is linearly associated with GA at birth, whereby smaller volumes are associated with earlier birth.

Derivation of high-resolution MRI atlases of the human cerebellum at 3T and segmentation using multiple automatically generated templates.

The cerebellum has classically been linked to motor learning and coordination. However, there is renewed interest in the role of the cerebellum in non-motor functions such as cognition and in the context of different neuropsychiatric disorders. The contribution of neuroimaging studies to advancing understanding of cerebellar structure and function has been limited, partly due to the cerebellum being understudied as a result of contrast and resolution limitations of standard structural magnetic resonance images (MRI). These limitations inhibit proper visualization of the highly compact and detailed cerebellar foliations. In addition, there is a lack of robust algorithms that automatically and reliably identify the cerebellum and its subregions, further complicating the design of large-scale studies of the cerebellum. As such, automated segmentation of the cerebellar lobules would allow detailed population studies of the cerebellum and its subregions. In this manuscript, we describe a novel set of high-resolution in vivo atlases of the cerebellum developed by pairing MR imaging with a carefully validated manual segmentation protocol. Using these cerebellar atlases as inputs, we validate a novel automated segmentation algorithm that takes advantage of the neuroanatomical variability that exists in a given population under study in order to automatically identify the cerebellum, and its lobules. Our automatic segmentation results demonstrate good accuracy in the identification of all lobules (mean Kappa [κ]=0.731; range 0.40-0.89), and the entire cerebellum (mean κ=0.925; range 0.90-0.94) when compared to "gold-standard" manual segmentations. These results compare favorably in comparison to other publically available methods for automatic segmentation of the cerebellum. The completed cerebellar atlases are available freely online (http://imaging-genetics.camh.ca/cerebellum) and can be customized to the unique neuroanatomy of different subjects using the proposed segmentation pipeline (https://github.com/pipitone/MAGeTbrain).

A novel in vivo atlas of human hippocampal subfields using high-resolution 3 T magnetic resonance imaging.

The hippocampus is a neuroanatomical structure that has been widely studied in the context of learning, memory, stress, and neurodegeneration. Neuroanatomically, the hippocampus is subdivided into several subfields with intricate morphologies and complex three-dimensional relationships. Recent studies have demonstrated that the identification of different subfields is possible with high-resolution and -contrast image volumes acquired using ex vivo specimens in a small bore 9.4 T scanner and, more recently, in vivo, at 7 T. In these studies, the neuroanatomical definitions of boundaries between subfields are based upon salient differences in image contrast. Typically, the definition of subfields has not been possible using commonly available magnetic resonance (MR) scanners (i.e.: 1.5 or 3T) due to resolution and contrast limitations. To overcome the limited availability of post-mortem specimens and expertise in state-of-the-art high-field imaging, we propose a coupling of MR acquisition and detailed segmentation techniques that allow for the reliable identification of hippocampal anatomy (including subfields). High-resolution and -contrast T1- and T2-weighted image volumes were acquired from 5 volunteers (2 male; 3 female; age range: 29-57, avg. 37) using a clinical research-grade 3T scanner and have final super-sampled isotropic voxel dimensions of 0.3mm. We demonstrate that by using these acquisition techniques, our data results in contrast-to-noise ratios that compare well with high-resolution images acquired with long scan times using post-mortem data at higher field strengths. For the subfields, the cornus ammonis (CA) 1, CA2/CA3, CA4/dentate gyrus, stratum radiatum/stratum lacunosum/stratum moleculare, and subiculum were all labeled as separate structures. Hippocampal volumes are reported for each of the substructures and the hippocampus as a whole (range for hippocampus: 2456.72-3325.02 mm(3)). Intra-rater reliability of our manual segmentation protocol demonstrates high reliability for the whole hippocampus (mean Dice Kappa of 0.91; range 0.90-0.92) and for each of the subfields (range of Dice Kappas: 0.64-0.83). We demonstrate that our reliability is better than the Dice Kappas produced by simulating the following errors: a translation by a single voxel in all cardinal directions and 1% volumetric shrinkage and expansion. The completed hippocampal atlases are available freely online (info2.camh.net/kf-tigr/index.php/Hippocampus) and can be coupled with novel computational neuroanatomy techniques that will allow for them to be customized to the unique neuroanatomy of different subjects, and ultimately be utilized in different analysis pipelines.