Development of a histologically validated segmentation protocol for the hippocampal body

Immunohistochemical field parcellation of the human hippocampus along its antero-posterior axis

The primate hippocampus includes the dentate gyrus, cornu ammonis (CA), and subiculum. CA is subdivided into four fields (CA1-CA3, plus CA3h/hilus of the dentate gyrus) with specific pyramidal cell morphology and connections. Work in non-human mammals has shown that hippocampal connectivity is precisely patterned both in the laminar and longitudinal axes. One of the main handicaps in the study of neuropathological semiology in the human hippocampus is the lack of clear laminar and longitudinal borders. The aim of this study was to explore a histochemical segmentation of the adult human hippocampus, integrating field (medio-lateral), laminar, and anteroposterior longitudinal patterning. We provide criteria for head-body-tail field and subfield parcellation of the human hippocampus based on immunodetection of Rabphilin3a (Rph3a), Purkinje-cell protein 4 (PCP4), Chromogranin A and Regulation of G protein signaling-14 (RGS-14). Notably, Rph3a and PCP4 allow to identify the border between CA3 and CA2, while Chromogranin A and RGS-14 give specific staining of CA2. We also provide novel histological data about the composition of human-specific regions of the anterior and posterior hippocampus. The data are given with stereotaxic coordinates along the longitudinal axis. This study provides novel insights for a detailed region-specific parcellation of the human hippocampus useful for human brain imaging and neuropathology.

The prosubiculum in the human hippocampus: A rostrocaudal, feature-driven, and systematic approach

The hippocampal subfield prosubiculum (ProS), is a conserved neuroanatomic region in mouse, monkey, and human. This area lies between CA1 and subiculum (Sub) and particularly lacks consensus on its boundaries; reports have varied on the description of its features and location. In this report, we review, refine, and evaluate four cytoarchitectural features that differentiate ProS from its neighboring subfields: (1) small neurons, (2) lightly stained neurons, (3) superficial clustered neurons, and (4) a cell sparse zone. ProS was delineated in all cases (n = 10). ProS was examined for its cytoarchitectonic features and location rostrocaudally, from the anterior head through the body in the hippocampus. The most common feature was small pyramidal neurons, which were intermingled with larger pyramidal neurons in ProS. We quantitatively measured ProS pyramidal neurons, which showed (average, width at pyramidal base = 14.31 µm, n = 400 per subfield). CA1 neurons averaged 15.57 µm and Sub neurons averaged 15.63 µm, both were significantly different than ProS (Kruskal-Wallis test, p < .0001). The other three features observed were lightly stained neurons, clustered neurons, and a cell sparse zone. Taken together, these findings suggest that ProS is an independent subfield, likely with distinct functional contributions to the broader interconnected hippocampal network. Our results suggest that ProS is a cytoarchitecturally varied subfield, both for features and among individuals. This diverse architecture in features and individuals for ProS could explain the long-standing complexity regarding the identification of this subfield.

Pentad: A reproducible cytoarchitectonic protocol and its application to parcellation of the human hippocampus

Introduction

The hippocampus is integral for learning and memory and is targeted by multiple diseases. Neuroimaging approaches frequently use hippocampal subfield volumes as a standard measure of neurodegeneration, thus making them an essential biomarker to study. Collectively, histologic parcellation studies contain various disagreements, discrepancies, and omissions. The present study aimed to advance the hippocampal subfield segmentation field by establishing the first histology based parcellation protocol, applied to n = 22 human hippocampal samples.

Methods

The protocol focuses on five cellular traits observed in the pyramidal layer of the human hippocampus. We coin this approach the pentad protocol. The traits were: chromophilia, neuron size, packing density, clustering, and collinearity. Subfields included were CA1, CA2, CA3, CA4, prosubiculum, subiculum, presubiculum, parasubiculum, as well as the medial (uncal) subfields Subu, CA1u, CA2u, CA3u, and CA4u. We also establish nine distinct anterior-posterior levels of the hippocampus in the coronal plane to document rostrocaudal differences.

Results

Applying the pentad protocol, we parcellated 13 subfields at nine levels in 22 samples. We found that CA1 had the smallest neurons, CA2 showed high neuronal clustering, and CA3 displayed the most collinear neurons of the CA fields. The border between presubiculum and subiculum was staircase shaped, and parasubiculum had larger neurons than presubiculum. We also demonstrate cytoarchitectural evidence that CA4 and prosubiculum exist as individual subfields.

Discussion

This protocol is comprehensive, regimented and supplies a high number of samples, hippocampal subfields, and anterior-posterior coronal levels. The pentad protocol utilizes the gold standard approach for the human hippocampus subfield parcellation.

Integrated 3d flow-based multi-atlas brain structure segmentation

MRI brain structure segmentation plays an important role in neuroimaging studies. Existing methods either spend much CPU time, require considerable annotated data, or fail in segmenting volumes with large deformation. In this paper, we develop a novel multi-atlas-based algorithm for 3D MRI brain structure segmentation. It consists of three modules: registration, atlas selection and label fusion. Both registration and label fusion leverage an integrated flow based on grayscale and SIFT features. We introduce an effective and efficient strategy for atlas selection by employing the accompanying energy generated in the registration step. A 3D sequential belief propagation method and a 3D coarse-to-fine flow matching approach are developed in both registration and label fusion modules. The proposed method is evaluated on five public datasets. The results show that it has the best performance in almost all the settings compared to competitive methods such as ANTs, Elastix, Learning to Rank and Joint Label Fusion. Moreover, our registration method is more than 7 times as efficient as that of ANTs SyN, while our label transfer method is 18 times faster than Joint Label Fusion in CPU time. The results on the ADNI dataset demonstrate that our method is applicable to image pairs that require a significant transformation in registration. The performance on a composite dataset suggests that our method succeeds in a cross-modality manner. The results of this study show that the integrated 3D flow-based method is effective and efficient for brain structure segmentation. It also demonstrates the power of SIFT features, multi-atlas segmentation and classical machine learning algorithms for a medical image analysis task. The experimental results on public datasets show the proposed method’s potential for general applicability in various brain structures and settings.

Automated hippocampal unfolding for morphometry and subfield segmentation with HippUnfold

Like neocortical structures, the archicortical hippocampus differs in its folding patterns across individuals. Here, we present an automated and robust BIDS-App, HippUnfold, for defining and indexing individual-specific hippocampal folding in MRI, analogous to popular tools used in neocortical reconstruction. Such tailoring is critical for inter-individual alignment, with topology serving as the basis for homology. This topological framework enables qualitatively new analyses of morphological and laminar structure in the hippocampus or its subfields. It is critical for refining current neuroimaging analyses at a meso- as well as micro-scale. HippUnfold uses state-of-the-art deep learning combined with previously developed topological constraints to generate uniquely folded surfaces to fit a given subject's hippocampal conformation. It is designed to work with commonly employed sub-millimetric MRI acquisitions, with possible extension to microscopic resolution. In this paper, we describe the power of HippUnfold in feature extraction, and highlight its unique value compared to several extant hippocampal subfield analysis methods.

Surface-based hippocampal subfield segmentation

Though it is often termed 'subcortical,' the hippocampus is composed of a folded 'archicortical' sheet contiguous with the neocortex. The human hippocampus varies considerably in its internal folding configuration, creating major challenges in interindividual alignment and parcellation into subfields. In this opinion article, we discuss surface-based methods that aim to explicitly model hippocampal folding, similar to methods used in the neocortex, allowing interindividual alignment in an unfolded or flat-mapped 2D space. Such an approach enables detailed morphological characterization, constrains the problem of subfield segmentation, and provides a way to visualize data without occlusions. We argue that, when applied to magnetic resonance imaging (MRI) data, such methods overcome pitfalls of more conventional manual or registration-based subfield segmentation approaches.

Pixel-Wise Classification in Hippocampus Histological Images

This paper presents a method for pixel-wise classification applied for the first time on hippocampus histological images. The goal is achieved by representing pixels in a 14-D vector, composed of grey-level information and moment invariants. Then, several popular machine learning models are used to categorize them, and multiple metrics are computed to evaluate the performance of the different models. The multilayer perceptron, random forest, support vector machine, and radial basis function networks were compared, achieving the multilayer perceptron model the highest result on accuracy metric, AUC, and F ₁ score with highly satisfactory results for substituting a manual classification task, due to an expert opinion in the hippocampus histological images.

Clear and Consistent Imaging of Hippocampal Internal Architecture With High Resolution Multiple Image Co-registration and Averaging (HR-MICRA)

Magnetic resonance imaging of hippocampal internal architecture (HIA) at 3T is challenging. HIA is defined by layers of gray and white matter that are less than 1 mm thick in the coronal plane. To visualize HIA, conventional MRI approaches have relied on sequences with high in-plane resolution (≤0.5 mm) but comparatively thick slices (2–5 mm). However, thicker slices are prone to volume averaging effects that result in loss of HIA clarity and blurring of the borders of the hippocampal subfields in up to 61% of slices as has been reported. In this work we describe an approach to hippocampal imaging that provides consistently high HIA clarity using a commonly available sequence and post-processing techniques that is flexible and may be applicable to any MRI platform. We refer to this approach as High Resolution Multiple Image Co-registration and Averaging (HR-MICRA). This approach uses a variable flip angle turbo spin echo sequence to repeatedly acquire a whole brain T2w image volume with high resolution in three dimensions in a relatively short amount of time, and then co-register the volumes to correct for movement and average the repeated scans to improve SNR. We compared the averages of 4, 9, and 16 individual scans in 20 healthy controls using a published HIA clarity rating scale. In the body of the hippocampus, the proportion of slices with good or excellent HIA clarity was 90%, 83%, and 67% for the 16x, 9x, and 4x HR-MICRA images, respectively. Using the 4x HR-MICRA images as a baseline, the 9x HR-MICRA images were 2.6 times and 16x HR-MICRA images were 3.2 times more likely to have high HIA ratings (p < 0.001) across all hippocampal segments (head, body, and tail). The thin slices of the HR-MICRA images allow reformatting in any plane with clear visualization of hippocampal dentation in the sagittal plane. Clear and consistent visualization of HIA will allow application of this technique to future hippocampal structure research, as well as more precise manual or automated segmentation.

Hippocampal subfield measurement and ILAE hippocampal sclerosis subtype classification with in vivo 4.7 tesla MRI

OBJECTIVE

Neuropathological studies indicate that hippocampal sclerosis (HS) consists of three subtypes (ILAE types 1-3 HS). However, HS subtypes currently can only be diagnosed by pathological analysis of hippocampal tissue resected during epilepsy surgery or at autopsy. In vivo diagnosis of HS subtypes holds potential to improve our understanding of these variants in the ipsilateral as well as contralateral hippocampus. In this study, we aimed to: i) evaluate the reliability of our histology-derived segmentation protocol when applied to in vivo MRI; and ii) characterize variability of HS subtypes along the hippocampal long axis in patients with epilepsy.

METHODS

Eleven subjects with unilateral HS were compared with ten healthy controls. We used 4.7 T MRI to acquire high resolution MR Images of the hippocampus in each subject. In vivo MRI-based diagnoses of HS subtypes were then determined in each patient by two methods: i) hippocampal subfield volumetry of the entire hippocampal body; and ii) subfield area analysis at multiple thin slices throughout the hippocampal body.

RESULTS

Hippocampal body subfield segmentation demonstrated excellent reliability and volumetry of the symptomatic hippocampus revealed abnormalities in all eleven patients. Six subjects demonstrated findings consistent with type 1 HS while five subjects had volumetry-defined atypical HS (two with type 2 HS & three with type 3 HS) in the symptomatic hippocampus, while five subjects were found to have type 3 HS in the contralateral hippocampus. Subfield area analyses demonstrated remarkable variability of HS subtypes along the hippocampal long axis, both ipsilateral and contralateral to the seizure focus.

SIGNIFICANCE

Our results provide preliminary evidence that determining HS Subtype using in vivo MRI may allow preoperative diagnosis of ILAE HS subtypes. Further studies are essential to determine the pathological correlates of these neuroimaging findings. The heterogeneity of abnormalities observed along the long axis of the hippocampus is consistent with previous autopsy studies and highlights the necessity of studying the entire hippocampus both ipsilateral and contralateral to the seizure focus in these future studies.

Curved multiplanar reformatting provides improved visualization of hippocampal anatomy

There is a growing body of literature studying changes in hippocampal subfields in a variety of different neurological conditions, but this work has mainly focused on the hippocampal body given challenges in visualization of hippocampal anatomy in the head and tail when sectioned in the typical coronal image plane. Curved multiplanar reformatting (CMPR) is an image reconstruction method that can improve visualization of complex three‐dimensional structures. The objective of this study was to determine whether CMPR could facilitate visualization of the human hippocampal anatomy along the entire caudal–rostral axis. CMPR was applied to high‐resolution magnetic resonance imaging acquired ex vivo on four cadaveric hippocampal specimens at 4.7 T (T2‐weighted, 0.2 × 0.2 × 0.5 mm³). CMPR provided clear visualization of the classic “interlocking C” appearance of the dentate gyrus and cornu ammonis along the entire caudal–rostral axis including the head and tail, which otherwise show complex anatomy on the standard coronal slices. CMPR facilitated visualization of hippocampal anatomy providing the impetus to develop simplified approaches to delineate subfields along the entire hippocampus including the usually neglected head and tail.

Characterization of hippocampal subfields using ex vivo MRI and histology data: Lessons for in vivo segmentation

Hippocampal subfield segmentation on in vivo MRI is of great interest for cognition, aging, and disease research. Extant subfield segmentation protocols have been based on neuroanatomical references, but these references often give limited information on anatomical variability. Moreover, there is generally a mismatch between the orientation of the histological sections and the often anisotropic coronal sections on in vivo MRI. To address these issues, we provide a detailed description of hippocampal anatomy using a postmortem dataset containing nine specimens of subjects with and without dementia, which underwent a 9.4 T MRI and histological processing. Postmortem MRI matched the typical orientation of in vivo images and segmentations were generated in MRI space, based on the registered annotated histological sections. We focus on the following topics: the order of appearance of subfields, the location of subfields relative to macroanatomical features, the location of subfields in the uncus and tail and the composition of the dark band, a hypointense layer visible in T2-weighted MRI. Our main findings are that: (a) there is a consistent order of appearance of subfields in the hippocampal head, (b) the composition of subfields is not consistent in the anterior uncus, but more consistent in the posterior uncus, (c) the dark band consists only of the CA-stratum lacunosum moleculare, not the strata moleculare of the dentate gyrus, (d) the subiculum/CA1 border is located at the middle of the width of the hippocampus in the body in coronal plane, but moves in a medial direction from anterior to posterior, and (e) the variable location and composition of subfields in the hippocampal tail can be brought back to a body-like appearance when reslicing the MRI scan following the curvature of the tail. Our findings and this publicly available dataset will hopefully improve anatomical accuracy of future hippocampal subfield segmentation protocols.

Manual Hippocampal Subfield Segmentation Using High-Field MRI: Impact of Different Subfields in Hippocampal Volume Loss of Temporal Lobe Epilepsy Patients

In patients with temporal lobe epilepsy (TLE), presurgical magnetic resonance imaging (MRI) often reveals hippocampal atrophy, while neuropathological assessment indicates the different types of hippocampal sclerosis (HS). Different HS types are not discriminated in MRI so far. We aimed to define the volume of each hippocampal subfield on MRI manually and to compare automatic and manual segmentations for the discrimination of HS types. The T2-weighted images from 14 formalin-fixed age-matched control hippocampi were obtained with 4.7T MRI to evaluate the volume of each subfield at the anatomical level of the hippocampal head, body, and tail. Formalin-fixed coronal sections at the level of the body of 14 control cases, as well as tissue samples from 24 TLE patients, were imaged with a similar high-resolution sequence at 3T. Presurgical three-dimensional (3D) T1-weighted images from TLE went through a FreeSurfer 6.0 hippocampal subfield automatic assessment. The manual delineation with the 4.7T MRI was identified using Luxol Fast Blue stained 10-μm-thin microscopy slides, collected at every millimeter. An additional section at the level of the body from controls and TLE cases was submitted to NeuN immunohistochemistry for neuronal density estimation. All TLE cases were classified according to the International League Against Epilepsy's (ILAE's) HS classification. Manual volumetry in controls revealed that the dentate gyrus (DG)+CA4 region, CA1, and subiculum accounted for almost 90% of the hippocampal volume. The manual 3T volumetry showed that all TLE patients with type 1 HS (TLE-HS1) had lower volumes for DG+CA4, CA2, and CA1, whereas those TLE patients with HS type 2 (TLE-HS2) had lower volumes only in CA1 (p ≤ 0.038). Neuronal cell densities always decreased in CA4, CA3, CA2, and CA1 of TLE-HS1 but only in CA1 of TLE-HS2 (p ≤ 0.003). In addition, TLE-HS2 had a higher volume (p = 0.016) and higher neuronal density (p < 0.001) than the TLE-HS1 in DG + CA4. Automatic segmentation failed to match the manual or histological findings and was unable to differentiate TLE-HS1 from TLE-HS2. Total hippocampal volume correlated with DG+CA4 and CA1 volumes and neuronal density. For the first time, we also identified subfield-specific pathology patterns in the manual evaluation of volumetric MRI scans, showing the importance of manual segmentation to assess subfield-specific pathology patterns.