BISON: Brain tissue segmentation pipeline using T1 -weighted magnetic resonance images and a random forest classifier

An hetero-modal deep learning framework for medical image synthesis applied to contrast and non-contrast MRI

Some pathologies such as cancer and dementia require multiple imaging modalities to fully diagnose and assess the extent of the disease. Magnetic resonance imaging offers this kind of polyvalence, but examinations take time and can require contrast agent injection. The flexible synthesis of these imaging sequences based on the available ones for a given patient could help reduce scan times or circumvent the need for contrast agent injection. In this work, we propose a deep learning architecture that can perform the synthesis of all missing imaging sequences from any subset of available images. The network is trained adversarially, with the generator consisting of parallel 3D U-Net encoders and decoders that optimally combines their multi-resolution representations with a fusion operation learned by an attention network trained conjointly with the generator network. We compare our synthesis performance with 3D networks using other types of fusion and a comparable number of trainable parameters, such as the mean/variance fusion. In all synthesis scenarios except one, the synthesis performance of the network using attention-guided fusion was better than the other fusion schemes. We also inspect the encoded representations and the attention network outputs to gain insights into the synthesis process, and uncover desirable behaviors such as prioritization of specific modalities, flexible construction of the representation when important modalities are missing, and modalities being selected in regions where they carry sequence-specific information. This work suggests that a better construction of the latent representation space in hetero-modal networks can be achieved by using an attention network.

Degeneration in Nucleus basalis of Meynert signals earliest stage of Alzheimer's disease progression

Nucleus Basalis of Meynert (NbM), a crucial source of cholinergic projection to the entorhinal cortex (EC) and hippocampus (HC), has shown sensitivity to neurofibrillary degeneration in the early stages of Alzheimer's Disease. Using deformation-based morphometry (DBM) on up-sampled MRI scans from 1447 Alzheimer's Disease Neuroimaging Initiative participants, we aimed to quantify NbM degeneration along the disease trajectory. Results from cross-sectional analysis revealed significant differences of NbM volume between cognitively normal and early mild cognitive impairment cohorts, confirming recent studies suggesting that NbM degeneration happens before degeneration in the EC or HC. Longitudinal linear mixed-effect models were then used to compare trajectories of volume change after realigning all participants into a common timeline based on their cognitive decline. Results indicated the earliest deviations in NbM volumes from the cognitively healthy trajectory, challenging the prevailing idea that Alzheimer's originates in the EC. Converging evidence from cross-sectional and longitudinal models suggest that the NbM may be a focal target of early AD progression, which is often obscured by normal age-related decline.

Structural white matter properties and cognitive resilience to tau pathology

INTRODUCTION

We assessed whether macro- and/or micro-structural white matter properties are associated with cognitive resilience to Alzheimer's disease pathology years prior to clinical onset.

METHODS

We examined whether global efficiency, an indicator of communication efficiency in brain networks, and diffusion measurements within the limbic network and default mode network moderate the association between amyloid-β/tau pathology and cognitive decline. We also investigated whether demographic and health/risk factors are associated with white matter properties.

RESULTS

Higher global efficiency of the limbic network, as well as free-water corrected diffusion measures within the tracts of both networks, attenuated the impact of tau pathology on memory decline. Education, age, sex, white matter hyperintensities, and vascular risk factors were associated with white matter properties of both networks.

DISCUSSION

White matter can influence cognitive resilience against tau pathology, and promoting education and vascular health may enhance optimal white matter properties.

HIGHLIGHTS

Aβ and tau were associated with longitudinal memory change over ∼7.5 years. White matter properties attenuated the impact of tau pathology on memory change. Health/risk factors were associated with white matter properties.

Improving across-dataset brain tissue segmentation for MRI imaging using transformer

Brain tissue segmentation has demonstrated great utility in quantifying MRI data by serving as a precursor to further post-processing analysis. However, manual segmentation is highly labor-intensive, and automated approaches, including convolutional neural networks (CNNs), have struggled to generalize well due to properties inherent to MRI acquisition, leaving a great need for an effective segmentation tool. This study introduces a novel CNN-Transformer hybrid architecture designed to improve brain tissue segmentation by taking advantage of the increased performance and generality conferred by Transformers for 3D medical image segmentation tasks. We first demonstrate the superior performance of our model on various T1w MRI datasets. Then, we rigorously validate our model's generality applied across four multi-site T1w MRI datasets, covering different vendors, field strengths, scan parameters, and neuropsychiatric conditions. Finally, we highlight the reliability of our model on test-retest scans taken in different time points. In all situations, our model achieved the greatest generality and reliability compared to the benchmarks. As such, our method is inherently robust and can serve as a valuable tool for brain related T1w MRI studies. The code for the TABS network is available at: https://github.com/raovish6/TABS.

Machine learning in neuro-oncology: toward novel development fields

PURPOSE

Artificial Intelligence (AI) involves several and different techniques able to elaborate a large amount of data responding to a specific planned outcome. There are several possible applications of this technology in neuro-oncology.

METHODS

We reviewed, according to PRISMA guidelines, available studies adopting AI in different fields of neuro-oncology including neuro-radiology, pathology, surgery, radiation therapy, and systemic treatments.

RESULTS

Neuro-radiology presented the major number of studies assessing AI. However, this technology is being successfully tested also in other operative settings including surgery and radiation therapy. In this context, AI shows to significantly reduce resources and costs maintaining an elevated qualitative standard. Pathological diagnosis and development of novel systemic treatments are other two fields in which AI showed promising preliminary data.

CONCLUSION

It is likely that AI will be quickly included in some aspects of daily clinical practice. Possible applications of these techniques are impressive and cover all aspects of neuro-oncology.

Ventricular features as reliable differentiators between bvFTD and other dementias

•

Our results showed a consistent pattern of ventricle enlargement in the bvFTD patients, particularly in the anterior parts of the frontal and temporal horns of the lateral ventricles.

•

The estimation of the proposed ventricular anteroposterior ratio (APR) resulted in statistically significant difference compared to all other groups.

•

Our study proposes an easy to obtain and generalizable ventricle-based feature (APR) from T1-weighted structural MRI (routinely acquired and available in the clinic) that can be used not only to differentiate bvFTD from normal subjects, but also from other FTD variants (SV and PNFA), MCI, and AD patients.

•

We have made our ventricle feature estimation and bvFTD diagnosis tool (VentRa) publicly available, allowing application of our model in other studies.

•

If validated in a prospective study, VentRa has the potential to aid bvFTD diagnosis, particularly in settings where access to specialized FTD care is limited.

Introduction

Lateral ventricles are reliable and sensitive indicators of brain atrophy and disease progression in behavioral variant frontotemporal dementia (bvFTD). We aimed to investigate whether an automated tool using ventricular features could improve diagnostic accuracy in bvFTD across neurodegenerative diseases.

Methods

Using 678 subjects −69 bvFTD, 38 semantic variant, 37 primary non-fluent aphasia, 218 amyloid + mild cognitive impairment, 74 amyloid + Alzheimer’s Dementia and 242 normal controls- with a total of 2750 timepoints, lateral ventricles were segmented and differences in ventricular features were assessed between bvFTD, normal controls and other dementia cohorts.

Results

Ventricular antero-posterior ratio (APR) was the only feature that was significantly different and increased faster in bvFTD compared to all other cohorts. We achieved a 10-fold cross-validation accuracy of 80% (77% sensitivity, 82% specificity) in differentiating bvFTD from all other cohorts with other ventricular features (i.e., total ventricular volume and left–right lateral ventricle ratios), and 76% accuracy using only the single APR feature.

Discussion

Ventricular features, particularly the APR, might be reliable and easy-to-implement markers for bvFTD diagnosis. We have made our ventricle feature estimation and bvFTD diagnostic tool publicly available, allowing application of our model in other studies.

Diffusely abnormal white matter converts to T2 lesion volume in the absence of MRI-detectable acute inflammation

Diffusely abnormal white matter (DAWM), characterised by biochemical changes of myelin in the absence of frank demyelination, has been associated with clinical progression in secondary progressive MS (SPMS). However, little is known about changes of DAWM over time and their relation to focal white matter lesions (FWML).

The objectives of this work were: 1) To characterize the longitudinal evolution of FWML, DAWM, and DAWM that transforms into FWML, and 2) To determine whether gadolinium enhancement, known to be associated with the development of new FWML, is also related to DAWM voxels that transform into FWML.

Our data included 4220 MRI scans of 689 SPMS participants, followed for 156 weeks and 2677 scans of 686 RRMS participants, followed for 96 weeks. FWML and DAWM were segmented using a previously validated, automatic thresholding technique based on normalized T2 intensity values. Using longitudinally registered images, DAWM voxels at each visit that transformed into FWML on the last MRI scan as well as their overlap with gadolinium enhancing lesion masks were identified.

Our results showed that the average yearly rate of conversion of DAWM-to-FWML was 1.27 cc for SPMS and 0.80 cc for RRMS. FWML in SPMS participants significantly increased (t = 3.9; p = 0.0001) while DAWM significantly decreased (t = −4.3 p < 0.0001) and the ratio FWML:DAWM increased (t = 12.7; p < 0.00001). RRMS participants also showed an increase in the FWML:DAWM Ratio (t = 6.9; p < 0.00001) but without a significant change of the individual volumes. Gadolinium enhancement was associated with 7.3% and 18.7% of focal New T2 lesion formation in the infrequent scans of the RRMS and SPMS cohorts, respectively. In comparison, only 0.1% and 0.0% of DAWM-to-FWML voxels overlapped with gadolinium enhancement.

We conclude that DAWM transforms into FWML over time, in both RRMS and SPMS. DAWM appears to represent a form of pre-lesional pathology that contributes to T2 lesion volume increase over time, independent of new focal inflammation and gadolinium enhancement.

Non-parametric MRI Brain Atlas for the Polish Population

Introduction: The application of magnetic resonance imaging (MRI) to acquire detailed descriptions of the brain morphology in vivo is a driving force in brain mapping research. Most atlases are based on parametric statistics, however, the empirical results indicate that the population brain tissue distributions do not exhibit exactly a Gaussian shape. Our aim was to verify the population voxel-wise distribution of three main tissue classes: gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF), and to construct the brain templates for the Polish (Upper Silesian) healthy population with the associated non-parametric tissue probability maps (TPMs) taking into account the sex and age influence. Material and Methods: The voxel-wise distributions of these tissues were analyzed using the Shapiro-Wilk test. The non-parametric atlases were generated from 96 brains of the ethnically homogeneous, neurologically healthy, and radiologically verified group examined in a 3-Tesla MRI system. The standard parametric tissue proportion maps were also calculated for the sake of comparison. The maps were compared using the Wilcoxon signed-rank test and Kolmogorov-Smirnov test. The volumetric results segmented with the parametric and non-parametric templates were also analyzed. Results: The results confirmed that in each brain structure (regardless of the studied sub-population) the data distribution is skewed and apparently not Gaussian. The determined non-parametric and parametric templates were statistically compared, and significant differences were found between the maps obtained using both measures (the maps of GM, WM, and CSF). The impacts of applying the parametric and non-parametric TPMs on the segmentation process were also compared. The GM volumes are significantly greater when using the non-parametric atlas in the segmentation procedure, while the CSF volumes are smaller. Discussion and Conclusion: To determine the population atlases the parametric measures are uncritically and widely used. However, our findings suggest that the mean and parametric measures of such skewed distribution may not be the most appropriate summary statistic to find the best spatial representations of the structures in a standard space. The non-parametric methodology is more relevant and universal than the parametric approach in constructing the MRI brain atlases.

MNI-FTD templates, unbiased average templates of frontotemporal dementia variants

Standard templates are widely used in human neuroimaging processing pipelines to facilitate group-level analyses and comparisons across subjects/populations. MNI-ICBM152 template is the most commonly used standard template, representing an average of 152 healthy young adult brains. However, in patients with neurodegenerative diseases such as frontotemporal dementia (FTD), high atrophy levels lead to significant differences between individuals' brain shapes and MNI-ICBM152 template. Such differences might inevitably lead to registration errors or subtle biases in downstream analyses and results. Disease-specific templates are therefore desirable to reflect the anatomical characteristics of the populations of interest and reduce potential registration errors. Here, we present MNI-FTD136, MNI-bvFTD70, MNI-svFTD36, and MNI-pnfaFTD30, four unbiased average templates of 136 FTD patients, 70 behavioural variant (bv), 36 semantic variant (sv), and 30 progressive nonfluent aphasia (pnfa) variant FTD patients and a corresponding age-matched template of 133 controls (MNI-CN133), along with probabilistic tissue maps for each template. Public availability of these templates will facilitate analyses of FTD cohorts and enable comparisons between different studies in an appropriate common standardized space.

Automatic brain extraction and hemisphere segmentation in rat brain MR images after stroke using deformable models

PURPOSE

Experimental ischemic stroke models play an essential role in understanding the mechanisms of cerebral ischemia and evaluating the development of pathological extent. An important precursor to the investigation of ischemic strokes associated with rodents is the brain extraction and hemisphere segmentation in rat brain diffusion-weighted imaging (DWI) and T2-weighted MRI (T2WI) images. Accurate and reliable image segmentation tools for extracting the rat brain and hemispheres in the MR images are critical in subsequent processes, such as lesion identification and injury analysis. This study is an attempt to investigate rat brain extraction and hemisphere segmentation algorithms that are practicable in both DWI and T2WI images.

METHODS

To automatically perform brain extraction, the proposed framework is based on an efficient geometric deformable model. By introducing an additional image force in response to the rat brain characteristics into the skull stripping model, we establish a unique rat brain extraction scheme in DWI and T2WI images. For the subsequent hemisphere segmentation, we develop an efficient brain feature detection algorithm to approximately separate the rat brain. A refinement process is enforced by constructing a gradient vector flow in the proximity of the midsurface, where a parametric active contour is attracted to achieve hemisphere segmentation.

RESULTS

Extensive experiments with 55 DWI and T2WI subjects were executed in comparison with the state-of-the-art methods. Experimental results indicated that our rat brain extraction and hemisphere segmentation schemes outperformed the competitive methods and exhibited high performance both qualitatively and quantitatively. For rat brain extraction, the average Dice scores were 97.13% and 97.42% in DWI and T2WI image volumes, respectively. Rat hemisphere segmentation results based on the Hausdorff distance metric revealed average values of 0.17 and 0.15 mm for DWI and T2WI subjects, respectively.

CONCLUSIONS

We believe that the established frameworks are advantageous to facilitate preclinical stroke investigation and relevant neuroscience research that requires accurate brain extraction and hemisphere segmentation using rat DWI and T2WI images.

Automated multiclass tissue segmentation of clinical brain MRIs with lesions

•

A U-Net incorporating spatial prior information can successfully segment 6 brain tissue types.

•

The U-Net was able to segment gray and white matter in the presence of lesions.

•

The U-Net surpassed the performance of its source algorithm in an external dataset.

•

Segmentations were produced in a hundredth of the time of its predecessor algorithm.

Delineation and quantification of normal and abnormal brain tissues on Magnetic Resonance Images is fundamental to the diagnosis and longitudinal assessment of neurological diseases. Here we sought to develop a convolutional neural network for automated multiclass tissue segmentation of brain MRIs that was robust at typical clinical resolutions and in the presence of a variety of lesions. We trained a 3D U-Net for full brain multiclass tissue segmentation from a prior atlas-based segmentation method on an internal dataset that consisted of 558 clinical T1-weighted brain MRIs (453/52/53; training/validation/test) of patients with one of 50 different diagnostic entities (n = 362) or with a normal brain MRI (n = 196). We then used transfer learning to refine our model on an external dataset that consisted of 7 patients with hand-labeled tissue types. We evaluated the tissue-wise and intra-lesion performance with different loss functions and spatial prior information in the validation set and applied the best performing model to the internal and external test sets. The network achieved an average overall Dice score of 0.87 and volume similarity of 0.97 in the internal test set. Further, the network achieved a median intra-lesion tissue segmentation accuracy of 0.85 inside lesions within white matter and 0.61 inside lesions within gray matter. After transfer learning, the network achieved an average overall Dice score of 0.77 and volume similarity of 0.96 in the external dataset compared to human raters. The network had equivalent or better performance than the original atlas-based method on which it was trained across all metrics and produced segmentations in a hundredth of the time. We anticipate that this pipeline will be a useful tool for clinical decision support and quantitative analysis of clinical brain MRIs in the presence of lesions.