Dataset Information

Progressive multifocal leukoencephalopathy lesion and brain parenchymal segmentation from MRI using serial deep convolutional neural networks.

ABSTRACT: Progressive multifocal leukoencephalopathy (PML) is a rare opportunistic brain infection caused by the JC virus and associated with substantial morbidity and mortality. Accurate MRI assessment of PML lesion burden and brain parenchymal atrophy is of decisive value in monitoring the disease course and response to therapy. However, there are currently no validated automatic methods for quantification of PML lesion burden or associated parenchymal volume loss. Furthermore, manual brain or lesion delineations can be tedious, require the use of valuable time resources by radiologists or trained experts, and are often subjective. In this work, we introduce JCnet (named after the causative viral agent), an end-to-end, fully automated method for brain parenchymal and lesion segmentation in PML using consecutive 3D patch-based convolutional neural networks. The network architecture consists of multi-view feature pyramid networks with hierarchical residual learning blocks containing embedded batch normalization and nonlinear activation functions. The feature maps across the bottom-up and top-down pathways of the feature pyramids are merged, and an output probability membership generated through convolutional pathways, thus rendering the method fully convolutional. Our results show that this approach outperforms and improves longitudinal consistency compared to conventional, state-of-the-art methods of healthy brain and multiple sclerosis lesion segmentation, utilized here as comparators given the lack of available methods validated for use in PML. The ability to produce robust and accurate automated measures of brain atrophy and lesion segmentation in PML is not only valuable clinically but holds promise toward including standardized quantitative MRI measures in clinical trials of targeted therapies. Code is available at: https://github.com/omarallouz/JCnet.

SUBMITTER: Al-Louzi O

PROVIDER: S-EPMC7708929 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Progressive multifocal leukoencephalopathy lesion and brain parenchymal segmentation from MRI using serial deep convolutional neural networks.

Al-Louzi Omar O Roy Snehashis S Osuorah Ikesinachi I Parvathaneni Prasanna P Smith Bryan R BR Ohayon Joan J Sati Pascal P Pham Dzung L DL Jacobson Steven S Nath Avindra A Reich Daniel S DS Cortese Irene I

NeuroImage. Clinical 20201111

Progressive multifocal leukoencephalopathy (PML) is a rare opportunistic brain infection caused by the JC virus and associated with substantial morbidity and mortality. Accurate MRI assessment of PML lesion burden and brain parenchymal atrophy is of decisive value in monitoring the disease course and response to therapy. However, there are currently no validated automatic methods for quantification of PML lesion burden or associated parenchymal volume loss. Furthermore, manual brain or lesion de ...[more]

PMID: 33395989

Similar Datasets

Project description:Background and purposeThere are no validated imaging criteria for the diagnosis of progressive multifocal leukoencephalopathy in the cerebellum. Here we introduce the MR imaging shrimp sign, a cerebellar white matter lesion identifiable in patients with cerebellar progressive multifocal leukoencephalopathy, and we evaluate its sensitivity and specificity.Materials and methodsWe first identified patients with progressive multifocal leukoencephalopathy seen at Massachusetts General Hospital between 1998 and 2019 whose radiology reports included the term "cerebellum." Drawing on a priori knowledge, 2 investigators developed preliminary diagnostic criteria for the shrimp sign. These criteria were revised and validated in 2 successive stages by 4 additional blinded investigators. After defining the MR imaging shrimp sign, we assessed its sensitivity, specificity, positive predictive value, and negative predictive value.ResultsWe identified 20 patients with cerebellar progressive multifocal leukoencephalopathy: 16 with definite progressive multifocal leukoencephalopathy (mean, 46.4 [SD, 9.2] years of age; 5 women), and 4 with possible progressive multifocal leukoencephalopathy (mean, 45.8 [SD, 8.5] years of age; 1 woman). We studied 40 disease controls (mean, 43.6 [SD, 21.0] years of age; 16 women) with conditions known to affect the cerebellar white matter. We defined the MR imaging shrimp sign as a T2- and FLAIR-hyperintense, T1-hypointense, discrete cerebellar white matter lesion abutting-but-sparing the dentate nucleus. MR imaging shrimp sign sensitivity was 0.85; specificity, 1; positive predictive value, 1; and negative predictive value, 0.93. The shrimp sign was also seen in fragile X-associated tremor ataxia syndrome, but radiographic and clinical features distinguished it from progressive multifocal leukoencephalopathy.ConclusionsIn the right clinical context, the MR imaging shrimp sign has excellent sensitivity and specificity for cerebellar progressive multifocal leukoencephalopathy, providing a new radiologic marker of the disease.

Project description:Progressive multifocal leukoencephalopathy (PML) is a rare demyelinating disorder of the brain caused by reactivation of the JC virus (JCV), a polyomavirus that infects at least 60% of the population but is asymptomatic or results in benign symptoms in most people. PML occurs as a secondary disease in a variety of disorders or as a serious adverse event from immunosuppressant agents, but is mainly found in three groups: HIV-infected patients, patients with hematological malignancies, or multiple sclerosis (MS) patients on the immunosuppressant therapy natalizumab. It is severely debilitating and is deadly in ~50% HIV cases, ~90% of hematological malignancy cases, and ~24% of MS-natalizumab cases. A PML risk prediction test would have clinical utility in all at risk patient groups but would be particularly beneficial in patients considering therapy with immunosuppressant agents known to cause PML, such as natalizumab, rituximab, and others. While a JC antibody test is currently used in the clinical decision process for natalizumab, it is suboptimal because of its low specificity and requirement to periodically retest patients for seroconversion or to assess if a patient's JCV index has increased. Whereas a high specificity genetic risk prediction test comprising host genetic risk variants (i.e., germline variants occurring at higher frequency in PML patients compared to the general population) could be administered one time to provide clinicians with additional risk prediction information that is independent of JCV serostatus. Prior PML case reports support the hypothesis that PML risk is greater in patients with a genetically caused immunodeficiency disorder. To identify germline PML risk variants, we performed exome sequencing on 185 PML cases (70 in a discovery cohort and 115 in a replication cohort) and used the gnomAD variant database for interpretation. Our study yielded 19 rare variants (maximum allele frequency of 0.02 in gnomAD ethnically matched populations) that impact 17 immune function genes (10 are known to cause inborn errors of immunity). Modeling of these variants in a PML genetic risk test for MS patients considering natalizumab treatment indicates that at least a quarter of PML cases may be preventable.

Dataset Information

Progressive multifocal leukoencephalopathy lesion and brain parenchymal segmentation from MRI using serial deep convolutional neural networks.

Publications

Progressive multifocal leukoencephalopathy lesion and brain parenchymal segmentation from MRI using serial deep convolutional neural networks.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets