ABSTRACT
BACKGROUND: Cortical lesion subtypes' occurrence and distribution across networks may shed light on cognitive impairment (CI) in multiple sclerosis (MS). METHODS: In 332 people with MS, lesions were classified as intracortical, leukocortical or juxtacortical based on artificially generated double inversion-recovery images. RESULTS: CI-related leukocortical lesion count increases were greatest within sensorimotor and cognitive networks (p < 0.001). Only intracortical lesion count could distinguish between cognitive groups (p = 0.024). Effect sizes were two- to four-fold larger than differences between MS phenotypes. CONCLUSION: In CI-MS, leukocortical lesions predominate, whereas intracortical lesions distinguish cognitive groups. Lesions' grey matter (GM) involvement might be decisive for cognition in MS, surpassing overall disease burden.
ABSTRACT
BACKGROUND: Extended interval dosing (EID) of natalizumab treatment is increasingly used in multiple sclerosis. Besides the clear anti-inflammatory effect, natalizumab is considered to have neuroprotective properties as well. OBJECTIVES: This study aimed to study the longitudinal effects of EID compared to standard interval dosing (SID) and natalizumab drug concentrations on brain atrophy. METHODS: Patients receiving EID or SID of natalizumab with a minimum radiological follow-up of 2 years were included. Changes in brain atrophy measures over time were derived from clinical routine 3D-Fluid Attenuated Inversion Recovery (FLAIR)-weighted magnetic resonance imaging (MRI) scans using SynthSeg. RESULTS: We found no differences between EID (n = 32) and SID (n = 50) for whole brain (-0.21% vs -0.16%, p = 0.42), ventricular (1.84% vs 1.13%, p = 0.24), and thalamic (-0.32% vs -0.32%, p = 0.97) annualized volume change over a median follow-up of 3.2 years. No associations between natalizumab drug concentration and brain atrophy rate were found. CONCLUSION: We found no clear evidence that EID compared to SID or lower natalizumab drug concentrations have a negative impact on the development of brain atrophy over time.
Subject(s)
Central Nervous System Diseases , Leukoencephalopathy, Progressive Multifocal , Multiple Sclerosis, Relapsing-Remitting , Multiple Sclerosis , Humans , Natalizumab/therapeutic use , Multiple Sclerosis/diagnostic imaging , Multiple Sclerosis/drug therapy , Multiple Sclerosis/chemically induced , Leukoencephalopathy, Progressive Multifocal/chemically induced , Brain/diagnostic imaging , Brain/pathology , Magnetic Resonance Imaging , Atrophy/pathology , Multiple Sclerosis, Relapsing-Remitting/drug therapy , Immunologic Factors/therapeutic useABSTRACT
Background Cortical multiple sclerosis lesions are clinically relevant but inconspicuous at conventional clinical MRI. Double inversion recovery (DIR) and phase-sensitive inversion recovery (PSIR) are more sensitive but often unavailable. In the past 2 years, artificial intelligence (AI) was used to generate DIR and PSIR from standard clinical sequences (eg, T1-weighted, T2-weighted, and fluid-attenuated inversion-recovery sequences), but multicenter validation is crucial for further implementation. Purpose To evaluate cortical and juxtacortical multiple sclerosis lesion detection for diagnostic and disease monitoring purposes on AI-generated DIR and PSIR images compared with MRI-acquired DIR and PSIR images in a multicenter setting. Materials and Methods Generative adversarial networks were used to generate AI-based DIR (n = 50) and PSIR (n = 43) images. The number of detected lesions between AI-generated images and MRI-acquired (reference) images was compared by randomized blinded scoring by seven readers (all with >10 years of experience in lesion assessment). Reliability was expressed as the intraclass correlation coefficient (ICC). Differences in lesion subtype were determined using Wilcoxon signed-rank tests. Results MRI scans of 202 patients with multiple sclerosis (mean age, 46 years ± 11 [SD]; 127 women) were retrospectively collected from seven centers (February 2020 to January 2021). In total, 1154 lesions were detected on AI-generated DIR images versus 855 on MRI-acquired DIR images (mean difference per reader, 35.0% ± 22.8; P < .001). On AI-generated PSIR images, 803 lesions were detected versus 814 on MRI-acquired PSIR images (98.9% ± 19.4; P = .87). Reliability was good for both DIR (ICC, 0.81) and PSIR (ICC, 0.75) across centers. Regionally, more juxtacortical lesions were detected on AI-generated DIR images than on MRI-acquired DIR images (495 [42.9%] vs 338 [39.5%]; P < .001). On AI-generated PSIR images, fewer juxtacortical lesions were detected than on MRI-acquired PSIR images (232 [28.9%] vs 282 [34.6%]; P = .02). Conclusion Artificial intelligence-generated double inversion-recovery and phase-sensitive inversion-recovery images performed well compared with their MRI-acquired counterparts and can be considered reliable in a multicenter setting, with good between-reader and between-center interpretative agreement. Published under a CC BY 4.0 license. Supplemental material is available for this article. See also the editorial by Zivadinov and Dwyer in this issue.
Subject(s)
Multiple Sclerosis , Humans , Female , Middle Aged , Multiple Sclerosis/diagnostic imaging , Multiple Sclerosis/pathology , Artificial Intelligence , Retrospective Studies , Reproducibility of Results , Magnetic Resonance Imaging/methodsABSTRACT
BACKGROUND: There is a need in Relapsing-Remitting Multiple Sclerosis (RRMS) treatment for biomarkers that monitor neuroinflammation, neurodegeneration, treatment response, and disease progression despite treatment. OBJECTIVE: To assess the value of serum glial fibrillary acidic protein (sGFAP) as a biomarker for clinical disease progression and brain volume measurements in natalizumab-treated RRMS patients. METHODS: sGFAP and neurofilament light (sNfL) were measured in an observational cohort of natalizumab-treated RRMS patients at baseline, +3, +12, and +24 months and at the last sample follow-up (median 5.17 years). sGFAP was compared between significant clinical progressors and non-progressors and related to magnetic resonance imaging (MRI)-derived volumes of the whole brain, ventricle, thalamus, and lesion. The relationship between sGFAP and sNfL was assessed. RESULTS: A total of 88 patients were included, and 47.7% progressed. sGFAP levels at baseline were higher in patients with gadolinium enhancement (1.3-fold difference, p = 0.04) and decreased in 3 months of treatment (adj. p < 0.001). No association was found between longitudinal sGFAP levels and progressor status. sGFAP at baseline and 12 months was significantly associated with normalized ventricular (positively), thalamic (negatively), and lesion volumes (positively). Baseline and 12-month sGFAP predicted annualized ventricle volume change rate after 1 year of treatment. sGFAP correlated with sNfL at baseline (p < 0.001) and last sample follow-up (p < 0.001) but stabilized earlier. DISCUSSION: sGFAP levels related to MRI markers of neuroinflammation and neurodegeneration.
Subject(s)
Multiple Sclerosis, Relapsing-Remitting , Humans , Biomarkers , Contrast Media/metabolism , Disease Progression , Gadolinium , Glial Fibrillary Acidic Protein , Intermediate Filaments/metabolism , Multiple Sclerosis, Relapsing-Remitting/diagnostic imaging , Multiple Sclerosis, Relapsing-Remitting/drug therapy , Multiple Sclerosis, Relapsing-Remitting/metabolism , Natalizumab/therapeutic use , Neurofilament Proteins , Neuroinflammatory DiseasesABSTRACT
PURPOSE: Volume measurement using MRI is important to assess brain atrophy in multiple sclerosis (MS). However, differences between scanners, acquisition protocols, and analysis software introduce unwanted variability of volumes. To quantify theses effects, we compared within-scanner repeatability and between-scanner reproducibility of three different MR scanners for six brain segmentation methods. METHODS: Twenty-one people with MS underwent scanning and rescanning on three 3 T MR scanners (GE MR750, Philips Ingenuity, Toshiba Vantage Titan) to obtain 3D T1-weighted images. FreeSurfer, FSL, SAMSEG, FastSurfer, CAT-12, and SynthSeg were used to quantify brain, white matter and (deep) gray matter volumes both from lesion-filled and non-lesion-filled 3D T1-weighted images. We used intra-class correlation coefficient (ICC) to quantify agreement; repeated-measures ANOVA to analyze systematic differences; and variance component analysis to quantify the standard error of measurement (SEM) and smallest detectable change (SDC). RESULTS: For all six software, both between-scanner agreement (ICCs ranging 0.4-1) and within-scanner agreement (ICC range: 0.6-1) were typically good, and good to excellent (ICC > 0.7) for large structures. No clear differences were found between filled and non-filled images. However, gray and white matter volumes did differ systematically between scanners for all software (p < 0.05). Variance component analysis yielded within-scanner SDC ranging from 1.02% (SAMSEG, whole-brain) to 14.55% (FreeSurfer, CSF); and between-scanner SDC ranging from 4.83% (SynthSeg, thalamus) to 29.25% (CAT12, thalamus). CONCLUSION: Volume measurements of brain, GM and WM showed high repeatability, and high reproducibility despite substantial differences between scanners. Smallest detectable change was high, especially between different scanners, which hampers the clinical implementation of atrophy measurements.
Subject(s)
Multiple Sclerosis , Humans , Multiple Sclerosis/diagnostic imaging , Multiple Sclerosis/pathology , Gray Matter/pathology , Cross-Sectional Studies , Reproducibility of Results , Brain/diagnostic imaging , Brain/pathology , Magnetic Resonance Imaging/methods , Atrophy/pathology , SoftwareABSTRACT
BACKGROUND: Despite highly effective treatment strategies for patients with relapsing-remitting multiple sclerosis (RRMS), long-term neurodegeneration and disease progression are often considerable. Accurate blood-based biomarkers that predict long-term neurodegeneration are lacking. OBJECTIVE: To assess the predictive value of serum neurofilament-light (sNfL) and serum contactin-1 (sCNTN1) for long-term magnetic resonance imaging (MRI)-derived neurodegeneration in natalizumab-treated patients with RRMS. METHODS: sNfL and sCNTN1 were measured in an observational cohort of natalizumab-treated patients with RRMS at baseline (first dose) and at 3 months, Year 1, Year 2, and last follow-up (median = 5.2 years) of treatment. Disability progression was quantified using "EDSS-plus" criteria. Neurodegeneration was measured by calculating annualized percentage brain, ventricular, and thalamic volume change (PBVC, VVC, and TVC, respectively). Linear regression analysis was performed to identify longitudinal predictors of neurodegeneration. RESULTS: In total, 88 patients (age = 37 ± 9 years, 75% female) were included, of whom 48% progressed. Year 1 sNfL level (not baseline or 3 months) was associated with PBVC (standardized (std.) ß = -0.26, p = 0.013), VVC (standardized ß = 0.36, p < 0.001), and TVC (standardized ß = -0.24, p = 0.02). For sCNTN1, only 3-month level was associated with VVC (standardized ß = -0.31, p = 0.002). CONCLUSION: Year 1 (but not baseline) sNfL level was predictive for long-term brain atrophy in patients treated with natalizumab. sCNTN1 level did not show a clear predictive value.
Subject(s)
Brain , Contactin 1 , Multiple Sclerosis, Relapsing-Remitting , Adult , Female , Humans , Male , Middle Aged , Atrophy , Brain/diagnostic imaging , Brain/pathology , Multiple Sclerosis, Relapsing-Remitting/diagnostic imaging , Multiple Sclerosis, Relapsing-Remitting/drug therapy , Natalizumab/adverse effects , Contactin 1/metabolismABSTRACT
BACKGROUND: Robust predictive models of clinical impairment and worsening in multiple sclerosis (MS) are needed to identify patients at risk and optimize treatment strategies. OBJECTIVE: To evaluate whether machine learning (ML) methods can classify clinical impairment and predict worsening in people with MS (pwMS) and, if so, which combination of clinical and magnetic resonance imaging (MRI) features and ML algorithm is optimal. METHODS: We used baseline clinical and structural MRI data from two MS cohorts (Berlin: n = 125, Amsterdam: n = 330) to evaluate the capability of five ML models in classifying clinical impairment at baseline and predicting future clinical worsening over a follow-up of 2 and 5 years. Clinical worsening was defined by increases in the Expanded Disability Status Scale (EDSS), Timed 25-Foot Walk Test (T25FW), 9-Hole Peg Test (9HPT), or Symbol Digit Modalities Test (SDMT). Different combinations of clinical and volumetric MRI measures were systematically assessed in predicting clinical outcomes. ML models were evaluated using Monte Carlo cross-validation, area under the curve (AUC), and permutation testing to assess significance. RESULTS: The ML models significantly determined clinical impairment at baseline for the Amsterdam cohort, but did not reach significance for predicting clinical worsening over a follow-up of 2 and 5 years. High disability (EDSS ≥ 4) was best determined by a support vector machine (SVM) classifier using clinical and global MRI volumes (AUC = 0.83 ± 0.07, p = 0.015). Impaired cognition (SDMT Z-score ≤ -1.5) was best determined by a SVM using regional MRI volumes (thalamus, ventricles, lesions, and hippocampus), reaching an AUC of 0.73 ± 0.04 (p = 0.008). CONCLUSION: ML models could aid in classifying pwMS with clinical impairment and identify relevant biomarkers, but prediction of clinical worsening is an unmet need.
Subject(s)
Disease Progression , Machine Learning , Magnetic Resonance Imaging , Multiple Sclerosis , Humans , Male , Female , Adult , Multiple Sclerosis/diagnostic imaging , Multiple Sclerosis/classification , Multiple Sclerosis/physiopathology , Multiple Sclerosis/diagnosis , Middle Aged , Cohort Studies , Brain/diagnostic imaging , Brain/physiopathology , Disability Evaluation , Follow-Up StudiesABSTRACT
BACKGROUND: Cognitive impairment, a common and debilitating symptom in people with multiple sclerosis (MS), is especially related to cortical damage. However, the impact of regional cortical damage remains poorly understood. Our aim was to evaluate structural (network) integrity in lesional and non-lesional cortex in people with MS, and its relationship with cognitive dysfunction. METHODS: In this cross-sectional study, 176 people with MS and 48 healthy controls underwent MRI, including double inversion recovery and diffusion-weighted scans, and neuropsychological assessment. Cortical integrity was assessed based on fractional anisotropy (FA) and mean diffusivity (MD) within 212 regions split into lesional or non-lesional cortex, and grouped into seven cortical networks. Integrity was compared between people with MS and controls, and across cognitive groups: cognitively-impaired (CI; ≥ two domains at Z ≤ - 2 below controls), mildly CI (≥ two at - 2 < Z ≤ - 1.5), or cognitively-preserved (CP). RESULTS: Cortical lesions were observed in 87.5% of people with MS, mainly in ventral attention network, followed by limbic and default mode networks. Compared to controls, in non-lesional cortex, MD was increased in people with MS, but mean FA did not differ. Within the same individual, MD and FA were increased in lesional compared to non-lesional cortex. CI-MS exhibited higher MD than CP-MS in non-lesional cortex of default mode, frontoparietal and sensorimotor networks, of which the default mode network could best explain cognitive performance. CONCLUSION: Diffusion differences in lesional cortex were more severe than in non-lesional cortex. However, while most people with MS had cortical lesions, diffusion differences in CI-MS were more prominent in non-lesional cortex than lesional cortex, especially within default mode, frontoparietal and sensorimotor networks.
Subject(s)
Cerebral Cortex , Cognitive Dysfunction , Multiple Sclerosis , Nerve Net , Humans , Male , Female , Cross-Sectional Studies , Adult , Multiple Sclerosis/diagnostic imaging , Multiple Sclerosis/pathology , Multiple Sclerosis/physiopathology , Multiple Sclerosis/complications , Middle Aged , Cerebral Cortex/diagnostic imaging , Cerebral Cortex/pathology , Cerebral Cortex/physiopathology , Cognitive Dysfunction/etiology , Cognitive Dysfunction/diagnostic imaging , Cognitive Dysfunction/physiopathology , Cognitive Dysfunction/pathology , Nerve Net/diagnostic imaging , Nerve Net/physiopathology , Nerve Net/pathology , Magnetic Resonance Imaging , Neuropsychological Tests , Diffusion Magnetic Resonance ImagingABSTRACT
BACKGROUND: Digital monitoring of people with multiple sclerosis (PwMS) using smartphone-based monitoring tools is a promising method to assess disease activity and progression. OBJECTIVE: To study cross-sectional and longitudinal associations between active and passive digital monitoring parameters and MRI volume measures in PwMS. METHODS: In this prospective study, 92 PwMS were included. Clinical tests [Expanded Disability Status Scale (EDSS), Timed 25 Foot Walk test (T25FW), 9-Hole Peg Test (NHPT), and Symbol Digit Modalities Test (SDMT)] and structural MRI scans were performed at baseline (M0) and 12-month follow-up (M12). Active monitoring included the smartphone-based Symbol Digit Modalities Test (sSDMT) and 2 Minute Walk Test (s2MWT), while passive monitoring was based on smartphone keystroke dynamics (KD). Linear regression analyses were used to determine cross-sectional and longitudinal relations between digital and clinical outcomes and brain volumes, with age, disease duration and sex as covariates. RESULTS: In PwMS, both sSDMT and SDMT were associated with thalamic volumes and lesion volumes. KD were related to brain, ventricular, thalamic and lesion volumes. No relations were found between s2MWT and MRI volumes. NHPT scores were associated with lesion volumes only, while EDSS and T25FW were not related to MRI. No longitudinal associations were found for any of the outcome measures between M0 and M12. CONCLUSION: Our results show clear cross-sectional correlations between digital biomarkers and brain volumes in PwMS, which were not all present for conventional clinical outcomes, supporting the potential added value of digital monitoring tools.
ABSTRACT
BACKGROUND AND OBJECTIVES: Disability and cognitive impairment are known to be related to brain atrophy in multiple sclerosis (MS), but 3D-T1 imaging required for brain volumetrics is often unavailable in clinical protocols, unlike 3D-FLAIR. Here our aim was to investigate whether brain volumes derived from 3D-FLAIR images result in similar associations with disability and cognition in MS as do those derived from 3D-T1 images. METHODS: 3T-MRI scans of 329 MS patients and 76 healthy controls were included in this cross-sectional study. Brain volumes were derived using FreeSurfer on 3D-T1 and compared with brain volumes derived with SynthSeg and SAMSEG on 3D-FLAIR. Relative agreement was evaluated by calculating the intraclass correlation coefficient (ICC) of the 3D-T1 and 3D-FLAIR volumes. Consistency of relations with disability and average cognition was assessed using linear regression, while correcting for age and sex. The findings were corroborated in an independent validation cohort of 125 MS patients. RESULTS: The ICC between volume measured with FreeSurfer and those measured on 3D-FLAIR for brain, ventricle, cortex, total deep gray matter and thalamus was above 0.74 for SAMSEG and above 0.91 for SynthSeg. Worse disability and lower average cognition were similarly associated with brain (adj. R2 = 0.24-0.27, p < 0.01; adj. R2 = 0.26-0.29, p < 0.001) ventricle (adj. R2 = 0.27-0.28, p < 0.001; adj. R2 = 0.19-0.20, p < 0.001) and deep gray matter volumes (adj. R2 = 0.24-0.28, p < 0.001; adj. R2 = 0.27-0.28, p < 0.001) determined with all methods, except for cortical volumes derived from 3D-FLAIR. DISCUSSION: In this cross-sectional study, brain volumes derived from 3D-FLAIR and 3D-T1 show similar relationships to disability and cognitive dysfunction in MS, highlighting the potential of these techniques in clinical datasets.
Subject(s)
Multiple Sclerosis , Humans , Multiple Sclerosis/complications , Multiple Sclerosis/diagnostic imaging , Gray Matter/pathology , Cross-Sectional Studies , Feasibility Studies , Brain/pathology , Magnetic Resonance Imaging/methods , Cognition , Atrophy/pathologyABSTRACT
Purpose: Pathological conditions associated with the optic nerve (ON) can cause structural changes in the nerve. Quantifying these changes could provide further understanding of disease mechanisms. We aim to develop a framework that automatically segments the ON separately from its surrounding cerebrospinal fluid (CSF) on magnetic resonance imaging (MRI) and quantifies the diameter and cross-sectional area along the entire length of the nerve. Approach: Multicenter data were obtained from retinoblastoma referral centers, providing a heterogeneous dataset of 40 high-resolution 3D T2-weighted MRI scans with manual ground truth delineations of both ONs. A 3D U-Net was used for ON segmentation, and performance was assessed in a tenfold cross-validation (n=32) and on a separate test-set (n=8) by measuring spatial, volumetric, and distance agreement with manual ground truths. Segmentations were used to quantify diameter and cross-sectional area along the length of the ON, using centerline extraction of tubular 3D surface models. Absolute agreement between automated and manual measurements was assessed by the intraclass correlation coefficient (ICC). Results: The segmentation network achieved high performance, with a mean Dice similarity coefficient score of 0.84, median Hausdorff distance of 0.64 mm, and ICC of 0.95 on the test-set. The quantification method obtained acceptable correspondence to manual reference measurements with mean ICC values of 0.76 for the diameter and 0.71 for the cross-sectional area. Compared with other methods, our method precisely identifies the ON from surrounding CSF and accurately estimates its diameter along the nerve's centerline. Conclusions: Our automated framework provides an objective method for ON assessment in vivo.
ABSTRACT
BACKGROUND AND OBJECTIVES: Although MRI-based markers of neuroinflammation have proven crucial for the diagnosis of multiple sclerosis (MS), predicting clinical progression with inflammation remains difficult. Neurodegenerative markers such as brain volume loss show stronger clinical (predictive) correlations, but also harbor age-related variation that must be disentangled from disease duration. In this study we investigated how clinical disability is related to volumetric MRI measures in a cohort of MS patients and healthy controls (HC) of the same age: Project Y. METHODS: This study included 234 MS patients born in 1966 and 112 HC born between 1965 and 1967 in the Netherlands. Disability was quantified using the expanded disability status scale (EDSS), nine hole peg test (9HPT), and timed 25 foot walking test (T25FWT). Volumes were quantified on 3T MRI as normalized whole brain (NBV) and regional gray matter (GM) volumes using the same scanner and MRI protocol: cortical (normalized cortical gray matter volume; NCGMV), deep (NDGMV), thalamic (NThalV), and cerebellar (NCbV) GM volumes. In addition, mean upper cervical cord area (MUCCA), white matter lesion volume (LV), and spinal cord lesions were assessed. These measures were compared between patients and HC, and related to disability measures using linear regression. RESULTS: Mean age of people with MS (PwMS) was 52.8 years (SD 0.9) and median disease duration 15.8 years (IQR 8.7-24.8). All global and regional brain measures were lower in MS patients compared to HC. Univariate regression models showed that NDGMV (ß = -0.20) and MUCCA (ß = -0.38) were most strongly related to the EDSS in all PwMS. After subtype stratification, MUCCA was most strongly related to the EDSS (ß = -0.60) and 9HPT (ß = -0.55) in secondary progressive PwMS. Multivariate regression models demonstrated that in all PwMS, the EDSS was best explained by lower MUCCA, longer disease durations and a progressive disease course (adjusted-R (Sastre-Garriga et al., 2017) = 0.26, p < 0.001). MUCCA was a consistent correlate in separate models of the EDSS for all PwMS, relapsing and progressive onset PwMS. The 9HPT (adjusted-R (Sastre-Garriga et al., 2017) = 0.20, p < 0.001) was best explained by lower MUCCA, higher LV and pack years, while lower limb disability (adjusted-R (Sastre-Garriga et al., 2017) = 0.11, p < 0.001) was best explained by lower MUCCA, progressive onset MS and female sex. DISCUSSION: Our results indicate that in a cohort unbiased by age differences, spinal cord and deep gray matter volumes best related to physical disability. Our results support the use of these measures in clinical practice and trials.
Subject(s)
Multiple Sclerosis, Chronic Progressive , Multiple Sclerosis , Humans , Female , Middle Aged , Multiple Sclerosis/diagnostic imaging , Multiple Sclerosis/pathology , Magnetic Resonance Imaging/methods , Gray Matter/pathology , Multiple Sclerosis, Chronic Progressive/pathology , Brain/diagnostic imaging , Brain/pathology , Disability Evaluation , Atrophy/pathologyABSTRACT
BACKGROUND : Accurate segmentation of head and neck squamous cell cancer (HNSCC) is important for radiotherapy treatment planning. Manual segmentation of these tumors is time-consuming and vulnerable to inconsistencies between experts, especially in the complex head and neck region. The aim of this study is to introduce and evaluate an automatic segmentation pipeline for HNSCC using a multi-view CNN (MV-CNN). METHODS: The dataset included 220 patients with primary HNSCC and availability of T1-weighted, STIR and optionally contrast-enhanced T1-weighted MR images together with a manual reference segmentation of the primary tumor by an expert. A T1-weighted standard space of the head and neck region was created to register all MRI sequences to. An MV-CNN was trained with these three MRI sequences and evaluated in terms of volumetric and spatial performance in a cross-validation by measuring intra-class correlation (ICC) and dice similarity score (DSC), respectively. RESULTS: The average manual segmented primary tumor volume was 11.8±6.70 cm3 with a median [IQR] of 13.9 [3.22-15.9] cm3. The tumor volume measured by MV-CNN was 22.8±21.1 cm3 with a median [IQR] of 16.0 [8.24-31.1] cm3. Compared to the manual segmentations, the MV-CNN scored an average ICC of 0.64±0.06 and a DSC of 0.49±0.19. Improved segmentation performance was observed with increasing primary tumor volume: the smallest tumor volume group (<3 cm3) scored a DSC of 0.26±0.16 and the largest group (>15 cm3) a DSC of 0.63±0.11 (p<0.001). The automated segmentation tended to overestimate compared to the manual reference, both around the actual primary tumor and in false positively classified healthy structures and pathologically enlarged lymph nodes. CONCLUSION: An automatic segmentation pipeline was evaluated for primary HNSCC on MRI. The MV-CNN produced reasonable segmentation results, especially on large tumors, but overestimation decreased overall performance. In further research, the focus should be on decreasing false positives and make it valuable in treatment planning.