Multimodal Magnetic Resonance Imaging Research Articles

DeepKOA: a deep-learning model for predicting progression in knee osteoarthritis using multimodal magnetic resonance images from the osteoarthritis initiative.

No investigations have thoroughly explored the feasibility of combining magnetic resonance (MR) images and deep-learning methods for predicting the progression of knee osteoarthritis (KOA). We thus aimed to develop a potential deep-learning model for predicting OA progression based on MR images for the clinical setting. A longitudinal case-control study was performed using data from the Foundation for the National Institutes of Health (FNIH), composed of progressive cases [182 osteoarthritis (OA) knees with both radiographic and pain progression for 24-48 months] and matched controls (182 OA knees not meeting the case definition). DeepKOA was developed through 3-dimensional (3D) DenseNet169 to predict KOA progression over 24-48 months based on sagittal intermediate-weighted turbo-spin echo sequences with fat-suppression (SAG-IW-TSE-FS), sagittal 3D dual-echo steady-state water excitation (SAG-3D-DESS-WE) and its axial and coronal multiplanar reformation, and their combined MR images with patient-level labels at baseline, 12, and 24 months to eventually determine the probability of progression. The classification performance of the DeepKOA was evaluated using 5-fold cross-validation. An X-ray-based model and traditional models that used clinical variables via multilayer perceptron were built. Combined models were also constructed, which integrated clinical variables with DeepKOA. The area under the curve (AUC) was used as the evaluation metric. The performance of SAG-IW-TSE-FS in predicting OA progression was similar or higher to that of other single and combined sequences. The DeepKOA based on SAG-IW-TSE-FS achieved an AUC of 0.664 (95% CI: 0.585-0.743) at baseline, 0.739 (95% CI: 0.703-0.775) at 12 months, and 0.775 (95% CI: 0.686-0.865) at 24 months. The X-ray-based model achieved an AUC ranging from 0.573 to 0.613 at 3 time points. However, adding clinical variables to DeepKOA did not improve performance (P>0.05). Initial visualizations from gradient-weighted class activation mapping (Grad-CAM) indicated that the frequency with which the patellofemoral joint was highlighted increased as time progressed, which contrasted the trend observed in the tibiofemoral joint. The meniscus, the infrapatellar fat pad, and muscles posterior to the knee were highlighted to varying degrees. This study initially demonstrated the feasibility of DeepKOA in the prediction of KOA progression and identified the potential responsible structures which may enlighten the future development of more clinically practical methods.

Association between enhanced carbonyl stress and decreased apparent axonal density in schizophrenia by multimodal white matter imaging

Carbonyl stress is a condition featuring increased rich reactive carbonyl compounds, which facilitate the formation of advanced glycation end products including pentosidine. We previously reported the relationship between enhanced carbonyl stress and disrupted white matter integrity in schizophrenia, although which microstructural component is disrupted remained unclear. In this study, 32 patients with schizophrenia (SCZ) and 45 age- and gender-matched healthy volunteers (HC) were recruited. We obtained blood samples for carbonyl stress markers (plasma pentosidine and serum pyridoxal) and multi-modal magnetic resonance imaging measures of white matter microstructures including apparent axonal density (intra-cellular volume fraction (ICVF)) and orientation (orientation dispersion index (ODI)), and inflammation (free water (FW)). In SCZ, the plasma pentosidine level was significantly increased. Group comparison revealed that mean white matter values were decreased for ICVF, and increased for FW. We found a significant negative correlation between the plasma pentosidine level and mean ICVF values in SCZ, and a significant negative correlation between the serum pyridoxal level and mean ODI value in HC, regardless of age. Our results suggest an association between enhanced carbonyl stress and axonal abnormality in SCZ.

Multimodality MRI Radiomics Based on Machine Learning for Identifying True Tumor Recurrence and Treatment-Related Effects in Patients with Postoperative Glioma.

Conventional magnetic resonance imaging (MRI) features have difficulty distinguishing glioma true tumor recurrence (TuR) from treatment-related effects (TrE). We aimed to develop a machine-learning model based on multimodality MRI radiomics to help improve the efficiency of identifying glioma TuR. A total of 131 patients were enrolled and randomly divided into the training set (n=91) and the test set (n=40). Radiomic features were extracted from the postoperative enhancement (PoE) region and edema (ED) region from four routine MRI sequences. After analyses of Spearman's rank correlation coefficient, and least absolute shrinkage and selection operator, the key radiomic features were selected to construct support vector machine (SVM) and k-nearest neighbor (KNN) models. Decision curve analysis (DCA) and receiver operating characteristic (ROC) curves were used to analyze the performance. The PoE model had a significantly higher area under curve (AUC) than the ED model (p<0.05). Among the models constructed with a single sequence, the model using PoE regional features from CE-T1WI was superior to other models, with an AUC of 0.905 for SVM and 0.899 for KNN. In multimodality models, the PoE model outperformed the ED model with an AUC of 0.931 for SVM and 0.896 for KNN. The multimodality model, which combined routine sequences and the whole regional features, showed a slightly better performance with an AUC of 0.965 for SVM and 0.955 for KNN. Decision curve analysis showed the good clinical utility of multimodal radiomics models. Multimodality radiomics can identify glioma TuR and TrE, potentially aiding clinical decision-making for individualized treatment. And edematous regions may provide useful information for recognizing recurrence. 2021.04.15, No:2020039.

Diagnostic efficiency of multi-modal MRI based deep learning with Sobel operator in differentiating benign and malignant breast mass lesions-a retrospective study.

To compare the diagnostic efficiencies of deep learning single-modal and multi-modal for the classification of benign and malignant breast mass lesions. We retrospectively collected data from 203 patients (207 lesions, 101 benign and 106 malignant) with breast tumors who underwent breast magnetic resonance imaging (MRI) before surgery or biopsy between January 2014 and October 2020. Mass segmentation was performed based on the three dimensions-region of interest (3D-ROI) minimum bounding cube at the edge of the lesion. We established single-modal models based on a convolutional neural network (CNN) including T2WI and non-fs T1WI, the dynamic contrast-enhanced (DCE-MRI) first phase was pre-contrast T1WI (d1), and Phases 2, 4, and 6 were post-contrast T1WI (d2, d4, d6); and Multi-modal fusion models with a Sobel operator (four_mods:T2WI, non-fs-T1WI, d1, d2). Training set (n=145), validation set (n=22), and test set (n=40). Five-fold cross validation was performed. Accuracy, sensitivity, specificity, negative predictive value, positive predictive value, and area under the ROC curve (AUC) were used as evaluation indicators. Delong's test compared the diagnostic performance of the multi-modal and single-modal models. All models showed good performance, and the AUC values were all greater than 0.750. Among the single-modal models, T2WI, non-fs-T1WI, d1, and d2 had specificities of 77.1%, 77.2%, 80.2%, and 78.2%, respectively. d2 had the highest accuracy of 78.5% and showed the best diagnostic performance with an AUC of 0.827. The multi-modal model with the Sobel operator performed better than single-modal models, with an AUC of 0.887, sensitivity of 79.8%, specificity of 86.1%, and positive prediction value of 85.6%. Delong's test showed that the diagnostic performance of the multi-modal fusion models was higher than that of the six single-modal models (T2WI, non-fs-T1WI, d1, d2, d4, d6); the difference was statistically significant (p = 0.043, 0.017, 0.006, 0.017, 0.020, 0.004, all were greater than 0.05). Multi-modal fusion deep learning models with a Sobel operator had excellent diagnostic value in the classification of breast masses, and further increase the efficiency of diagnosis.

Alteration and clinical potential in gut microbiota in patients with cerebral small vessel disease.

Cerebral small vessel disease (CSVD) is a cluster of microvascular disorders with unclear pathological mechanisms. The microbiota-gut-brain axis is an essential regulatory mechanism between gut microbes and their host. Therefore, the compositional and functional gut microbiota alterations lead to cerebrovascular disease pathogenesis. The current study aims to determine the alteration and clinical value of the gut microbiota in CSVD patients. Sixty-four CSVD patients and 18 matched healthy controls (HCs) were included in our study. All the participants underwent neuropsychological tests, and the multi-modal magnetic resonance imaging depicted the changes in brain structure and function. Plasma samples were collected, and the fecal samples were analyzed with 16S rRNA gene sequencing. Based on the alpha diversity analysis, the CSVD group had significantly decreased Shannon and enhanced Simpson compared to the HC group. At the genus level, there was a significant increase in the relative abundances of Parasutterella, Anaeroglobus, Megasphaera, Akkermansia, Collinsella, and Veillonella in the CSVD group. Moreover, these genera with significant differences in CSVD patients revealed significant correlations with cognitive assessments, plasma levels of the blood-brain barrier-/inflammation-related indexes, and structural/functional magnetic resonance imaging changes. Functional prediction demonstrated that lipoic acid metabolism was significantly higher in CSVD patients than HCs. Additionally, a composite biomarker depending on six gut microbiota at the genus level displayed an area under the curve of 0.834 to distinguish CSVD patients from HCs using the least absolute shrinkage and selection operator (LASSO) algorithm. The evident changes in gut microbiota composition in CSVD patients were correlated with clinical features and pathological changes of CSVD. Combining these gut microbiota using the LASSO algorithm helped identify CSVD accurately.

The impact of HCN4 channels on CNS brain networks as a new target in pain development.

While it is well established that the isoform 2 of the hyperpolarization-activated cyclic nucleotide-gated cation channel (HCN2) plays an important role in the development and maintenance of pain, the role of the closely related HCN4 isoform in the processing of nociceptive signals is not known. HCN4 channels are highly expressed in the thalamus, a region important for stimulus transmission and information processing. We used a brain-specific HCN4-knockout mouse line (HCN4-KO) to explore the role of HCN4 channels in acute nociceptive processing using several behavioral tests as well as a multimodal magnetic resonance imaging (MRI) approach. Functional MRI (fMRI) brain responses were measured during acute peripheral thermal stimulation complemented by resting state (RS) before and after stimulation. The data were analyzed by conventional and graph-theoretical approaches. Finally, high-resolution anatomical brain data were acquired. HCN4-KO animals showed a central thermal, but not a mechanical hypersensitivity in behavioral experiments. The open field analysis showed no significant differences in motor readouts between HCN4-KO and controls but uncovered increased anxiety in the HCN4-KO mice. Thermal stimulus-driven fMRI (s-fMRI) data revealed increased response volumes and response amplitudes for HCN4-KO, most pronounced at lower stimulation temperatures in the subcortical input, the amygdala as well as in limbic/hippocampal regions, and in the cerebellum. These findings could be cross-validated by graph-theoretical analyses. Assessment of short-term RS before and after thermal stimulation revealed that stimulation-related modulations of the functional connectivity only occurred in control animals. This was consistent with the finding that the hippocampus was found to be smaller in HCN4-KO. In summary, the deletion of HCN4 channels impacts on processing of acute nociception, which is remarkably manifested as a thermal hypersensitive phenotype. This was mediated by the key regions hypothalamus, somatosensory cortex, cerebellum and the amygdala. As consequence, HCN4-KO mice were more anxious, and their brain-wide RS functional connectivity could not be modulated by thermal nociceptive stimulation.

Contrast-Enhanced Liver Magnetic Resonance Image Synthesis Using Gradient Regularized Multi-Modal Multi-Discrimination Sparse Attention Fusion GAN.

To provide abdominal contrast-enhanced MR image synthesis, we developed an gradient regularized multi-modal multi-discrimination sparse attention fusion generative adversarial network (GRMM-GAN) to avoid repeated contrast injections to patients and facilitate adaptive monitoring. With IRB approval, 165 abdominal MR studies from 61 liver cancer patients were retrospectively solicited from our institutional database. Each study included T2, T1 pre-contrast (T1pre), and T1 contrast-enhanced (T1ce) images. The GRMM-GAN synthesis pipeline consists of a sparse attention fusion network, an image gradient regularizer (GR), and a generative adversarial network with multi-discrimination. The studies were randomly divided into 115 for training, 20 for validation, and 30 for testing. The two pre-contrast MR modalities, T2 and T1pre images, were adopted as inputs in the training phase. The T1ce image at the portal venous phase was used as an output. The synthesized T1ce images were compared with the ground truth T1ce images. The evaluation metrics include peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and mean squared error (MSE). A Turing test and experts' contours evaluated the image synthesis quality. The proposed GRMM-GAN model achieved a PSNR of 28.56, an SSIM of 0.869, and an MSE of 83.27. The proposed model showed statistically significant improvements in all metrics tested with p-values < 0.05 over the state-of-the-art model comparisons. The average Turing test score was 52.33%, which is close to random guessing, supporting the model's effectiveness for clinical application. In the tumor-specific region analysis, the average tumor contrast-to-noise ratio (CNR) of the synthesized MR images was not statistically significant from the real MR images. The average DICE from real vs. synthetic images was 0.90 compared to the inter-operator DICE of 0.91. We demonstrated the function of a novel multi-modal MR image synthesis neural network GRMM-GAN for T1ce MR synthesis based on pre-contrast T1 and T2 MR images. GRMM-GAN shows promise for avoiding repeated contrast injections during radiation therapy treatment.

Aberrant dentato-rubro-thalamic pathway in action tremor but not rest tremor: A multi-modality magnetic resonance imaging study.

The purpose of this study was to clarify the dentato-rubro-thalamic (DRT) pathway in action tremor in comparison to normal controls (NC) and disease controls (i.e., rest tremor) by using multi-modality magnetic resonance imaging (MRI). This study included 40 essential tremor (ET) patients, 57 Parkinson's disease (PD) patients (29 with rest tremor, 28 without rest tremor), and 41 NC. We used multi-modality MRI to comprehensively assess major nuclei and fiber tracts of the DRT pathway, which included decussating DRT tract (d-DRTT) and non-decussating DRT tract (nd-DRTT), and compared the differences in DRT pathway components between action and rest tremor. Bilateral dentate nucleus (DN) in the ET group had excessive iron deposition compared with the NC group. Compared with the NC group, significantly decreased mean diffusivity and radial diffusivity were observed in the left nd-DRTT in the ET group, which were negatively correlated with tremor severity. No significant difference in each component of the DRT pathway was observed between the PD subgroup or the PD and NC. Aberrant changes in the DRT pathway may be specific to action tremor and were indicating that action tremor may be related to pathological overactivation of the DRT pathway.

Connectivity reveals homology between the visual systems of the human and macaque brains.

The visual systems of humans and nonhuman primates share many similarities in both anatomical and functional organization. Understanding the homology and differences between the two systems can provide important insights into the neural basis of visual perception and cognition. This research aims to investigate the homology between human and macaque visual systems based on connectivity, using diffusion tensor imaging and resting-state functional magnetic resonance imaging to construct structural and functional connectivity fingerprints of the visual systems in humans and macaques, and quantitatively analyze the connectivity patterns. By integrating multimodal magnetic resonance imaging, this research explored the homology and differences between the two systems. The results showed that 9 brain regions in the macaque visual system formed highly homologous mapping relationships with 11 brain regions in the human visual system, and the related brain regions between the two species showed highly structure homologous, with their functional organization being essentially conserved across species. Finally, this research generated a homology information map of the visual system for humans and macaques, providing a new perspective for subsequent cross-species analysis.

Joint learning-based feature reconstruction and enhanced network for incomplete multi-modal brain tumor segmentation

Multimodal Magnetic Resonance Imaging (MRI) can provide valuable complementary information and substantially enhance the performance of brain tumor segmentation. However, it is common for certain modalities to be absent or missing during clinical diagnosis, which can significantly impair segmentation techniques that rely on complete modalities. Current advanced methods attempt to address this challenge by developing shared feature representations via modal fusion to handle different missing modality situations. Considering the importance of missing modality information in multimodal segmentation, this paper utilize a feature reconstruction method to recover the missing information, and proposes a joint learning-based feature reconstruction and enhancement method for incomplete modality brain tumor segmentation. The method leverages an information learning mechanism to transfer information from the complete modality to a single modality, enabling it to obtain complete brain tumor information, even without the support of other modalities. Additionally, the method incorporates a module for reconstructing missing modality features, which recovers fused features of the absent modality through utilizing the abundant potential information obtained from the available modalities. Furthermore, the feature enhancement mechanism improves shared feature representation by utilizing the information obtained from the missing modalities that have been reconstructed. These processes enable the method to obtain more comprehensive information regarding brain tumors in various missing modality circumstances, thereby enhancing the model’s robustness. The performance of the proposed model was evaluated on BraTS datasets and compared with other deep learning algorithms using Dice similarity scores. On the BraTS2018 dataset, the proposed algorithm achieved a Dice similarity score of 86.28%, 77.02%, and 59.64% for whole tumors, tumor cores, and enhanced tumors, respectively. These results demonstrate the superiority of our framework over state-of-the-art methods in missing modalities situations.

Automated segmentation of the human supraclavicular fat depot via deep neural network in water-fat separated magnetic resonance images.

Human brown adipose tissue (BAT), mostly located in the cervical/supraclavicular region, is a promising target in obesity treatment. Magnetic resonance imaging (MRI) allows for mapping the fat content quantitatively. However, due to the complex heterogeneous distribution of BAT, it has been difficult to establish a standardized segmentation routine based on magnetic resonance (MR) images. Here, we suggest using a multi-modal deep neural network to detect the supraclavicular fat pocket. A total of 50 healthy subjects [median age/body mass index (BMI) =36 years/24.3 kg/m2] underwent MRI scans of the neck region on a 3 T Ingenia scanner (Philips Healthcare, Best, Netherlands). Manual segmentations following fixed rules for anatomical borders were used as ground truth labels. A deep learning-based method (termed as BAT-Net) was proposed for the segmentation of BAT on MRI scans. It jointly leveraged two-dimensional (2D) and three-dimensional (3D) convolutional neural network (CNN) architectures to efficiently encode the multi-modal and 3D context information from multi-modal MRI scans of the supraclavicular region. We compared the performance of BAT-Net to that of 2D U-Net and 3D U-Net. For 2D U-Net, we analyzed the performance difference of implementing 2D U-Net in three different planes, denoted as 2D U-Net (axial), 2D U-Net (coronal), and 2D U-Net (sagittal). The proposed model achieved an average dice similarity coefficient (DSC) of 0.878 with a standard deviation of 0.020. The volume segmented by the network was smaller compared to the ground truth labels by 9.20 mL on average with a mean absolute increase in proton density fat fraction (PDFF) inside the segmented regions of 1.19 percentage points. The BAT-Net outperformed all implemented 2D U-Nets and the 3D U-Nets with average DSC enhancement ranging from 0.016 to 0.023. The current work integrates a deep neural network-based segmentation into the automated segmentation of supraclavicular fat depot for quantitative evaluation of BAT. Experiments show that the presented multi-modal method benefits from leveraging both 2D and 3D CNN architecture and outperforms the independent use of 2D or 3D networks. Deep learning-based segmentation methods show potential towards a fully automated segmentation of the supraclavicular fat depot.

Multimodal quantitative magnetic resonance imaging of the thalamus in tinnitus patients with different outcomes after sound therapy.

This study systematically investigated structural and functional alterations in the thalamus and its subregions using multimodal magnetic resonance imaging (MRI) and examined its clinical relevance in tinnitus patients with different outcomes after sound therapy (narrowband noise). In total, 60 patients with persistent tinnitus and 57 healthy controls (HCs) were recruited. Based on treatment efficacy, 28 patients were categorized into the effective group and 32 into the ineffective group. Five MRI measurements of the thalamus and its seven subregions, including gray matter volume, fractional anisotropy, fractional amplitude of low-frequency fluctuation, and functional connectivity (FC), were obtained for each participant and compared between the groups. Patients in both the groups exhibited widespread functional and diffusion abnormalities in the whole thalamus and several subregions, with more obvious changes observed in the effective group. All tinnitus patients had abnormal FC compared with the HCs; FC differences between the two patient groups were only observed in the striatal network, auditory-related cortex, and the core area of the limbic system. We combined the multimodal quantitative thalamic alterations and used it as an imaging indicator to evaluate prognosis before sound therapy and achieved a sensitivity of 71.9% and a specificity of 85.7%. Similar patterns of thalamic alterations were identified in tinnitus patients with different outcomes, with more obvious changes observed in the effective group. Our findings support the tinnitus generation hypothesis of frontostriatal gating system dysfunction. A combination of multimodal quantitative thalamic properties may be used as indicators to predict tinnitus prognosis before sound therapy.

How does the brain age in individuals with multiple sclerosis? A systematic review.

Multiple Sclerosis (MS) is a complex neurological disorder that involves demyelination, lesions and atrophy in both white and gray matter. Such changes in the central nervous system are diagnostic in MS and has a strong relationship with both physical and cognitive symptoms. As a result, magnetic resonance imaging (MRI) scans as a metric of brain atrophy have emerged as an important outcome measure in MS studies. Recently, research has begun to focus on the contribution of aging to the structural changes in the brain associated with MS; prompting questions about whether there is an amplifying effect of aging superimposed on MS-related brain atrophy. To examine current evidence of how the brain ages in individuals with MS, a systematic review of the literature was performed. Specific questions were focused on how aging affects gray and white matter structure, whether patterns of brain atrophy differ in younger and older cohorts and if there are structural differences in the brain as a function of sex in aging people with MS. This review considered studies that used MRI to examine the effects of aging in adults with MS. Twenty-one studies met eligibility criteria. Findings across these studies revealed that gray matter atrophy was more pronounced in older adults with MS, particularly in subcortical regions such as the thalamus; that the rates of atrophy were similar but varied by region for younger and older cohorts; and that males may experience more brain atrophy than females. Further studies that use multimodal MRI acquisition methods are needed to capture changes in both males and females over time, particularly in middle to older adulthood.

M3AE: Multimodal Representation Learning for Brain Tumor Segmentation with Missing Modalities

Multimodal magnetic resonance imaging (MRI) provides complementary information for sub-region analysis of brain tumors. Plenty of methods have been proposed for automatic brain tumor segmentation using four common MRI modalities and achieved remarkable performance. In practice, however, it is common to have one or more modalities missing due to image corruption, artifacts, acquisition protocols, allergy to contrast agents, or simply cost. In this work, we propose a novel two-stage framework for brain tumor segmentation with missing modalities. In the first stage, a multimodal masked autoencoder (M3AE) is proposed, where both random modalities (i.e., modality dropout) and random patches of the remaining modalities are masked for a reconstruction task, for self-supervised learning of robust multimodal representations against missing modalities. To this end, we name our framework M3AE. Meanwhile, we employ model inversion to optimize a representative full-modal image at marginal extra cost, which will be used to substitute for the missing modalities and boost performance during inference. Then in the second stage, a memory-efficient self distillation is proposed to distill knowledge between heterogenous missing-modal situations while fine-tuning the model for supervised segmentation. Our M3AE belongs to the ‘catch-all’ genre where a single model can be applied to all possible subsets of modalities, thus is economic for both training and deployment. Extensive experiments on BraTS 2018 and 2020 datasets demonstrate its superior performance to existing state-of-the-art methods with missing modalities, as well as the efficacy of its components. Our code is available at: https://github.com/ccarliu/m3ae.

A unified hybrid transformer for joint MRI sequences super-resolution and missing data imputation

Objective. High-resolution multi-modal magnetic resonance imaging (MRI) is crucial in clinical practice for accurate diagnosis and treatment. However, challenges such as budget constraints, potential contrast agent deposition, and image corruption often limit the acquisition of multiple sequences from a single patient. Therefore, the development of novel methods to reconstruct under-sampled images and synthesize missing sequences is crucial for clinical and research applications. Approach. In this paper, we propose a unified hybrid framework called SIFormer, which utilizes any available low-resolution MRI contrast configurations to complete super-resolution (SR) of poor-quality MR images and impute missing sequences simultaneously in one forward process. SIFormer consists of a hybrid generator and a convolution-based discriminator. The generator incorporates two key blocks. First, the dual branch attention block combines the long-range dependency building capability of the transformer with the high-frequency local information capture capability of the convolutional neural network in a channel-wise split manner. Second, we introduce a learnable gating adaptation multi-layer perception in the feed-forward block to optimize information transmission efficiently. Main results. Comparative evaluations against six state-of-the-art methods demonstrate that SIFormer achieves enhanced quantitative performance and produces more visually pleasing results for image SR and synthesis tasks across multiple datasets. Significance. Extensive experiments conducted on multi-center multi-contrast MRI datasets, including both healthy individuals and brain tumor patients, highlight the potential of our proposed method to serve as a valuable supplement to MRI sequence acquisition in clinical and research settings.

Prediction of 1p/19q codeletion status in diffuse lower-grade glioma using multimodal MRI radiomics

To develop a noninvasive method for prediction of 1p/19q codeletion in diffuse lower-grade glioma (DLGG) based on multimodal magnetic resonance imaging (MRI) radiomics. We collected MRI data from 104 patients with pathologically confirmed DLGG between October, 2015 and September, 2022. A total of 535 radiomics features were extracted from T2WI, T1WI, FLAIR, CE-T1WI and DWI, including 70 morphological features, 90 first order features, and 375 texture features. We constructed logistic regression (LR), logistic regression least absolute shrinkage and selection operator (LRlasso), support vector machine (SVM) and Linear Discriminant Analysis (LDA) radiomics models and compared their predictive performance after 10-fold cross validation. The MRI images were reviewed by two radiologists independently for predicting the 1p/19q status. Receiver operating characteristic curves were used to evaluate classification performance of the radiomics models and the radiologists. The 4 radiomics models (LR, LRlasso, SVM and LDA) achieved similar area under the curve (AUC) in the validation dataset (0.833, 0.819, 0.824 and 0.819, respectively; P>0.1), and their predictive performance was all superior to that of resident physicians of radiology (AUC=0.645, P=0.011, 0.022, 0.016, 0.030, respectively) and similar to that of attending physicians of radiology (AUC=0.838, P>0.05). Multiparametric MRI radiomics models show good performance for noninvasive prediction of 1p/19q codeletion status in patients with in diffuse lower-grade glioma.

477-P: Thalamic Mitochondrial and Neuronal Function Is Preserved in Painful Diabetic Peripheral Neuropathy—A Multimodal Magnetic Resonance Imaging Study

Aims: Previous neuroimaging studies have demonstrated cerebral structural, functional, and vascular alterations in Painful-Diabetic Peripheral Neuropathy (DPN). We present a multi-modal Magnetic Resonance Imaging (MRI) study to determine the structure, and neuronal and mitochondrial function of the thalamus in Painful-DPN. Methods: A total of 49 participants were enrolled, 38 with type 2 diabetes (9 without DPN [No-DPN]; 11 with Painless-DPN; and 18 with Painful-DPN) and 11 healthy volunteers (HV). Participants underwent detailed clinical and neurophysiological assessments, and multi-modal MR imaging of the thalami bilaterally including: Proton-Magnetic Spectroscopy (1H-MRS); 31-Phosphorus MRS (31P-MRS); and brain morphometric analysis. We calculated the N-acetyl aspartate to choline (NAA:Cho), a measure of neuronal function, and inorganic phosphate (Pi) to ATP (Pi:ATP), a measure of mitochondrial function. Results: There were no significant differences in thalamic volume between groups (ANOVA, p=0.927). There was a significant group effect in both metabolite ratios: NAA:Cho (ANOVA, p=0.013); Pi:ATP (p=0.021). The NAA:Cho was numerically the lowest in Painless-DPN, reaching significance compared with HV (p=0.001) and Painful-DPN (p=0.019). Additionally, the Pi:ATP was significantly higher in Painless-DPN compared with HV (p=0.013) and Painful-DPN (p=0.008). There was also a significant correlation between the NAA:Cho and Pi:PCr (Pearson’s r -0.336, p=0.034). Conclusion: This is the first study to determine perform cerebral 1H- and 31P-MRS in human DPN. We have demonstrated altered 1H- and 31P-MRS ratios in Painless- but not Painful-DPN, suggesting that mitochondrial dysfunction may underlie thalamic neuronal dysfunction in Painless-DPN. Conversely, in Painful-DPN, preservation of thalamic mitochondrial and neuronal function may be involved in the maintenance of neuropathic pain signaling. Disclosure G.P.Sloan: None. K.Teh: None. D.Selvarajah: None. S.Tesfaye: Advisory Panel; Bayer Inc., Wörwag Pharma GmbH &amp; Co. KG, Angelini, Speaker's Bureau; Viatris Inc., Nevro Corp., Procter &amp; Gamble, Novo Nordisk, Berlin-Chemie AG.

A neuromarker for deficit syndrome in schizophrenia from a combination of structural and functional magnetic resonance imaging.

Deficit schizophrenia (DS), defined by primary and enduring negative symptoms, has been proposed as a promising homogeneous subtype of schizophrenia. It has been demonstrated that unimodal neuroimaging characteristics of DS were different from non-deficit schizophrenia (NDS), however, whether multimodal-based neuroimaging features could identify deficit syndrome remains to be determined. Functional and structural multimodal magnetic resonance imaging of DS, NDS and healthy controls were scanned. Voxel-based features of gray matter volume, fractional amplitude of low-frequency fluctuations, and regional homogeneity were extracted. The support vector machine classification models were constructed using these features separately and jointly. The most discriminative features were defined as the first 10% of features with the greatest weights. Moreover, relevance vector regression was applied to explore the predictive values of these top-weighted features in predicting negative symptoms. The multimodal classifier achieved a higher accuracy (75.48%) compared with the single modal model in distinguishing DS from NDS. The most predictive brain regions were mainly located in the default mode and visual networks, exhibiting differences between functional and structural features. Further, the identified discriminative features significantly predicted scores of diminished expressivity factor in DS but not NDS. The present study demonstrated that local properties of brain regions extracted from multimodal imaging data could distinguish DS from NDS with a machine learning-based approach and confirmed the relationship between distinctive features and the negative symptoms subdomain. These findings may improve the identification of potential neuroimaging signatures and improve the clinical assessment of the deficit syndrome.

Brain structure-function coupling associated with cognitive impairment in cerebral small vessel disease.

Cerebral small vessel disease (CSVD) is a common chronic and progressive disease that can lead to mental and cognitive impairment. Damage to brain structure and function may play an important role in the neuropsychiatric disorders of patients with CSVD. Increasing evidence suggests that functional changes are accompanied by structural changes in corresponding brain regions. Thus, normal structure-function coupling is essential for optimal brain performance, and disrupted structure-function coupling can be found in many neurological and psychiatric disorders. To date, most studies on patients with CSVD have focused on separate structures or functions, including reductions in white matter volume and blood flow, which lead to cognitive dysfunction. However, there are few studies on brain structure-function coupling in patients with CSVD. In recent years, with the rapid development of multilevel (voxel-wise, neurovascular, regional level, and network level) brain structure-functional coupling analysis methods based on multimodal magnetic resonance imaging (MRI), new evidence has been provided to reveal the correlation between brain function and structural abnormalities and cognitive impairment. Therefore, studying brain structure-function coupling has a potential significance in the exploration and elucidation of the neurobiological mechanism of cognitive impairment in patients with CSVD. This article mainly describes the currently popular brain structure-function coupling analysis technology based on multimodal MRI and the important research progress of these coupling technologies on CSVD and cognitive impairment to provide a perspective for the study of the pathogenesis and early diagnosis of CSVD.

A single stage knowledge distillation network for brain tumor segmentation on limited MR image modalities

Background and objectivePrecisely segmenting brain tumors using multimodal Magnetic Resonance Imaging (MRI) is an essential task for early diagnosis, disease monitoring, and surgical planning. Unfortunately, the complete four image modalities utilized in the well-known BraTS benchmark dataset: T1, T2, Fluid-Attenuated Inversion Recovery (FLAIR), and T1 Contrast-Enhanced (T1CE) are not regularly acquired in clinical practice due to the high cost and long acquisition time. Rather, it is common to utilize limited image modalities for brain tumor segmentation. MethodsIn this paper, we propose a single stage learning of knowledge distillation algorithm that derives information from the missing modalities for better segmentation of brain tumors. Unlike the previous works that adopted a two-stage framework to distill the knowledge from a pre-trained network into a student network, where the latter network is trained on limited image modality, we train both models simultaneously using a single-stage knowledge distillation algorithm. We transfer the information by reducing the redundancy from a teacher network trained on full image modalities to the student network using Barlow Twins loss on a latent-space level. To distill the knowledge on the pixel level, we further employ a deep supervision idea that trains the backbone networks of both teacher and student paths using Cross-Entropy loss. ResultsWe demonstrate that the proposed single-stage knowledge distillation approach enables improving the performance of the student network in each tumor category with overall dice scores of 91.11% for Tumor Core, 89.70% for Enhancing Tumor, and 92.20% for Whole Tumor in the case of only using the FLAIR and T1CE images, outperforming the state-of-the-art segmentation methods. ConclusionsThe outcomes of this work prove the feasibility of exploiting the knowledge distillation in segmenting brain tumors using limited image modalities and hence make it closer to clinical practices.