Brain Magnetic Resonance Imaging Data Research Articles

Magnetic resonance imaging based neurosurgical planning on hololens 2: A feasibility study in a paediatric hospital.

The goal of this work is to show how to implement a mixed reality application (app) for neurosurgery planning based on neuroimaging data, highlighting the strengths and weaknesses of its design. Our workflow explains how to handle neuroimaging data, including how to load morphological, functional and diffusion tensor imaging data into a mixed reality environment, thus creating a first guide of this kind. Brain magnetic resonance imaging data from a paediatric patient were acquired using a 3 T Siemens Magnetom Skyra scanner. Initially, this raw data underwent specific software pre-processing and were subsequently transformed to ensure seamless integration with the mixed reality app. After that, we created three-dimensional models of brain structures and the mixed reality environment using Unity™ engine together with Microsoft® HoloLens 2™ device. To get an evaluation of the app we submitted a questionnaire to four neurosurgeons. To collect data concerning the performance of a user session we used Unity Performance Profiler. The use of the interactive features, such as rotating, scaling and moving models and browsing through menus, provided by the app had high scores in the questionnaire, and their use can still be improved as suggested by the performance data collected. The questionnaire's average scores were high, so the overall experiences of using our mixed reality app were positive. We have successfully created a valuable and easy-to-use neuroimaging data mixed reality app, laying the foundation for more future clinical uses, as more models and data derived from various biomedical images can be imported.

Deep Learning for Brain MRI Confirms Patterned Pathological Progression in Alzheimer's Disease.

Deep learning (DL) on brain magnetic resonance imaging (MRI) data has shown excellent performance in differentiating individuals with Alzheimer's disease (AD). However, the value of DL in detecting progressive structural MRI (sMRI) abnormalities linked to AD pathology has yet to be established. In this study, an interpretable DL algorithm named the Ensemble of 3-dimensional convolutional neural network (Ensemble 3DCNN) with enhanced parsing techniques is proposed to investigate the longitudinal trajectories of whole-brain sMRI changes denoting AD onset and progression. A set of 2369 T1-weighted images from the multi-centre Alzheimer's Disease Neuroimaging Initiative and Open Access Series of Imaging Studies cohorts are applied to model derivation, validation, testing, and pattern analysis. An Ensemble-3DCNN-based P-score is generated, based on which multiple brain regions, including amygdala, insular, parahippocampal, and temporal gyrus, exhibit early and connected progressive neurodegeneration. Complex individual variability in the sMRI is also observed. This study combining non-invasive sMRI and interpretable DL in detecting patterned sMRI changes confirmed AD pathological progression, shedding new light on predicting AD progression using whole-brainsMRI.

Semi-automatic segmentation of the fetal brain from magnetic resonance imaging.

Volumetric measurements of fetal brain maturation in the third trimester of pregnancy are key predictors of developmental outcomes. Improved understanding of fetal brain development trajectories may aid in identifying and clinically managing at-risk fetuses. Currently, fetal brain structures in magnetic resonance images (MRI) are often manually segmented, which requires both time and expertise. To facilitate the targeting and measurement of brain structures in the fetus, we compared the results of five segmentation methods applied to fetal brain MRI data to gold-standard manual tracings. Adult women with singleton pregnancies (n = 21), of whom five were scanned twice, approximately 3 weeks apart, were recruited [26 total datasets, median gestational age (GA) = 34.8, IQR = 30.9-36.6]. T2-weighted single-shot fast spin echo images of the fetal brain were acquired on 1.5T and 3T MRI scanners. Images were first combined into a single 3D anatomical volume. Next, a trained tracer manually segmented the thalamus, cerebellum, and total cerebral volumes. The manual segmentations were compared with five automatic methods of segmentation available within Advanced Normalization Tools (ANTs) and FMRIB's Linear Image Registration Tool (FLIRT) toolboxes. The manual and automatic labels were compared using Dice similarity coefficients (DSCs). The DSC values were compared using Friedman's test for repeated measures. Comparing cerebellum and thalamus masks against the manually segmented masks, the median DSC values for ANTs and FLIRT were 0.72 [interquartile range (IQR) = 0.6-0.8] and 0.54 (IQR = 0.4-0.6), respectively. A Friedman's test indicated that the ANTs registration methods, primarily nonlinear methods, performed better than FLIRT (p < 0.001). Deformable registration methods provided the most accurate results relative to manual segmentation. Overall, this semi-automatic subcortical segmentation method provides reliable performance to segment subcortical volumes in fetal MR images. This method reduces the costs of manual segmentation, facilitating the measurement of typical and atypical fetal brain development.

Increased volume of cerebral oedema is associated with risk of acute seizure activity and adverse neurological outcomes in encephalitis – regional and volumetric analysis in a multi-centre cohort

BackgroundSeizures can occur unpredictably in patients with acute encephalitis syndrome (AES), and many suffer from poor long-term neurological sequelae. Establishing factors associated with acute seizures risk and poor outcomes could support clinical care. We aimed to conduct regional and volumetric analysis of cerebral oedema on magnetic resonance imaging (MRI) in patients with AES. We assessed the relationship of brain oedema with acute seizure activity and long-term neurological outcome.MethodsIn a multi-centre cohort study, adults and children presenting with an AES were recruited in the UK. The clinical and brain MRI data were retrospectively reviewed. The outcomes variables were inpatient acute seizure activity and neurological disability at six-months post-discharge. A poor outcome was defined as a Glasgow outcome score (GOS) of 1–3. We quantified regional brain oedema on MRI through stereological examination of T2-weighted images using established methodology by independent and blinded assessors. Clinical and neuroimaging variables were analysed by multivariate logistic regression to assess for correlation with acute seizure activity and outcome.ResultsThe study cohort comprised 69 patients (mean age 31.8 years; 53.6% female), of whom 41 (59.4%) had acute seizures as inpatients. A higher Glasgow coma scale (GCS) score on admission was a negative predictor of seizures (OR 0.61 [0.46–0.83], p = 0.001). Even correcting for GCS on admission, the presence of cortical oedema was a significant risk factor for acute seizure activity (OR 5.48 [1.62–18.51], p = 0.006) and greater volume of cerebral oedema in these cortical structures increased the risk of acute seizures (OR 1.90 [1.12–3.21], p = 0.017). At six-month post-discharge, 21 (30.4%) had a poor neurological outcome. Herpes simplex virus encephalitis was associated with higher risk of poor outcomes in univariate analysis (OR 3.92 [1.08–14.20], p = 0.038). When controlling for aetiology, increased volume of cerebral oedema was an independent risk factor for adverse neurological outcome at 6 months (OR 1.73 [1.06–2.83], p = 0.027).ConclusionsBoth the presence and degree of cerebral oedema on MRIs of patients with AES may help identify patients at risk of acute seizure activity and subsequent long-term morbidity.

VP.78 Clinical and genetic study of LAMA2-related muscular dystrophy patients with seizures

<i>LAMA2</i>-related muscular dystrophy (LAMA2-MD) is an autosomal recessive muscular dystrophy caused by pathogenic variants in <i>LAMA2</i> gene, including two subgroups: severe, early-onset <i>LAMA2</i>-related congenital muscular dystrophy (LAMA2-CMD) and mild, late-onset autosomal recessive limb-girdle muscular dystrophy-23 (LGMD R23). Seizures frequently occurred in LAMA2-MD, however, the relationship between seizures and genotype in LAMA2-MD is not fully known. The clinical information of seizures including seizure type, frequency, treament, EEG changes, brain MRI and genetic data of LAMA2-MD patients from Peking University First Hospital between January 2003 and March 2022 were collected and analyzed. Among 141 LAMA2-MD patients, seizures were observed in 10.5% (13/124) of LAMA2-CMD patients and 29.4% (5/17) of LGMD R23 patients, including 14 patients with epilepsy and 4 patients with febrile seizures. The median (range) age of first seizure episode was 5 (0.1-20) years in LAMA2-CMD and 14 (1-24) years in LGMD R23. Eight patients with epilepsy showed focal seizures. Abnormal results of electroencephalography were found in 8/10 patients, manifesting as temporal, occipital or temporo-occipital epileptiform abnormalities. Brain magnetic resonance imaging (MRI) was performed in 16 patients with seizures, showing occipital pachygyria in 4 patients, cerebellum and brain stem dysplasia in 2 patients. Brain MRI of 41 LAMA2-MD patients without seizures were reassessed, showing occipital lobe pachygyria in 6 patients. Therefore, there was no significant association between seizures and occipital pachygyria in our cohort (P=0.18). Nine patients received antiepileptic drugs, while patients with occipital lobe pachygyria showed poor seizure control. The pathogenic variants included splice mutation (n=8), nonsense mutation (n=6) and frameshift mutation(n=6) in LAMA2-CMD patients, while missense mutation (n=5), frameshift mutation(n=4) in LGMD R23 patients. Compared with LAMA2-CMD, seizures were more frequently occurred in LGMD R23 patients and mainly manifested as focal seizures. No correlation between genotype and seizures was observed. In this study, we found that seizures occurred in 10.5% of LAMA2-CMD and 29.4% of LGMD R23 patients, and described the features of seizures in LAMA2-MD patients.

A computationally efficient learning model to classify audio signal attributes

&lt;p&gt;The era of machine learning has opened up groundbreaking realities and opportunities in the field of medical diagnosis. However, it is also observed that faster and proper diagnosis of any diseases/medical conditions require proper analysis and classification of digital signal data. It indicates the proper identification of tumors in the brain. Brain magnetic resonance imaging (MRI) data has to be appropriately classified, and similarly, pulse signal analysis is required to evaluate the human heart operating condition. Several studies have used machine learning (ML) modeling to classify speech signals, but very few studies have explored the classification of audio signal attributes in the context of intelligent healthcare monitoring. The study thereby aims to introduce novel mathematical modeling to analyze and classify synthetic pulse audio signal attributes with cost-effective computation. The numerical modeling is composed of several functional blocks where deep neural network-based learning (DNNL) plays a crucial role during the training phase, and also it is further combined with a recurrent structure of long-short term memory (R-LSTM) feedback connections (FCs). The design approaches further experiment in a numerical computing environment in terms of accuracy and computational aspects. The classification outcome of the proposed approach shows that it attains approximately 85% accuracy, which is comparable to the baseline approaches and execution time.&lt;/p&gt;

MCNN: a multi-level CNN model for the classification of brain tumors in IoT-healthcare system.

The classification of brain tumors is significantly important for diagnosing and treating brain tumors in IoT healthcare systems. In this work, we have proposed a robust classification model for brain tumors employing deep learning techniques. In the design of the proposed method, an improved Convolutional neural network is used to classify Meningioma, Glioma, and Pituitary types of brain tumors. To test the multi-level convolutional neural network model, brain magnetic resonance image data is utilized. The MCNN model classification results were improved using data augmentation and transfer learning methods. In addition, hold-out and performance evaluation metrics have been employed in the proposed MCNN model. The experimental results show that the proposed model obtained higher outcomes than the state-of-the-art techniques and achieved 99.89% classification accuracy. Due to the higher results of the proposed approach, we recommend it for the identification of brain cancer in IoT-healthcare systems.

DACBT: deep learning approach for classification of brain tumors using MRI data in IoT healthcare environment

The classification of brain tumors (BT) is significantly essential for the diagnosis of Brian cancer (BC) in IoT-healthcare systems. Artificial intelligence (AI) techniques based on Computer aided diagnostic systems (CADS) are mostly used for the accurate detection of brain cancer. However, due to the inaccuracy of artificial diagnostic systems, medical professionals are not effectively incorporating them into the diagnosis process of Brain Cancer. In this research study, we proposed a robust brain tumor classification method using Deep Learning (DL) techniques to address the lack of accuracy issue in existing artificial diagnosis systems. In the design of the proposed approach, an improved convolution neural network (CNN) is used to classify brain tumors employing brain magnetic resonance (MR) image data. The model classification performance has improved by incorporating data augmentation and transfer learning methods. The results confirmed that the model obtained high accuracy compared to the baseline models. Based on high predictive results we suggest the proposed model for brain cancer diagnosis in IoT-healthcare systems.

Anomalous cerebral morphology of pregnant women with cleft fetuses.

ObjectivePregnancy leads to long-lasting changes in brain structure for healthy women; however, little is known regarding alterations in the cortical features of pregnant women with malformed fetuses. Isolated clefts of the lip and/or palate (ICL/P) are the most common congenital anomaly in the craniofacial region, which is highly gene-associated. We speculated that pregnant women carrying fetuses with ICL/P may have associated risk genes and specific brain changes during pregnancy.MethodsIn this study, we investigated T1-weighted brain magnetic resonance imaging data from 48 pregnant women: 24 women carrying fetuses with ICL/P (ICL/P group) and 24 women carrying normal fetuses (normal controls), then explored intergroup differences in gray matter volume (GMV), cortical thickness (CT) and cortical complexity (gyrification).ResultsCompared with controls, the ICL/P group had decreased total intracranial volume (TIV) than normal controls; besides, they exhibited increased GMV in the left cuneus, decreased GMV in the right superior temporal gyrus; increased CT in the left precuneus and left superior parietal gyrus, decreased CT involving parsopercularis, fusiform, middle temporal in the left hemisphere and supramarginal, precentral gyrus (PreCG) in the right hemisphere; increased gyrification in the left insula and PreCG, the left middle temporal, and the right supratemporal gyrus.ConclusionPregnant women with ICL/P fetuses had brain morphology changes involving language, auditory, vision, and sensory cortex, which may be their special brain changes compared to normal pregnant women. This study may provide clues for the early detection of fetuses with ICL/P, and be vital for preconception and prenatal counseling with non-invasive methods.

Different loneliness types, cognitive function, and brain structure in midlife: Findings from the Framingham Heart Study.

SummaryBackgroundIt remains unclear whether persistent loneliness is related to brain structures that are associated with cognitive decline and development of Alzheimer's disease (AD). This study aimed to investigate the relationships between different loneliness types, cognitive functioning, and regional brain volumes.MethodsLoneliness was measured longitudinally, using the item from the Center for Epidemiologic Studies Depression Scale in the Framingham Heart Study, Generation 3, with participants’ average age of 46·3 ± 8·6 years. Robust regression models tested the association between different loneliness types with longitudinal neuropsychological performance (n = 2,609) and regional magnetic resonance imaging brain data (n = 1,829) (2002-2019). Results were stratified for sex, depression, and Apolipoprotein E4 (ApoE4).FindingsPersistent loneliness, but not transient loneliness, was strongly associated with cognitive decline, especially memory and executive function. Persistent loneliness was negatively associated with temporal lobe volume (β = −0.18, 95%CI [−0.32, −0.04], P = 0·01). Among women, persistent loneliness was associated with smaller frontal lobe (β = −0.19, 95%CI [−0.38, −0.01], P = 0·04), temporal lobe (β = −0.20, 95%CI [−0.37, −0.03], P = 0·02), and hippocampus volumes (β = −0.23, 95%CI [−0.40, −0.06], P = 0·007), and larger lateral ventricle volume (β = 0.15, 95%CI [0.02, 0.28], P = 0·03). The higher cumulative loneliness scores across three exams, the smaller parietal, temporal, and hippocampus volumes and larger lateral ventricle were evident, especially in the presence of ApoE4.InterpretationPersistent loneliness in midlife was associated with atrophy in brain regions responsible for memory and executive dysfunction. Interventions to reduce the chronicity of loneliness may mitigate the risk of age-related cognitive decline and AD.FundingUS National Institute on Aging.

Prediction of Alzheimer\'s Disease Using CNN

Abstract: Alzheimer’s disease is an incurable, progressive neurological brain disorder. Earlier detection of Alzheimer’s disease can help with proper treatment and prevent brain tissue damage. Several statistical and machine learning models have been exploited by researchers for Alzheimer’s disease diagnosis. Analysing magnetic resonance imaging (MRI) is a common practice for Alzheimer’s disease diagnosis in clinical research. Detection of Alzheimer’s disease is exacting due to the similarity in Alzheimer’s disease MRI data and standard healthy MRI data of older people. Recently, advanced deep learning techniques have successfully demonstrated human-level performance in numerous fields including medical image analysis. We propose a deep convolutional neural network for Alzheimer’s disease diagnosis using brain MRI data analysis. While most of the existing approaches perform binary classification, our model can identify different stages of Alzheimer’s disease and obtains superior performance for early-stage diagnosis.

Impact of a multidomain lifestyle intervention on regional spontaneous brain activity

In the SoUth Korean study to PrEvent cognitive impaiRment and protect BRAIN health through lifestyle intervention in at-risk elderly people (SUPERBRAIN), we evaluated the impact of multidomain lifestyle intervention on regional homogeneity (ReHo) in resting-state functional brain magnetic resonance imaging (MRI) data. Of 152 participants aged 60–79 years without dementia assigned to either facility-based multidomain intervention (FMI), home-based MI, or controls, we analyzed 56 scanned MRIs at baseline and 24 weeks. ReHo values from regions with significant longitudinal changes were compared between the intervention and control groups and their correlations with the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) or serum brain-derived neurotrophic factor (BDNF) were evaluated. ReHo values in the left medial orbitofrontal gyrus and right superior parietal lobule were increased [p = 0.021, correlated positively with serum BDNF changes (r = 0.504, p = 0.047)] and decreased [p = 0.021, correlated negatively with changes in the total (r = −0.509, p = 0.044) and attention (r = −0.562, p = 0.023). RBANS], respectively, in the participants assigned to the FMI group than those of the controls. Our results suggest that facility-based group preventive strategies may have cognitive benefits through neuroplastic changes in functional processing circuits in the brain areas which play a crucial role in the adaptive learning and internally directed cognition.

Multimodal MRI-Based Whole-Brain Assessment in Patients In Anoxoischemic Coma by Using 3D Convolutional Neural Networks

BackgroundThere is an unfulfilled need to find the best way to automatically capture, analyze, organize, and merge structural and functional brain magnetic resonance imaging (MRI) data to ultimately extract relevant signals that can assist the medical decision process at the bedside of patients in postanoxic coma. We aimed to develop and validate a deep learning model to leverage multimodal 3D MRI whole-brain times series for an early evaluation of brain damages related to anoxoischemic coma.MethodsThis proof-of-concept, prospective, cohort study was undertaken at the intensive care unit affiliated with the University Hospital (Toulouse, France), between March 2018 and May 2020. All patients were scanned in coma state at least 2 days (4 ± 2 days) after cardiac arrest. Over the same period, age-matched healthy volunteers were recruited and included. Brain MRI quantification encompassed both “functional data” from regions of interest (precuneus and posterior cingulate cortex) with whole-brain functional connectivity analysis and “structural data” (gray matter volume, T1-weighted, fractional anisotropy, and mean diffusivity). A specifically designed 3D convolutional neuronal network (CNN) was created to allow conscious state discrimination (coma vs. controls) by using raw MRI indices as the input. A voxel-wise visualization method based on the study of convolutional filters was applied to support CNN outcome. The Ethics Committee of the University Teaching Hospital of Toulouse, France (2018-A31) approved the study and informed consent was obtained from all participants.ResultsThe final cohort consisted of 29 patients in postanoxic coma and 34 healthy volunteers. Coma patients were successfully discerned from controls by using 3D CNN in combination with different MR indices. The best accuracy was achieved by functional MRI data, in particular with resting-state functional MRI of the posterior cingulate cortex, with an accuracy of 0.96 (range 0.94–0.98) on the test set from 10-time repeated tenfold cross-validation. Even more satisfactory performances were achieved through the majority voting strategy, which was able to compensate for mistakes from single MR indices. Visualization maps allowed us to identify the most relevant regions for each MRI index, notably regions previously described as possibly being involved in consciousness emergence. Interestingly, a posteriori analysis of misclassified patients indicated that they may present some common functional MRI traits with controls, which suggests further favorable outcomes.ConclusionsA fully automated identification of clinically relevant signals from complex multimodal neuroimaging data is a major research topic that may bring a radical paradigm shift in the neuroprognostication of patients with severe brain injury. We report for the first time a successful discrimination between patients in postanoxic coma patients from people serving as controls by using 3D CNN whole-brain structural and functional MRI data.Clinical Trial Numberhttp://ClinicalTrials.gov (No. NCT03482115).

SM-SegNet: A Lightweight Squeeze M-SegNet for Tissue Segmentation in Brain MRI Scans.

In this paper, we propose a novel squeeze M-SegNet (SM-SegNet) architecture featuring a fire module to perform accurate as well as fast segmentation of the brain on magnetic resonance imaging (MRI) scans. The proposed model utilizes uniform input patches, combined-connections, long skip connections, and squeeze–expand convolutional layers from the fire module to segment brain MRI data. The proposed SM-SegNet architecture involves a multi-scale deep network on the encoder side and deep supervision on the decoder side, which uses combined-connections (skip connections and pooling indices) from the encoder to the decoder layer. The multi-scale side input layers support the deep network layers’ extraction of discriminative feature information, and the decoder side provides deep supervision to reduce the gradient problem. By using combined-connections, extracted features can be transferred from the encoder to the decoder resulting in recovering spatial information, which makes the model converge faster. Long skip connections were used to stabilize the gradient updates in the network. Owing to the adoption of the fire module, the proposed model was significantly faster to train and offered a more efficient memory usage with 83% fewer parameters than previously developed methods, owing to the adoption of the fire module. The proposed method was evaluated using the open-access series of imaging studies (OASIS) and the internet brain segmentation registry (IBSR) datasets. The experimental results demonstrate that the proposed SM-SegNet architecture achieves segmentation accuracies of 95% for cerebrospinal fluid, 95% for gray matter, and 96% for white matter, which outperforms the existing methods in both subjective and objective metrics in brain MRI segmentation.

COMPARITIVE ANALYSIS OF K++GKFCM CLUSTERING FOR BRAIN TUMOUR DETECTION

Due to the complicated structure of brain tumors, hazy borders, and outside variables like noise, inferring tumor and edema regions from brain magnetic resonance imaging (MRI) data is still difficult. In this paper, a powerful hybrid clustering technique together with morphological procedures is suggested for segmenting brain tumors in order to reduce noise sensitivity and enhance segmentation stability. The following are the paper's key contributions: initially, adaptive Wiener filtering is employed for de-noising, and morphological procedures are applied for deleting non-brain tissue, thereby minimizing the method’s susceptibility to noise. Second, to segment pictures, the fuzzy C-means technique based on a Gaussian kernel is used with K-means++ clustering. This clustering decreases the sensitivity of the clustering parameters while simultaneously enhancing the stability of the algorithm. To produce accurate representations of brain tumours, the retrieved tumour pictures are lastly post processed utilising morphological procedures and median filtering. The suggested approach was also contrasted with other existing segmentation algorithms.

Visual rating of cortical atrophy in migraine and tension-type headache

Background: Studies of cortical atrophy in migraine and tension-type headache (TTH) have reported inconsistent results. Some studies have reported decreased cortical thickness or volume, while others have reported thickening in the somatosensory and visual cortices in migraine patients. The same controversy is true with respect to episodic and chronic headaches. This study evaluated and compared cortical atrophy in migraine and TTH as well as in episodic and chronic headache. Methods: The study included 43 (37 female) migraine and 15 (12 female) TTH patients. Demographic, clinical, and brain magnetic resonance imaging (MRI) data were evaluated. 1.5 Tesla MRI brain coronal T2 sequences were reviewed by two neurologists, and atrophy was graded according to a visual rating scale (VRS). Two neurologists jointly evaluated and scored cortical thickness on a Likert scale of 0 to 3 or 4 at the following sites bilaterally: olfactory sulcus, anterior cingulate sulcus, circular insular sulcus, anterior temporal and medial temporal lobes, and posterior cingulate and parieto-occipital sulci. The scores at all sites were then summed into a total atrophy score (TAS) for each subject. Results: There was no significant difference between the migraine and TTH groups regarding demographics, clinical findings, and atrophy scores in any particular region or with respect to the TAS. When we divided our study subjects into episodic and chronic headache groups, there was no significant difference between these groups in terms of demographic and clinical findings or individual and total atrophy scores. Conclusions: These results are in line with previous studies that have reported no differences between migraine and TTH in terms of cortical atrophy.

Bleeding Complications in Neonates Receiving Extracorporeal Membrane Oxygenation and Controlled Hypothermia.

Safety and efficacy data on controlled hypothermia (CH) for neonates with moderate to severe hypoxic ischemic encephalopathy has been extrapolated to a subgroup of these patients who also require extracorporeal membrane oxygenation for refractory persistent pulmonary hypertension of the newborn (PPHN). However, safety data on the concomitant use of CH and extracorporeal membrane oxygenation (ECMO) are lacking. This is a single-center retrospective study of neonates ≥35 weeks' gestation with refractory PPHN who required ECMO between January 2010 and December 2020. Study groups were divided into those receiving CH/ECMO versus ECMO only. Baseline characteristics, short-term outcomes, and brain magnetic resonance imaging (MRI) data were compared. A total of 36 neonates who received ECMO for refractory PPHN were included. Of these, 44.4% (n = 16) received CH/ECMO and 55.6% (n = 20) received ECMO only. Bleeding complications were more common in CH/ECMO group 50% (n = 8) versus ECMO only 15% (n = 3, p = 0.023). T1 brain MRI severity scores were higher in CH/ECMO group versus ECMO only group, however, there were no statistical difference in T2 and diffusion-weighted image scores. Functional status and survival to discharge were comparable between groups. In our cohort, neonates who received CH/ECMO had higher bleeding complications than ECMO only group with comparable functional status and survival at discharge. · Safety data on the concomitant use of CH and ECMO are lacking in neonates.. · In our cohort, neonates who received CH/ECMO had higher bleeding complications than ECMO only group.. · Functional status and survival to discharge were no differences between the two groups..

Generation of Synthetic Rat Brain MRI Scans with a 3D Enhanced Alpha Generative Adversarial Network

Translational brain research using Magnetic Resonance Imaging (MRI) is becoming increasingly popular as animal models are an essential part of scientific studies and more ultra-high-field scanners are becoming available. Some disadvantages of MRI are the availability of MRI scanners and the time required for a full scanning session. Privacy laws and the 3Rs ethics rule also make it difficult to create large datasets for training deep learning models. To overcome these challenges, an adaptation of the alpha Generative Adversarial Networks (GANs) architecture was used to test its ability to generate realistic 3D MRI scans of the rat brain in silico. As far as the authors are aware, this was the first time a GAN-based approach was used to generate synthetic MRI data of the rat brain. The generated scans were evaluated using various quantitative metrics, a Turing test, and a segmentation test. The last two tests proved the realism and applicability of the generated scans to real problems. Therefore, by using the proposed new normalisation layer and loss functions, it was possible to improve the realism of the generated rat MRI scans, and it was shown that using the generated data improved the segmentation model more than using the conventional data augmentation.

Atlas-ISTN: Joint segmentation, registration and atlas construction with image-and-spatial transformer networks.

Deep learning models for semantic segmentation are able to learn powerful representations for pixel-wise predictions, but are sensitive to noise at test time and may lead to implausible topologies. Image registration models on the other hand are able to warp known topologies to target images as a means of segmentation, but typically require large amounts of training data, and have not widely been benchmarked against pixel-wise segmentation models. We propose the Atlas Image-and-Spatial Transformer Network (Atlas-ISTN), a framework that jointly learns segmentation and registration on 2D and 3D image data, and constructs a population-derived atlas in the process. Atlas-ISTN learns to segment multiple structures of interest and to register the constructed atlas labelmap to an intermediate pixel-wise segmentation. Additionally, Atlas-ISTN allows for test time refinement of the model's parameters to optimize the alignment of the atlas labelmap to an intermediate pixel-wise segmentation. This process both mitigates for noise in the target image that can result in spurious pixel-wise predictions, as well as improves upon the one-pass prediction of the model. Benefits of the Atlas-ISTN framework are demonstrated qualitatively and quantitatively on 2D synthetic data and 3D cardiac computed tomography and brain magnetic resonance image data, out-performing both segmentation and registration baseline models. Atlas-ISTN also provides inter-subject correspondence of the structures of interest.

Backdoor Attacks Against Transfer Learning With Pre-Trained Deep Learning Models

Transfer learning provides an effective solution for feasibly and fast customize accurate \textit{Student} models, by transferring the learned knowledge of pre-trained \textit{Teacher} models over large datasets via fine-tuning. Many pre-trained Teacher models used in transfer learning are publicly available and maintained by public platforms, increasing their vulnerability to backdoor attacks. In this paper, we demonstrate a backdoor threat to transfer learning tasks on both image and time-series data leveraging the knowledge of publicly accessible Teacher models, aimed at defeating three commonly-adopted defenses: \textit{pruning-based}, \textit{retraining-based} and \textit{input pre-processing-based defenses}. Specifically, (A) ranking-based selection mechanism to speed up the backdoor trigger generation and perturbation process while defeating \textit{pruning-based} and/or \textit{retraining-based defenses}. (B) autoencoder-powered trigger generation is proposed to produce a robust trigger that can defeat the \textit{input pre-processing-based defense}, while guaranteeing that selected neuron(s) can be significantly activated. (C) defense-aware retraining to generate the manipulated model using reverse-engineered model inputs. We launch effective misclassification attacks on Student models over real-world images, brain Magnetic Resonance Imaging (MRI) data and Electrocardiography (ECG) learning systems. The experiments reveal that our enhanced attack can maintain the $98.4\%$ and $97.2\%$ classification accuracy as the genuine model on clean image and time series inputs respectively while improving $27.9\%-100\%$ and $27.1\%-56.1\%$ attack success rate on trojaned image and time series inputs respectively in the presence of pruning-based and/or retraining-based defenses.