Unpaired intra-operative OCT (iOCT) video super-resolution with contrastive learning

Regenerative therapies show promise in reversing sight loss caused by degenerative eye diseases. Their precise subretinal delivery can be facilitated by robotic systems alongside with Intra-operative Optical Coherence Tomography (iOCT). However, iOCT's real-time retinal layer information is compromised by inferior image quality. To address this limitation, we introduce an unpaired video super-resolution methodology for iOCT quality enhancement. A recurrent network is proposed to leverage temporal information from iOCT sequences, and spatial information from pre-operatively acquired OCT images. Additionally, a patchwise contrastive loss enables unpaired super-resolution. Extensive quantitative analysis demonstrates that our approach outperforms existing state-of-the-art iOCT super-resolution models. Furthermore, ablation studies showcase the importance of temporal aggregation and contrastive loss in elevating iOCT quality. A qualitative study involving expert clinicians also confirms this improvement. The comprehensive evaluation demonstrates our method's potential to enhance the iOCT image quality, thereby facilitating successful guidance for regenerative therapies.

Glioblastoma and Radiotherapy: a multi-center AI study for Survival Predictions from MRI (GRASP study)

BACKGROUND

The aim was to predict survival of glioblastoma at eight months after radiotherapy (a period allowing for completing a typical course of adjuvant temozolomide), by applying deep learning to the first brain MRI after radiotherapy completion.

METHODS

Retrospective and prospective data were collected from 206 consecutive glioblastoma, IDH-wildtype patients diagnosed between March 2014-February 2022 across 11 UK centers. Models were trained on 158 retrospective patients from three centers. Holdout test sets were retrospective (n=19; internal validation), and prospective (n=29; external validation from eight distinct centers).Neural network branches for T2-weighted and contrast-enhanced T1-weighted inputs were concatenated to predict survival. A non-imaging branch (demographics/MGMT/treatment data) was also combined with the imaging model. We investigated the influence of individual MR sequences; non-imaging features; and weighted dense blocks pretrained for abnormality detection.

RESULTS

The imaging model outperformed the non-imaging model in all test sets (area under the receiver-operating characteristic curve, AUC p=0.038) and performed similarly to a combined imaging/non-imaging model (p>0.05). Imaging, non-imaging, and combined models applied to amalgamated test sets gave AUCs of 0.93, 0.79, and 0.91. Initializing the imaging model with pretrained weights from 10,000s of brain MRIs improved performance considerably (amalgamated test sets without pretraining 0.64; p=0.003).

CONCLUSIONS

A deep learning model using MRI images after radiotherapy, reliably and accurately determined survival of glioblastoma. The model serves as a prognostic biomarker identifying patients who will not survive beyond a typical course of adjuvant temozolomide, thereby stratifying patients into those who might require early second-line or clinical trial treatment.

Placental differences between uncomplicated and complicated monochorionic diamniotic pregnancies on diffusion and multicompartment Magnetic Resonance Imaging

INTRODUCTION

Twin-twin transfusion syndrome (TTTS) and selective fetal growth restriction (sFGR) are common complications in monochorionic diamniotic (MCDA) pregnancies. The Diffusion-rElaxation Combined Imaging for Detailed Placental Evaluation (DECIDE) model, a placental-specific model, separates the T2 values of the fetal and maternal blood from the background tissue and estimates the fetal blood oxygen saturation. This study investigates diffusion and relaxation differences in uncomplicated MCDA pregnancies and MCDA pregnancies complicated by TTTS and sFGR in mid-pregnancy.

METHODS

This prospective monocentric cohort study included uncomplicated MCDA pregnancies and pregnancies complicated by TTTS and sFGR. We performed MRI with conventional diffusion-weighted imaging (DWI) and combined relaxometry - DWI-intravoxel incoherent motion. DECIDE analysis was used to quantify different parameters within the placenta related to the fetal, placental, and maternal compartments.

RESULTS

We included 99 pregnancies, of which 46 were uncomplicated, 12 were complicated by sFGR and 41 by TTTS. Conventional DWI did not find differences between or within cohorts. On DECIDE imaging, fetoplacental oxygen saturation was significantly lower in the smaller member of sFGR (p = 0.07) and in both members of TTTS (p = 0.01 and p = 0.004) compared to the uncomplicated pairs. Additionally, average T2 relaxation time was significantly lower in the smaller twin of the sFGR (p = 0.004) compared to the uncomplicated twins (p = 0.03).

CONCLUSION

Multicompartment functional MRI showed significant differences in several MRI parameters between the placenta of uncomplicated MCDA pregnancies and those complicated by sFGR and TTTS in mid-pregnancy.

Application of Automatic Segmentation on Super-Resolution Reconstruction MR Images of the Abnormal Fetal Brain

BACKGROUND AND PURPOSE

Fetal brain MR imaging is clinically used to characterize fetal brain abnormalities. Recently, algorithms have been proposed to reconstruct high-resolution 3D fetal brain volumes from 2D slices. By means of these reconstructions, convolutional neural networks have been developed for automatic image segmentation to avoid labor-intensive manual annotations, usually trained on data of normal fetal brains. Herein, we tested the performance of an algorithm specifically developed for segmentation of abnormal fetal brains.

MATERIALS AND METHODS

This was a single-center retrospective study on MR images of 16 fetuses with severe CNS anomalies (gestation, 21-39 weeks). T2-weighted 2D slices were converted to 3D volumes using a super-resolution reconstruction algorithm. The acquired volumetric data were then processed by a novel convolutional neural network to perform segmentations of white matter and the ventricular system and cerebellum. These were compared with manual segmentation using the Dice coefficient, Hausdorff distance (95th percentile), and volume difference. Using interquartile ranges, we identified outliers of these metrics and further analyzed them in detail.

RESULTS

The mean Dice coefficient was 96.2%, 93.7%, and 94.7% for white matter and the ventricular system and cerebellum, respectively. The Hausdorff distance was 1.1, 2.3, and 1.6 mm, respectively. The volume difference was 1.6, 1.4, and 0.3 mL, respectively. Of the 126 measurements, there were 16 outliers among 5 fetuses, discussed on a case-by-case basis.

CONCLUSIONS

Our novel segmentation algorithm obtained excellent results on MR images of fetuses with severe brain abnormalities. Analysis of the outliers shows the need to include pathologies underrepresented in the current data set. Quality control to prevent occasional errors is still needed.

Neuroprotection is improved by watertightness of fetal spina bifida repair in the sheep model

OBJECTIVES

A contributing factor to unsuccessful prenatal spina bifida aperta (SBA) repair via an open approach may be incomplete neurosurgical repair causing persistent in-utero leakage of cerebrospinal fluid (CSF) and exposure of the fetal spinal cord to amniotic fluid. We aimed to investigate the neurostructural and neurofunctional efficacy of watertight prenatal SBA repair in a validated SBA fetal lamb model.

METHODS

A well-powered superiority study was conducted in the validated SBA fetal lamb model (n = 7 per group). The outcomes of lambs which underwent watertight or non-watertight multilayer repair through an open approach were compared to those of unrepaired SBA lambs (historical controls) at delivery (term = 145 days). At ∼75 days, fetal lambs underwent standardized induction of lumbar SBA. At ∼100 days, they were assigned to an either watertight or non-watertight layered repair group based on an intraoperative watertightness test using subcutaneous fluorescein injection. At 1-2 days postnatally, as primary outcome, we assessed reversal of hindbrain herniation using magnetic resonance imaging (MRI). Secondary proxies of neuroprotection were: absence of CSF leakage at the repair site; hindlimb motor function based on joint-movement score, locomotor grade and Motor Evoked Potential (MEP); four-score neuroprotection scale, encompassing live birth, complete hindbrain herniation reversal, absence of CSF leakage and joint-movement score ≥ 9/15; and brain and spinal cord histology and immunohistochemistry. As the watertightness test cannot be used clinically due to its invasiveness, we developed a potential surrogate intraoperative three-score skin-repair-quality scale based on visual assessment of the quality of the skin repair (suture inter-run distance ≤ 3 mm, absence of tear and absence of ischemia), with high quality defined by a score ≥ 2/3 and low quality by a score < 2/3, and assessed its relationship with improved outcome.

RESULTS

Compared with unrepaired lambs, lambs with watertight repair achieved a high level of neuroprotection (neuroprotection score of 4/4 in 5/7 vs 0/7 lambs) as evidenced by: a significant 100% (vs 14%) reversal of hindbrain herniation on MRI; low CSF leakage (14% vs 100%); better hindlimb motor function, with higher joint-movement score, locomotor grade and MEP area under the curve and peak-to-peak amplitude; higher neuronal density in the hippocampus and corpus callosum; and higher reactive astrogliosis at the SBA lesion epicenter. Conversely, lambs with non-watertight SBA repair did not achieve the same level of neuroprotection (score of 4/4 in 1/7 lambs) compared with unrepaired lambs, with: a non-significant 86% (vs 14%) reversal of hindbrain herniation; high CSF leakage (43% vs 100%); no improvement in motor function; low brain neuron count in both the hippocampus and corpus callosum; and small spinal astroglial cell area at the epicenter. Both watertight layered repair and high (≥ 2/3) intraoperative skin-repair-quality score were associated with improved outcome, but the watertightness test and skin-repair-quality scale could not be used interchangeably due to result discrepancies.

CONCLUSIONS

Watertight layered fetal SBA repair is neuroprotective since it improves brain and spinal-cord structure and function in the fetal lamb model. This translational research has important clinical implications. A neurosurgical technique that achieves watertightness should be adopted in all fetal centers to improve neuroprotection. Future clinical studies could assess whether a high skin-repair-quality score (≥ 2/3) correlates with neuroprotection. © 2022 International Society of Ultrasound in Obstetrics and Gynecology.

CrossMoDA 2021 challenge: Benchmark of cross-modality domain adaptation techniques for vestibular schwannoma and cochlea segmentation

Domain Adaptation (DA) has recently been of strong interest in the medical imaging community. While a large variety of DA techniques have been proposed for image segmentation, most of these techniques have been validated either on private datasets or on small publicly available datasets. Moreover, these datasets mostly addressed single-class problems. To tackle these limitations, the Cross-Modality Domain Adaptation (crossMoDA) challenge was organised in conjunction with the 24th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2021). CrossMoDA is the first large and multi-class benchmark for unsupervised cross-modality Domain Adaptation. The goal of the challenge is to segment two key brain structures involved in the follow-up and treatment planning of vestibular schwannoma (VS): the VS and the cochleas. Currently, the diagnosis and surveillance in patients with VS are commonly performed using contrast-enhanced T1 (ceT₁) MR imaging. However, there is growing interest in using non-contrast imaging sequences such as high-resolution T2 (hrT₂) imaging. For this reason, we established an unsupervised cross-modality segmentation benchmark. The training dataset provides annotated ceT₁ scans (N=105) and unpaired non-annotated hrT₂ scans (N=105). The aim was to automatically perform unilateral VS and bilateral cochlea segmentation on hrT₂ scans as provided in the testing set (N=137). This problem is particularly challenging given the large intensity distribution gap across the modalities and the small volume of the structures. A total of 55 teams from 16 countries submitted predictions to the validation leaderboard. Among them, 16 teams from 9 different countries submitted their algorithm for the evaluation phase. The level of performance reached by the top-performing teams is strikingly high (best median Dice score - VS: 88.4%; Cochleas: 85.7%) and close to full supervision (median Dice score - VS: 92.5%; Cochleas: 87.7%). All top-performing methods made use of an image-to-image translation approach to transform the source-domain images into pseudo-target-domain images. A segmentation network was then trained using these generated images and the manual annotations provided for the source image.

Toward semi-automatic biologically effective dose treatment plan optimisation for Gamma Knife radiosurgery

Objective.Dose-rate effects in Gamma Knife radiosurgery treatments can lead to varying biologically effective dose (BED) levels for the same physical dose. The non-convex BED model depends on the delivery sequence and creates a non-trivial treatment planning problem. We investigate the feasibility of employing inverse planning methods to generate treatment plans exhibiting desirable BED characteristics using the per iso-centre beam-on times and delivery sequence.Approach.We implement two dedicated optimisation algorithms. One approach relies on mixed-integer linear programming (MILP) using a purposely developed convex underestimator for the BED to mitigate local minima issues at the cost of computational complexity. The second approach (local optimisation) is faster and potentially usable in a clinical setting but more prone to local minima issues. It sequentially executes the beam-on time (quasi-Newton method) and sequence optimisation (local search algorithm). We investigate the trade-off between time to convergence and solution quality by evaluating the resulting treatment plans' objective function values and clinical parameters. We also study the treatment time dependence of the initial and optimised plans using BED95(BED delivered to 95% of the target volume) values.Main results.When optimising the beam-on times and delivery sequence, the local optimisation approach converges several orders of magnitude faster than the MILP approach (minutes versus hours-days) while typically reaching within 1.2% (0.02-2.08%) of the final objective function value. The quality parameters of the resulting treatment plans show no meaningful difference between the local and MILP optimisation approaches. The presented optimisation approaches remove the treatment time dependence observed in the original treatment plans, and the chosen objectives successfully promote more conformal treatments.Significance.We demonstrate the feasibility of using an inverse planning approach within a reasonable time frame to ensure BED-based objectives are achieved across varying treatment times and highlight the prospect of further improvements in treatment plan quality.

Software tool for visualization of a probabilistic map of the epileptogenic zone from seizure semiologies

Around one third of epilepsies are drug-resistant. For these patients, seizures may be reduced or cured by surgically removing the epileptogenic zone (EZ), which is the portion of the brain giving rise to seizures. If noninvasive data are not sufficiently lateralizing or localizing, the EZ may need to be localized by precise implantation of intracranial electroencephalography (iEEG) electrodes. The choice of iEEG targets is influenced by clinicians' experience and personal knowledge of the literature, which leads to substantial variations in implantation strategies across different epilepsy centers. The clinical diagnostic pathway for surgical planning could be supported and standardized by an objective tool to suggest EZ locations, based on the outcomes of retrospective clinical cases reported in the literature. We present an open-source software tool that presents clinicians with an intuitive and data-driven visualization to infer the location of the symptomatogenic zone, that may overlap with the EZ. The likely EZ is represented as a probabilistic map overlaid on the patient's images, given a list of seizure semiologies observed in that specific patient. We demonstrate a case study on retrospective data from a patient treated in our unit, who underwent resective epilepsy surgery and achieved 1-year seizure freedom after surgery. The resected brain structures identified as EZ location overlapped with the regions highlighted by our tool, demonstrating its potential utility.

A spatio-temporal atlas of the developing fetal brain with spina bifida aperta

Background: Spina bifida aperta (SBA) is a birth defect associated with severe anatomical changes in the developing fetal brain. Brain magnetic resonance imaging (MRI) atlases are popular tools for studying neuropathology in the brain anatomy, but previous fetal brain MRI atlases have focused on the normal fetal brain. We aimed to develop a spatio-temporal fetal brain MRI atlas for SBA. Methods: We developed a semi-automatic computational method to compute the first spatio-temporal fetal brain MRI atlas for SBA. We used 90 MRIs of fetuses with SBA with gestational ages ranging from 21 to 35 weeks. Isotropic and motion-free 3D reconstructed MRIs were obtained for all the examinations. We propose a protocol for the annotation of anatomical landmarks in brain 3D MRI of fetuses with SBA with the aim of making spatial alignment of abnormal fetal brain MRIs more robust. In addition, we propose a weighted generalized Procrustes method based on the anatomical landmarks for the initialization of the atlas. The proposed weighted generalized Procrustes can handle temporal regularization and missing annotations. After initialization, the atlas is refined iteratively using non-linear image registration based on the image intensity and the anatomical land-marks. A semi-automatic method is used to obtain a parcellation of our fetal brain atlas into eight tissue types: white matter, ventricular system, cerebellum, extra-axial cerebrospinal fluid, cortical gray matter, deep gray matter, brainstem, and corpus callosum. Results: An intra-rater variability analysis suggests that the seven anatomical land-marks are sufficiently reliable. We find that the proposed atlas outperforms a normal fetal brain atlas for the automatic segmentation of brain 3D MRI of fetuses with SBA. Conclusions: We make publicly available a spatio-temporal fetal brain MRI atlas for SBA, available here:  https://doi.org/10.7303/syn25887675. This atlas can support future research on automatic segmentation methods for brain 3D MRI of fetuses with SBA.

Accelerated vascular aging: Ethnic differences in basilar artery length and diameter, and its association with cardiovascular risk factors and cerebral small vessel disease

Background and aims

Risk of stroke and dementia is markedly higher in people of South Asian and African Caribbean descent than white Europeans in the UK. This is unexplained by cardiovascular risk factors (CVRF). We hypothesized this might indicate accelerated early vascular aging (EVA) and that EVA might account for stronger associations between cerebral large artery characteristics and markers of small vessel disease.

Methods

360 participants in a tri-ethnic population-based study (120 per ethnic group) underwent cerebral and vertebral MRI. Length and median diameter of the basilar artery (BA) were derived from Time of Flight images, while white matter hyperintensities (WMH) volumes were obtained from T1 and FLAIR images. Associations between BA characteristics and CVRF were assessed using multivariable linear regression. Partial correlation coefficients between WMH load and BA characteristics were calculated after adjustment for CVRF and other potential confounders.

Results

BA diameter was strongly associated with age in South Asians (+11.3 μm/year 95% CI = [3.05; 19.62]; p = 0.008), with unconvincing relationships in African Caribbeans (3.4 μm/year [-5.26, 12.12]; p = 0.436) or Europeans (2.6 μm/year [-5.75, 10.87]; p = 0.543). BA length was associated with age in South Asians (+0.34 mm/year [0.02; 0.65]; p = 0.037) and African Caribbeans (+0.39 mm/year [0.12; 0.65]; p = 0.005) but not Europeans (+0.08 mm/year [-0.26; 0.41]; p = 0.653). BA diameter (rho = 0.210; p = 0.022) and length (rho = 0.261; p = 0.004) were associated with frontal WMH load in South Asians (persisting after multivariable adjustment for CVRF).

Conclusions

Compared with Europeans, the basilar artery undergoes more accelerated EVA in South Asians and in African Caribbeans, albeit to a lesser extent. Such EVA may contribute to the higher burden of CSVD observed in South Asians and excess risk of stroke, vascular cognitive impairment and dementia observed in these ethnic groups.

Advanced quantitative evaluation of PET systems using the ACR phantom and NiftyPET software

PURPOSE

A novel phantom-imaging platform, a set of software tools, for automated and high-precision imaging of the American College of Radiology (ACR) positron emission tomography (PET) phantom for PET/magnetic resonance (PET/MR) and PET/computed tomography (PET/CT) systems is proposed.

METHODS

The key feature of this platform is the vector graphics design that facilitates the automated measurement of the knife-edge response function and hence image resolution, using composite volume of interest templates in a 0.5 mm resolution grid applied to all inserts of the phantom. Furthermore, the proposed platform enables the generation of an accurate μ $\mu$ -map for PET/MR systems with a robust alignment based on two-stage image registration using specifically designed PET templates. The proposed platform is based on the open-source NiftyPET software package used to generate multiple list-mode data bootstrap realizations and image reconstructions to determine the precision of the two-stage registration and any image-derived statistics. For all the analyses, iterative image reconstruction was employed with and without modeled shift-invariant point spread function and with varying iterations of the ordered subsets expectation maximization (OSEM) algorithm. The impact of the activity outside the field of view (FOV) was assessed using two acquisitions of 30 min each, with and without the activity outside the FOV.

RESULTS

The utility of the platform has been demonstrated by providing a standard and an advanced phantom analysis including the estimation of spatial resolution using all cylindrical inserts. In the imaging planes close to the edge of the axial FOV, we observed deterioration in the quantitative accuracy, reduced resolution (FWHM increased by 1-2 mm), reduced contrast, and background uniformity due to the activity outside the FOV. Although it slows convergence, the PSF reconstruction had a positive impact on resolution and contrast recovery, but the degree of improvement depended on the regions. The uncertainty analysis based on bootstrap resampling of raw PET data indicated high precision of the two-stage registration.

CONCLUSIONS

We demonstrated that phantom imaging using the proposed methodology with the metric of spatial resolution and multiple bootstrap realizations may be helpful in more accurate evaluation of PET systems as well as in facilitating fine tuning for optimal imaging parameters in PET/MR and PET/CT clinical research studies.

Surgical biomicroscopy-guided intra-operative optical coherence tomography (iOCT) image super-resolution

Purpose

Intra-retinal delivery of novel sight-restoring therapies will require the precision of robotic systems accompanied by excellent visualisation of retinal layers. Intra-operative Optical Coherence Tomography (iOCT) provides cross-sectional retinal images in real time but at the cost of image quality that is insufficient for intra-retinal therapy delivery.This paper proposes a super-resolution methodology that improves iOCT image quality leveraging spatiotemporal consistency of incoming iOCT video streams.

Methods

To overcome the absence of ground truth high-resolution (HR) images, we first generate HR iOCT images by fusing spatially aligned iOCT video frames. Then, we automatically assess the quality of the HR images on key retinal layers using a deep semantic segmentation model. Finally, we use image-to-image translation models (Pix2Pix and CycleGAN) to enhance the quality of LR images via quality transfer from the estimated HR domain.

Results

Our proposed methodology generates iOCT images of improved quality according to both full-reference and no-reference metrics. A qualitative study with expert clinicians also confirms the improvement in the delineation of pertinent layers and in the reduction of artefacts. Furthermore, our approach outperforms conventional denoising filters and the learning-based state-of-the-art.

Conclusions

The results indicate that the learning-based methods using the estimated, through our pipeline, HR domain can be used to enhance the iOCT image quality. Therefore, the proposed method can computationally augment the capabilities of iOCT imaging helping this modality support the vitreoretinal surgical interventions of the future.

Deep homography estimation in dynamic surgical scenes for laparoscopic camera motion extraction

Current laparoscopic camera motion automation relies on rule-based approaches or only focuses on surgical tools. Imitation Learning (IL) methods could alleviate these shortcomings, but have so far been applied to oversimplified setups. Instead of extracting actions from oversimplified setups, in this work we introduce a method that allows to extract a laparoscope holder's actions from videos of laparoscopic interventions. We synthetically add camera motion to a newly acquired dataset of camera motion free da Vinci surgery image sequences through a novel homography generation algorithm. The synthetic camera motion serves as a supervisory signal for camera motion estimation that is invariant to object and tool motion. We perform an extensive evaluation of state-of-the-art (SOTA) Deep Neural Networks (DNNs) across multiple compute regimes, finding our method transfers from our camera motion free da Vinci surgery dataset to videos of laparoscopic interventions, outperforming classical homography estimation approaches in both, precision by 41 % , and runtime on a CPU by 43 % .

Probabilistic landscape of seizure semiology localizing values

Semiology describes the evolution of symptoms and signs during epileptic seizures and contributes to the evaluation of individuals with focal drug-resistant epilepsy for curative resection. Semiology varies in complexity from elementary sensorimotor seizures arising from primary cortex to complex behaviours and automatisms emerging from distributed cerebral networks. Detailed semiology interpreted by expert epileptologists may point towards the likely site of seizure onset, but this process is subjective. No study has captured the variances in semiological localizing values in a data-driven manner to allow objective and probabilistic determinations of implicated networks and nodes. We curated an open data set from the epilepsy literature, in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, linking semiology to hierarchical brain localizations. A total of 11 230 data points were collected from 4643 patients across 309 articles, labelled using ground truths (postoperative seizure-freedom, concordance of imaging and neurophysiology, and/or invasive EEG) and a designation method that distinguished between semiologies arising from a predefined cortical region and descriptions of neuroanatomical localizations responsible for generating a particular semiology. This allowed us to mitigate temporal lobe publication bias by filtering studies that preselected patients based on prior knowledge of their seizure foci. Using this data set, we describe the probabilistic landscape of semiological localizing values as forest plots at the resolution of seven major brain regions: temporal, frontal, cingulate, parietal, occipital, insula, and hypothalamus, and five temporal subregions. We evaluated the intrinsic value of any one semiology over all other ictal manifestations. For example, epigastric auras implicated the temporal lobe with 83% probability when not accounting for the publication bias that favoured temporal lobe epilepsies. Unbiased results for a prior distribution of cortical localizations revised the prevalence of temporal lobe epilepsies from 66% to 44%. Therefore, knowledge about the presence of epigastric auras updates localization to the temporal lobe with an odds ratio (OR) of 2.4 [CI_95% (1.9, 2.9); and specifically, mesial temporal structures OR: 2.8 (2.3, 2.9)], attesting the value of epigastric auras. As a further example, although head version is thought to implicate the frontal lobes, it did not add localizing value compared with the prior distribution of cortical localizations [OR: 0.9 (0.7, 1.2)]. Objectification of the localizing values of the 12 most common semiologies provides a complementary view of brain dysfunction to that of lesion-deficit mappings, as instead of linking brain regions to phenotypic-deficits, semiological phenotypes are linked back to brain sources. This work enables coupling of seizure propagation with ictal manifestations, and clinical support algorithms for localizing seizure phenotypes.

AutoProstate: Towards Automated Reporting of Prostate MRI for Prostate Cancer Assessment Using Deep Learning

Simple Summary

International guidelines recommend multiparametric magnetic resonance imaging (mpMRI) of the prostate for use by radiologists to identify lesions containing clinically significant prostate cancer, prior to confirmatory biopsy. Automatic assessment of prostate mpMRI using artificial intelligence algorithms holds a currently unrealized potential to improve the diagnostic accuracy achievable by radiologists alone, improve the reporting consistency between radiologists, and enhance reporting quality. In this work, we introduce AutoProstate: a deep learning-powered framework for automatic MRI-based prostate cancer assessment. In particular, AutoProstate utilizes patient data and biparametric MRI to populate an automatic web-based report which includes segmentations of the whole prostate, prostatic zones, and candidate clinically significant prostate cancer lesions, and in addition, several derived characteristics with clinical value are presented. Notably, AutoProstate performed well in external validation using the PICTURE study dataset, suggesting value in prospective multicentre validation, with a view towards future deployment into the prostate cancer diagnostic pathway.

Abstract

Multiparametric magnetic resonance imaging (mpMRI) of the prostate is used by radiologists to identify, score, and stage abnormalities that may correspond to clinically significant prostate cancer (CSPCa). Automatic assessment of prostate mpMRI using artificial intelligence algorithms may facilitate a reduction in missed cancers and unnecessary biopsies, an increase in inter-observer agreement between radiologists, and an improvement in reporting quality. In this work, we introduce AutoProstate, a deep learning-powered framework for automatic MRI-based prostate cancer assessment. AutoProstate comprises of three modules: Zone-Segmenter, CSPCa-Segmenter, and Report-Generator. Zone-Segmenter segments the prostatic zones on T2-weighted imaging, CSPCa-Segmenter detects and segments CSPCa lesions using biparametric MRI, and Report-Generator generates an automatic web-based report containing four sections: Patient Details, Prostate Size and PSA Density, Clinically Significant Lesion Candidates, and Findings Summary. In our experiment, AutoProstate was trained using the publicly available PROSTATEx dataset, and externally validated using the PICTURE dataset. Moreover, the performance of AutoProstate was compared to the performance of an experienced radiologist who prospectively read PICTURE dataset cases. In comparison to the radiologist, AutoProstate showed statistically significant improvements in prostate volume and prostate-specific antigen density estimation. Furthermore, AutoProstate matched the CSPCa lesion detection sensitivity of the radiologist, which is paramount, but produced more false positive detections.

Deep learning approach for hyperspectral image demosaicking, spectral correction and high-resolution RGB reconstruction

Hyperspectral imaging is one of the most promising techniques for intraoperative tissue characterisation. Snapshot mosaic cameras, which can capture hyperspectral data in a single exposure, have the potential to make a real-time hyperspectral imaging system for surgical decision-making possible. However, optimal exploitation of the captured data requires solving an ill-posed demosaicking problem and applying additional spectral corrections. In this work, we propose a supervised learning-based image demosaicking algorithm for snapshot hyperspectral images. Due to the lack of publicly available medical images acquired with snapshot mosaic cameras, a synthetic image generation approach is proposed to simulate snapshot images from existing medical image datasets captured by high-resolution, but slow, hyperspectral imaging devices. Image reconstruction is achieved using convolutional neural networks for hyperspectral image super-resolution, followed by spectral correction using a sensor-specific calibration matrix. The results are evaluated both quantitatively and qualitatively, showing clear improvements in image quality compared to a baseline demosaicking method using linear interpolation. Moreover, the fast processing time of 45 ms of our algorithm to obtain super-resolved RGB or oxygenation saturation maps per image for a state-of-the-art snapshot mosaic camera demonstrates the potential for its seamless integration into real-time surgical hyperspectral imaging applications.

Segmentation of vestibular schwannoma from MRI, an open annotated dataset and baseline algorithm

Automatic segmentation of vestibular schwannomas (VS) from magnetic resonance imaging (MRI) could significantly improve clinical workflow and assist patient management. We have previously developed a novel artificial intelligence framework based on a 2.5D convolutional neural network achieving excellent results equivalent to those achieved by an independent human annotator. Here, we provide the first publicly-available annotated imaging dataset of VS by releasing the data and annotations used in our prior work. This collection contains a labelled dataset of 484 MR images collected on 242 consecutive patients with a VS undergoing Gamma Knife Stereotactic Radiosurgery at a single institution. Data includes all segmentations and contours used in treatment planning and details of the administered dose. Implementation of our automated segmentation algorithm uses MONAI, a freely-available open-source framework for deep learning in healthcare imaging. These data will facilitate the development and validation of automated segmentation frameworks for VS and may also be used to develop other multi-modal algorithmic models.

OS08.6.A Glioblastoma treatment response machine learning monitoring biomarkers: a systematic review and meta-analysis

BACKGROUND

The aim of the systematic review was to assess recently published studies on diagnostic test accuracy of glioblastoma treatment response monitoring biomarkers in adults, developed through machine learning (ML).

MATERIAL AND METHODS

PRISMA methodology was followed. Articles published 09/2018-01/2021 (since previous reviews) were searched for using MEDLINE, EMBASE, and the Cochrane Register by two reviewers independently. Included study participants were adult patients with high grade glioma who had undergone standard treatment (maximal resection, radiotherapy with concomitant and adjuvant temozolomide) and subsequently underwent follow-up imaging to determine treatment response status (specifically, distinguishing progression/recurrence from progression/recurrence mimics - the target condition). Risk of bias and applicability was assessed with QUADAS 2. A third reviewer arbitrated any discrepancy. Contingency tables were created for hold-out test sets and recall, specificity, precision, F1-score, balanced accuracy calculated. A meta-analysis was performed using a bivariate model for recall, false positive rate and area-under the receiver operator characteristic curve (AUC).

RESULTS

Eighteen studies were included with 1335 patients in training sets and 384 in test sets. To determine whether there was progression or a mimic, the reference standard combination of follow-up imaging and histopathology at re-operation was applied in 67% (13/18) of studies. The small numbers of patient included in studies, the high risk of bias and concerns of applicability in the study designs (particularly in relation to the reference standard and patient selection due to confounding), and the low level of evidence, suggest that limited conclusions can be drawn from the data. Ten studies (10/18, 56%) had internal or external hold-out test set data that could be included in a meta-analysis of monitoring biomarker studies. The pooled sensitivity was 0.77 (0.65–0.86). The pooled false positive rate (1-specificity) was 0.35 (0.25–0.47). The summary point estimate for the AUC was 0.77.

CONCLUSION

There is likely good diagnostic performance of machine learning models that use MRI features to distinguish between progression and mimics. The diagnostic performance of ML using implicit features did not appear to be superior to ML using explicit features. There are a range of ML-based solutions poised to become treatment response monitoring biomarkers for glioblastoma. To achieve this, the development and validation of ML models require large, well-annotated datasets where the potential for confounding in the study design has been carefully considered. Therefore, multidisciplinary efforts and multicentre collaborations are necessary.

TorchIO: A Python library for efficient loading, preprocessing, augmentation and patch-based sampling of medical images in deep learning

BACKGROUND AND OBJECTIVE

Processing of medical images such as MRI or CT presents different challenges compared to RGB images typically used in computer vision. These include a lack of labels for large datasets, high computational costs, and the need of metadata to describe the physical properties of voxels. Data augmentation is used to artificially increase the size of the training datasets. Training with image subvolumes or patches decreases the need for computational power. Spatial metadata needs to be carefully taken into account in order to ensure a correct alignment and orientation of volumes.

METHODS

We present TorchIO, an open-source Python library to enable efficient loading, preprocessing, augmentation and patch-based sampling of medical images for deep learning. TorchIO follows the style of PyTorch and integrates standard medical image processing libraries to efficiently process images during training of neural networks. TorchIO transforms can be easily composed, reproduced, traced and extended. Most transforms can be inverted, making the library suitable for test-time augmentation and estimation of aleatoric uncertainty in the context of segmentation. We provide multiple generic preprocessing and augmentation operations as well as simulation of MRI-specific artifacts.

RESULTS

Source code, comprehensive tutorials and extensive documentation for TorchIO can be found at http://torchio.rtfd.io/. The package can be installed from the Python Package Index (PyPI) running pip install torchio. It includes a command-line interface which allows users to apply transforms to image files without using Python. Additionally, we provide a graphical user interface within a TorchIO extension in 3D Slicer to visualize the effects of transforms.

CONCLUSION

TorchIO was developed to help researchers standardize medical image processing pipelines and allow them to focus on the deep learning experiments. It encourages good open-science practices, as it supports experiment reproducibility and is version-controlled so that the software can be cited precisely. Due to its modularity, the library is compatible with other frameworks for deep learning with medical images.

TorchIO: A Python library for efficient loading, preprocessing, augmentation and patch-based sampling of medical images in deep learning

BACKGROUND AND OBJECTIVE

Processing of medical images such as MRI or CT presents different challenges compared to RGB images typically used in computer vision. These include a lack of labels for large datasets, high computational costs, and the need of metadata to describe the physical properties of voxels. Data augmentation is used to artificially increase the size of the training datasets. Training with image subvolumes or patches decreases the need for computational power. Spatial metadata needs to be carefully taken into account in order to ensure a correct alignment and orientation of volumes.

METHODS

We present TorchIO, an open-source Python library to enable efficient loading, preprocessing, augmentation and patch-based sampling of medical images for deep learning. TorchIO follows the style of PyTorch and integrates standard medical image processing libraries to efficiently process images during training of neural networks. TorchIO transforms can be easily composed, reproduced, traced and extended. Most transforms can be inverted, making the library suitable for test-time augmentation and estimation of aleatoric uncertainty in the context of segmentation. We provide multiple generic preprocessing and augmentation operations as well as simulation of MRI-specific artifacts.

RESULTS

Source code, comprehensive tutorials and extensive documentation for TorchIO can be found at http://torchio.rtfd.io/. The package can be installed from the Python Package Index (PyPI) running pip install torchio. It includes a command-line interface which allows users to apply transforms to image files without using Python. Additionally, we provide a graphical user interface within a TorchIO extension in 3D Slicer to visualize the effects of transforms.

CONCLUSION

TorchIO was developed to help researchers standardize medical image processing pipelines and allow them to focus on the deep learning experiments. It encourages good open-science practices, as it supports experiment reproducibility and is version-controlled so that the software can be cited precisely. Due to its modularity, the library is compatible with other frameworks for deep learning with medical images.

Intraoperative hyperspectral label-free imaging: from system design to first-in-patient translation

Despite advances in intraoperative surgical imaging, reliable discrimination of critical tissue during surgery remains challenging. As a result, decisions with potentially life-changing consequences for patients are still based on the surgeon's subjective visual assessment. Hyperspectral imaging (HSI) provides a promising solution for objective intraoperative tissue characterisation, with the advantages of being non-contact, non-ionising and non-invasive. However, while its potential to aid surgical decision-making has been investigated for a range of applications, to date no real-time intraoperative HSI (iHSI) system has been presented that follows critical design considerations to ensure a satisfactory integration into the surgical workflow. By establishing functional and technical requirements of an intraoperative system for surgery, we present an iHSI system design that allows for real-time wide-field HSI and responsive surgical guidance in a highly constrained operating theatre. Two systems exploiting state-of-the-art industrial HSI cameras, respectively using linescan and snapshot imaging technology, were designed and investigated by performing assessments against established design criteria and ex vivo tissue experiments. Finally, we report the use of our real-time iHSI system in a clinical feasibility case study as part of a spinal fusion surgery. Our results demonstrate seamless integration into existing surgical workflows.

Imitation learning for improved 3D PET/MR attenuation correction

The assessment of the quality of synthesised/pseudo Computed Tomography (pCT) images is commonly measured by an intensity-wise similarity between the ground truth CT and the pCT. However, when using the pCT as an attenuation map (μ-map) for PET reconstruction in Positron Emission Tomography Magnetic Resonance Imaging (PET/MRI) minimising the error between pCT and CT neglects the main objective of predicting a pCT that when used as μ-map reconstructs a pseudo PET (pPET) which is as similar as possible to the gold standard CT-derived PET reconstruction. This observation motivated us to propose a novel multi-hypothesis deep learning framework explicitly aimed at PET reconstruction application. A convolutional neural network (CNN) synthesises pCTs by minimising a combination of the pixel-wise error between pCT and CT and a novel metric-loss that itself is defined by a CNN and aims to minimise consequent PET residuals. Training is performed on a database of twenty 3D MR/CT/PET brain image pairs. Quantitative results on a fully independent dataset of twenty-three 3D MR/CT/PET image pairs show that the network is able to synthesise more accurate pCTs. The Mean Absolute Error on the pCT (110.98 HU ± 19.22 HU) compared to a baseline CNN (172.12 HU ± 19.61 HU) and a multi-atlas propagation approach (153.40 HU ± 18.68 HU), and subsequently lead to a significant improvement in the PET reconstruction error (4.74% ± 1.52% compared to baseline 13.72% ± 2.48% and multi-atlas propagation 6.68% ± 2.06%).

Artificial Intelligence Compared to Radiologists for the Initial Diagnosis of Prostate Cancer on Magnetic Resonance Imaging: A Systematic Review and Recommendations for Future Studies

Computer-aided diagnosis (CAD) of prostate cancer on multiparametric magnetic resonance imaging (mpMRI), using artificial intelligence (AI), may reduce missed cancers and unnecessary biopsies, increase inter-observer agreement between radiologists, and alleviate pressures caused by rising case incidence and a shortage of specialist radiologists to read prostate mpMRI. However, well-designed evaluation studies are required to prove efficacy above current clinical practice. A systematic search of the MEDLINE, EMBASE, and arXiv electronic databases was conducted for studies that compared CAD for prostate cancer detection or classification on MRI against radiologist interpretation and a histopathological reference standard, in treatment-naïve men with a clinical suspicion of prostate cancer. Twenty-seven studies were included in the final analysis. Due to substantial heterogeneities in the included studies, a narrative synthesis is presented. Several studies reported superior diagnostic accuracy for CAD over radiologist interpretation on small, internal patient datasets, though this was not observed in the few studies that performed evaluation using external patient data. Our review found insufficient evidence to suggest the clinical deployment of artificial intelligence algorithms at present. Further work is needed to develop and enforce methodological standards, promote access to large diverse datasets, and conduct prospective evaluations before clinical adoption can be considered.

Computer-aided diagnosis of prostate cancer using multiparametric MRI and clinical features: A patient-level classification framework

Computer-aided diagnosis (CAD) of prostate cancer (PCa) using multiparametric magnetic resonance imaging (mpMRI) is actively being investigated as a means to provide clinical decision support to radiologists. Typically, these systems are trained using lesion annotations. However, lesion annotations are expensive to obtain and inadequate for characterizing certain tumor types e.g. diffuse tumors and MRI invisible tumors. In this work, we introduce a novel patient-level classification framework, denoted PCF, that is trained using patient-level labels only. In PCF, features are extracted from three-dimensional mpMRI and derived parameter maps using convolutional neural networks and subsequently, combined with clinical features by a multi-classifier support vector machine scheme. The output of PCF is a probability value that indicates whether a patient is harboring clinically significant PCa (Gleason score ≥3+4) or not. PCF achieved mean area under the receiver operating characteristic curves of 0.79 and 0.86 on the PICTURE and PROSTATEx datasets respectively, using five-fold cross-validation. Clinical evaluation over a temporally separated PICTURE dataset cohort demonstrated comparable sensitivity and specificity to an experienced radiologist. We envision PCF finding most utility as a second reader during routine diagnosis or as a triage tool to identify low-risk patients who do not require a clinical read.

A self-supervised learning strategy for postoperative brain cavity segmentation simulating resections

PURPOSE

Accurate segmentation of brain resection cavities (RCs) aids in postoperative analysis and determining follow-up treatment. Convolutional neural networks (CNNs) are the state-of-the-art image segmentation technique, but require large annotated datasets for training. Annotation of 3D medical images is time-consuming, requires highly trained raters and may suffer from high inter-rater variability. Self-supervised learning strategies can leverage unlabeled data for training.

METHODS

We developed an algorithm to simulate resections from preoperative magnetic resonance images (MRIs). We performed self-supervised training of a 3D CNN for RC segmentation using our simulation method. We curated EPISURG, a dataset comprising 430 postoperative and 268 preoperative MRIs from 430 refractory epilepsy patients who underwent resective neurosurgery. We fine-tuned our model on three small annotated datasets from different institutions and on the annotated images in EPISURG, comprising 20, 33, 19 and 133 subjects.

RESULTS

The model trained on data with simulated resections obtained median (interquartile range) Dice score coefficients (DSCs) of 81.7 (16.4), 82.4 (36.4), 74.9 (24.2) and 80.5 (18.7) for each of the four datasets. After fine-tuning, DSCs were 89.2 (13.3), 84.1 (19.8), 80.2 (20.1) and 85.2 (10.8). For comparison, inter-rater agreement between human annotators from our previous study was 84.0 (9.9).

CONCLUSION

We present a self-supervised learning strategy for 3D CNNs using simulated RCs to accurately segment real RCs on postoperative MRI. Our method generalizes well to data from different institutions, pathologies and modalities. Source code, segmentation models and the EPISURG dataset are available at https://github.com/fepegar/resseg-ijcars .

22 Volumetric MRI; A Potential Outcome Measure of Muscle Reinnervation

Introduction

Improved outcome measures of muscle reinnervation would facilitate clinical translation of new therapies which hope to enhance human peripheral nerve repair. Valid outcome measures should be receptive to the biological process of muscle reinnervation and correlate with clinical assessments of muscular function. This study investigated the responsiveness of volumetric MRI to the biological process of muscle reinnervation and its relationship with clinical indices of muscular function.

Method

Twenty-five patients who underwent nerve transfer to reinnervate elbow flexor muscles were followed-up at a median time of 258 days (-86 to 1698 days) post-operatively for a mean of two (one to three) volumetric MRI assessments. Medical Research Council (MRC) grade, peak volitional force (PVF), muscular fatigue, co-contraction and Stanmore Percentage of Normal Elbow Assessment (SPONEA) was also measured at each appointment. The responsiveness of each parameter was compared using Pearson or Spearman correlation as appropriate.

Results

Elbow flexor muscle volume per unit BMI demonstrated responsiveness to the biological process of muscle reinnervation (R2=0.73, p &lt; 0.001) and correlated with patient reported impairments of reinnervated muscle; co-contraction (R2=0.63, p = 0.02) and muscle fatigue (R2=0.64, p = 0.04).

Conclusions

Volumetric MRI may is an excellent candidate as an outcome measure of muscle reinnervation.

Integrated multi-modality image-guided navigation for neurosurgery: open-source software platform using state-of-the-art clinical hardware

PURPOSE

Image-guided surgery (IGS) is an integral part of modern neuro-oncology surgery. Navigated ultrasound provides the surgeon with reconstructed views of ultrasound data, but no commercial system presently permits its integration with other essential non-imaging-based intraoperative monitoring modalities such as intraoperative neuromonitoring. Such a system would be particularly useful in skull base neurosurgery.

METHODS

We established functional and technical requirements of an integrated multi-modality IGS system tailored for skull base surgery with the ability to incorporate: (1) preoperative MRI data and associated 3D volume reconstructions, (2) real-time intraoperative neurophysiological data and (3) live reconstructed 3D ultrasound. We created an open-source software platform to integrate with readily available commercial hardware. We tested the accuracy of the system's ultrasound navigation and reconstruction using a polyvinyl alcohol phantom model and simulated the use of the complete navigation system in a clinical operating room using a patient-specific phantom model.

RESULTS

Experimental validation of the system's navigated ultrasound component demonstrated accuracy of [Formula: see text] and a frame rate of 25 frames per second. Clinical simulation confirmed that system assembly was straightforward, could be achieved in a clinically acceptable time of [Formula: see text] and performed with a clinically acceptable level of accuracy.

CONCLUSION

We present an integrated open-source research platform for multi-modality IGS. The present prototype system was tailored for neurosurgery and met all minimum design requirements focused on skull base surgery. Future work aims to optimise the system further by addressing the remaining target requirements.

144 Probable Delirium is A Presenting Symptom of COVID-19 in Frail, Older Adults: A Study of Hospitalised and Community-Based Cohorts

Introduction

COVID-19 exhibits a more severe disease course in older adults with frailty. Awareness of atypical presentations is critical to facilitate early disease identification. This study aimed to assess how frailty affects presenting symptoms of COVID-19 in older adults.

Methods

Observational study of two distinct cohorts: (i) Hospitalised patients aged 65 and over; unscheduled admission to a large London teaching hospital between March 1st, 2020-May 5th, 2020; COVID-19 confirmed by RT-PCR of nasopharyngeal swab (n = 322); (ii) Community-based adults aged 65 and over enrolled in the COVID Symptom Study mobile application between March 24th (application launch)-May 8th, 2020; self-report or report-by-proxy data; reported test-positive for COVID-19 (n = 535). Multivariable logistic regression analysis performed on age-matched samples of both cohorts to determine associations between frailty and symptoms of COVID-19 including delirium, fever and cough.

Results

Hospital cohort: there was a significantly higher prevalence of delirium amongst the frail sample, with no difference in fever or cough. Of those presenting with delirium, 10/53 (18.9%) presented with delirium as the only documented symptom. Community-based cohort: there was a significantly higher prevalence of probable delirium in the frail sample, and also of fatigue and shortness of breath. Of those reporting probable delirium, 28/84 (33%) did not report fever or cough.

Conclusions

This study demonstrates a higher prevalence of delirium as a presenting symptom of COVID-19 infection in older adults with frailty compared to their age-matched non-frail counterparts. Clinicians should suspect COVID-19 in frail older adults presenting with delirium. Early detection facilitates infection control measures to mitigate against catastrophic spread and preventable hospitalisations and deaths amongst this population. Our findings emphasise the need for systematic frailty assessment for all acutely ill older patients in both hospital and community settings, as well as systematic evaluation of any change in mental status.

Cranial findings detected by second-trimester ultrasound in fetuses with myelomeningocele: a systematic review

Background

Abnormal intracranial findings are often detected at mid‐trimester ultrasound (US) in fetuses with myelomeningocele (MMC). It is unclear whether these findings constitute a spectrum of the disease or are an independent finding, which should contraindicate fetal surgery.

Objective

To ascertain the spectrum and frequency of US‐detected cranial findings in fetuses with MMC.

Search strategy

MEDLINE, Embase, Web of Science and CENTRAL were searched from January 2000 to June 2020.

Selection criteria

Study reporting incidence of cranial US findings in consecutive cases of second‐trimester fetuses with MMC.

Data collection and analysis

Publication quality was assessed by Newcastle–Ottawa Scale (NOS) and modified NOS. Meta‐analysis could not be performed as a result of high clinical diversity and study heterogeneity.

Main results

Fourteen cranial US findings were reported in 15 studies. Findings in classic Chiari II malformation (CIIM) spectrum included posterior fossa funnelling (96%), small transcerebellar diameter (82–96%), ‘banana’ sign (50–100%), beaked tectum (65%) and ‘lemon’ sign (53–100%). Additional cranial findings were small biparietal diameter (BPD) and head circumference (HC) (<5th centile; 53 and 71%, respectively), ventriculomegaly (45–89%), abnormal pointed shape of the occipital horn (77–78%), thinning of the posterior cerebrum, perinodular heterotopia (11%), abnormal gyration (3%), corpus callosum disorders (60%) and midline interhemispheric cyst (42%).

Conclusions

We identified 14 cranial findings by second‐trimester US in fetuses with MMC. The relatively high incidence of these findings and their unclear prognostic significance might not contraindicate fetal surgery in the case of normal fetal genetic testing. Some cranial findings may independently affect postnatal outcome, however. Long‐term detailed follow‐up is required to investigate this.

Tweetable abstract

A high rate of cranial abnormalities found on second‐trimester ultrasound in fetuses with myelomeningocele.

A high rate of cranial abnormalities found on second‐trimester ultrasound in fetuses with myelomeningocele.

Diagnostic value of cutaneous manifestation of SARS-CoV-2 infection

Background

One of the challenging aspects of SARS‐CoV‐2 infection is its diverse multisystemic disease presentation.

Objectives

To evaluate the diagnostic value of cutaneous manifestations of SARS‐CoV‐2 infection and investigate their duration and timing in relation to other COVID‐19 symptoms.

Methods

We used data from 336 847 UK users of the COVID Symptom Study app to assess the diagnostic value of body rash or an acral rash in SARS‐CoV‐2 infection, and data from an independent online survey of 11 544 respondents to investigate skin‐specific symptoms and collect their photographs.

Results

Using data from the app, we show significant association between skin rashes and a positive swab test result (odds ratio 1·67, 95% confidence interval 1·42–1·97). Strikingly, among the respondents of the independent online survey, we found that 17% of SARS‐CoV‐2‐positive cases reported skin rashes as the first presentation, and 21% as the only clinical sign of COVID‐19. Together with the British Association of Dermatologists, we have compiled a catalogue of images of the most common skin manifestations of COVID‐19 from 400 individuals (https://covidskinsigns.com), which we have made publicly available to assist clinicians in recognition of this early clinical feature of COVID‐19.

Conclusions

Skin rashes cluster with other COVID‐19 symptoms, are predictive of a positive swab test, and occur in a significant number of cases, either alone or before other classical symptoms. Recognizing rashes is important in identifying new and earlier cases of COVID‐19.

What is already known about this topic?

Several studies conducted in hospital settings reported that patients with COVID‐19 presented with unusual skin rashes, including urticarial rashes, vesicular lesions and, less frequently, chilblains in fingers or toes.

What does this study add?

We confirmed, in a community‐based setting that also includes milder forms of the disease, that the presence of a skin rash is predictive of SARS‐CoV‐2 infection.

We provide a website with photos of skin manifestations to help healthcare professionals in diagnosing COVID‐19.

Skin rashes should be taken into account to provide a quick COVID‐19 diagnosis to curb the spread of the disease.

Linked Comment: Naldi. Br J Dermatol 2021; 184:793–794.

Anomaly detection for the individual analysis of brain PET images

Purpose: In clinical practice, positron emission tomography (PET) images are mostly analyzed visually, but the sensitivity and specificity of this approach greatly depend on the observer's experience. Quantitative analysis of PET images would alleviate this problem by helping define an objective limit between normal and pathological findings. We present an anomaly detection framework for the individual analysis of PET images. Approach: We created subject-specific abnormality maps that summarize the pathology's topographical distribution in the brain by comparing the subject's PET image to a model of healthy PET appearance that is specific to the subject under investigation. This model was generated from demographically and morphologically matched PET scans from a control dataset. Results: We generated abnormality maps for healthy controls, patients at different stages of Alzheimer's disease and with different frontotemporal dementia syndromes. We showed that no anomalies were detected for the healthy controls and that the anomalies detected from the patients with dementia coincided with the regions where abnormal uptake was expected. We also validated the proposed framework using the abnormality maps as inputs of a classifier and obtained higher classification accuracies than when using the PET images themselves as inputs. Conclusions: The proposed method was able to automatically locate and characterize the areas characteristic of dementia from PET images. The abnormality maps are expected to (i) help clinicians in their diagnosis by highlighting, in a data-driven fashion, the pathological areas, and (ii) improve the interpretability of subsequent analyses, such as computer-aided diagnosis or spatiotemporal modeling.

Symptom clusters in COVID-19: A potential clinical prediction tool from the COVID Symptom Study app

As no one symptom can predict disease severity or the need for dedicated medical support in coronavirus disease 2019 (COVID-19), we asked whether documenting symptom time series over the first few days informs outcome. Unsupervised time series clustering over symptom presentation was performed on data collected from a training dataset of completed cases enlisted early from the COVID Symptom Study Smartphone application, yielding six distinct symptom presentations. Clustering was validated on an independent replication dataset between 1 and 28 May 2020. Using the first 5 days of symptom logging, the ROC-AUC (receiver operating characteristic - area under the curve) of need for respiratory support was 78.8%, substantially outperforming personal characteristics alone (ROC-AUC 69.5%). Such an approach could be used to monitor at-risk patients and predict medical resource requirements days before they are required.

Optimal symptom combinations to aid COVID-19 case identification: Analysis from a community-based, prospective, observational cohort

Objectives

Diagnostic work-up following any COVID-19 associated symptom will lead to extensive testing, potentially overwhelming laboratory capacity whilst primarily yielding negative results. We aimed to identify optimal symptom combinations to capture most cases using fewer tests with implications for COVID-19 vaccine developers across different resource settings and public health.

Methods

UK and US users of the COVID-19 Symptom Study app who reported new-onset symptoms and an RT-PCR test within seven days of symptom onset were included. Sensitivity, specificity, and number of RT-PCR tests needed to identify one case (test per case [TPC]) were calculated for different symptom combinations. A multi-objective evolutionary algorithm was applied to generate combinations with optimal trade-offs between sensitivity and specificity.

Findings

UK and US cohorts included 122,305 (1,202 positives) and 3,162 (79 positive) individuals. Within three days of symptom onset, the COVID-19 specific symptom combination (cough, dyspnoea, fever, anosmia/ageusia) identified 69% of cases requiring 47 TPC. The combination with highest sensitivity (fatigue, anosmia/ageusia, cough, diarrhoea, headache, sore throat) identified 96% cases requiring 96 TPC.

Interpretation

We confirmed the significance of COVID-19 specific symptoms for triggering RT-PCR and identified additional symptom combinations with optimal trade-offs between sensitivity and specificity that maximize case capture given different resource settings.

Optimal symptom combinations to aid COVID-19 case identification: analysis from a community-based, prospective, observational cohort

OBJECTIVES

Diagnostic work-up following any COVID-19 associated symptom will lead to extensive testing, potentially overwhelming laboratory capacity whilst primarily yielding negative results. We aimed to identify optimal symptom combinations to capture most cases using fewer tests with implications for COVID-19 vaccine developers across different resource settings and public health.

METHODS

UK and US users of the COVID-19 Symptom Study app who reported new-onset symptoms and an RT-PCR test within seven days of symptom onset were included. Sensitivity, specificity, and number of RT-PCR tests needed to identify one case (test per case [TPC]) were calculated for different symptom combinations. A multi-objective evolutionary algorithm was applied to generate combinations with optimal trade-offs between sensitivity and specificity.

FINDINGS

UK and US cohorts included 122,305 (1,202 positives) and 3,162 (79 positive) individuals. Within three days of symptom onset, the COVID-19 specific symptom combination (cough, dyspnoea, fever, anosmia/ageusia) identified 69% of cases requiring 47 TPC. The combination with highest sensitivity (fatigue, anosmia/ageusia, cough, diarrhoea, headache, sore throat) identified 96% cases requiring 96 TPC.

INTERPRETATION

We confirmed the significance of COVID-19 specific symptoms for triggering RT-PCR and identified additional symptom combinations with optimal trade-offs between sensitivity and specificity that maximize case capture given different resource settings.

CA-Net: Comprehensive Attention Convolutional Neural Networks for Explainable Medical Image Segmentation

Accurate medical image segmentation is essential for diagnosis and treatment planning of diseases. Convolutional Neural Networks (CNNs) have achieved state-of-the-art performance for automatic medical image segmentation. However, they are still challenged by complicated conditions where the segmentation target has large variations of position, shape and scale, and existing CNNs have a poor explainability that limits their application to clinical decisions. In this work, we make extensive use of multiple attentions in a CNN architecture and propose a comprehensive attention-based CNN (CA-Net) for more accurate and explainable medical image segmentation that is aware of the most important spatial positions, channels and scales at the same time. In particular, we first propose a joint spatial attention module to make the network focus more on the foreground region. Then, a novel channel attention module is proposed to adaptively recalibrate channel-wise feature responses and highlight the most relevant feature channels. Also, we propose a scale attention module implicitly emphasizing the most salient feature maps among multiple scales so that the CNN is adaptive to the size of an object. Extensive experiments on skin lesion segmentation from ISIC 2018 and multi-class segmentation of fetal MRI found that our proposed CA-Net significantly improved the average segmentation Dice score from 87.77% to 92.08% for skin lesion, 84.79% to 87.08% for the placenta and 93.20% to 95.88% for the fetal brain respectively compared with U-Net. It reduced the model size to around 15 times smaller with close or even better accuracy compared with state-of-the-art DeepLabv3+. In addition, it has a much higher explainability than existing networks by visualizing the attention weight maps. Our code is available at https://github.com/HiLab-git/CA-Net.

A multi-channel uncertainty-aware multi-resolution network for MR to CT synthesis

Synthesising computed tomography (CT) images from magnetic resonance images (MRI) plays an important role in the field of medical image analysis, both for quantification and diagnostic purposes. Convolutional neural networks (CNNs) have achieved state-of-the-art results in image-to-image translation for brain applications. However, synthesising whole-body images remains largely uncharted territory involving many challenges, including large image size and limited field of view, complex spatial context, and anatomical differences between images acquired at different times. We propose the use of an uncertainty-aware multi-channel multi-resolution 3D cascade network specifically aiming for whole-body MR to CT synthesis. The Mean Absolute Error on the synthetic CT generated with the MultiRes unc network (73.90 HU) is compared to multiple baseline CNNs like 3D U-Net (92.89 HU), HighRes3DNet (89.05 HU) and deep boosted regression (77.58 HU) and shows superior synthesis performance. We ultimately exploit the extrapolation properties of the MultiRes networks on sub-regions of the body.

Robust parametric modeling of Alzheimer's disease progression

Quantitative characterization of disease progression using longitudinal data can provide long-term predictions for the pathological stages of individuals. This work studies the robust modeling of Alzheimer's disease progression using parametric methods. The proposed method linearly maps the individual's age to a disease progression score (DPS) and jointly fits constrained generalized logistic functions to the longitudinal dynamics of biomarkers as functions of the DPS using M-estimation. Robustness of the estimates is quantified using bootstrapping via Monte Carlo resampling, and the estimated inflection points of the fitted functions are used to temporally order the modeled biomarkers in the disease course. Kernel density estimation is applied to the obtained DPSs for clinical status classification using a Bayesian classifier. Different M-estimators and logistic functions, including a novel type proposed in this study, called modified Stannard, are evaluated on the data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) for robust modeling of volumetric magnetic resonance imaging (MRI) and positron emission tomography (PET) biomarkers, cerebrospinal fluid (CSF) measurements, as well as cognitive tests. The results show that the modified Stannard function fitted using the logistic loss achieves the best modeling performance with an average normalized mean absolute error (NMAE) of 0.991 across all biomarkers and bootstraps. Applied to the ADNI test set, this model achieves a multiclass area under the ROC curve (AUC) of 0.934 in clinical status classification. The obtained results for the proposed model outperform almost all state-of-the-art results in predicting biomarker values and classifying clinical status. Finally, the experiments show that the proposed model, trained using abundant ADNI data, generalizes well to data from the National Alzheimer's Coordinating Center (NACC) with an average NMAE of 1.182 and a multiclass AUC of 0.929.

A generative model of hyperelastic strain energy density functions for multiple tissue brain deformation

PURPOSE

Estimation of brain deformation is crucial during neurosurgery. Whilst mechanical characterisation captures stress-strain relationships of tissue, biomechanical models are limited by experimental conditions. This results in variability reported in the literature. The aim of this work was to demonstrate a generative model of strain energy density functions can estimate the elastic properties of tissue using observed brain deformation.

METHODS

For the generative model a Gaussian Process regression learns elastic potentials from 73 manuscripts. We evaluate the use of neo-Hookean, Mooney-Rivlin and 1-term Ogden meta-models to guarantee stability. Single and multiple tissue experiments validate the ability of our generative model to estimate tissue properties on a synthetic brain model and in eight temporal lobe resection cases where deformation is observed between pre- and post-operative images.

RESULTS

Estimated parameters on a synthetic model are close to the known reference with a root-mean-square error (RMSE) of 0.1 mm and 0.2 mm between surface nodes for single and multiple tissue experiments. In clinical cases, we were able to recover brain deformation from pre- to post-operative images reducing RMSE of differences from 1.37 to 1.08 mm on the ventricle surface and from 5.89 to 4.84 mm on the resection cavity surface.

CONCLUSION

Our generative model can capture uncertainties related to mechanical characterisation of tissue. When fitting samples from elastography and linear studies, all meta-models performed similarly. The Ogden meta-model performed the best on hyperelastic studies. We were able to predict elastic parameters in a reference model on a synthetic phantom. However, deformation observed in clinical cases is only partly explained using our generative model.

The impact of unscheduled gaps and iso-centre sequencing on the biologically effective dose in Gamma Knife radiosurgery

PURPOSE

Establish the impact of iso-centre sequencing and unscheduled gaps in Gamma Knife® (GK) radiosurgery on the biologically effective dose (BED).

METHODS

A BED model was used to study BED values on the prescription iso-surface of patients treated with GK Perfexion™ (Vestibular Schwannoma). The effect of a 15 min gap, simulated at varying points in the treatment delivery, and adjustments to the sequencing of iso-centre delivery, based on average dose-rate, was quantified in terms of the impact on BED.

RESULTS

Depending on the position of the gap and the average dose-rate profiles, the mean BED values were decreased by 0.1% to 9.9% of the value in the original plan. A heuristic approach to iso-centre sequencing showed variations in BED of up to 14.2%, relative to the mean BED of the original sequence.

CONCLUSION

The treatment variables, like the iso-centre sequence and unscheduled gaps, should be considered during GK radiosurgery treatments.

The future of digital health with federated learning

Data-driven machine learning (ML) has emerged as a promising approach for building accurate and robust statistical models from medical data, which is collected in huge volumes by modern healthcare systems. Existing medical data is not fully exploited by ML primarily because it sits in data silos and privacy concerns restrict access to this data. However, without access to sufficient data, ML will be prevented from reaching its full potential and, ultimately, from making the transition from research to clinical practice. This paper considers key factors contributing to this issue, explores how federated learning (FL) may provide a solution for the future of digital health and highlights the challenges and considerations that need to be addressed.

Magnetic resonance imaging measurement of placental perfusion and oxygen saturation in early-onset fetal growth restriction

OBJECTIVE

We hypothesised that a multi-compartment magnetic resonance imaging (MRI) technique that is sensitive to fetal blood oxygenation would identify changes in placental blood volume and fetal blood oxygenation in pregnancies complicated by early-onset fetal growth restriction (FGR).

DESIGN

Case-control study.

SETTING

London, UK.

POPULATION

Women with uncomplicated pregnancies (estimated fetal weight [EFW] >10th centile for gestational age [GA] and normal maternal and fetal Doppler ultrasound, n = 12) or early-onset FGR (EFW <3rd centile with or without abnormal Doppler ultrasound <32 weeks GA, n = 12) were studied.

METHODS

All women underwent MRI examination. Using a multi-compartment MRI technique, we quantified fetal and maternal blood volume and feto-placental blood oxygenation.

MAIN OUTCOME MEASURES

Disease severity was stratified according to Doppler pulsatility index and the relationship to the MRI parameters was investigated, including the influence of GA at scan.

RESULTS

The FGR group (mean GA 27⁺⁵  weeks, range 24⁺² to 33⁺⁶  weeks) had a significantly lower EFW compared with the control group (mean GA 29⁺¹  weeks; -705 g, 95% CI -353 to -1057 g). MRI-derived feto-placental oxygen saturation was higher in controls compared with FGR (75 ± 9.6% versus 56 ± 16.2%, P = 0.02, 95% CI 7.8-30.3%). Feto-placental oxygen saturation estimation correlated strongly with GA at scan in controls (r = -0.83).

CONCLUSION

Using a novel multimodal MRI protocol we demonstrated reduced feto-placental blood oxygen saturation in pregnancies complicated by early-onset FGR. The degree of abnormality correlated with disease severity defined by ultrasound Doppler findings. Gestational age-dependent changes in oxygen saturation were also present in normal pregnancies.

TWEETABLE ABSTRACT

MRI reveals differences in feto-placental oxygen saturation between normal and FGR pregnancy that is associated with disease severity.

A multi-shell multi-tissue diffusion study of brain connectivity in early multiple sclerosis

BACKGROUND

The potential of multi-shell diffusion imaging to produce accurate brain connectivity metrics able to unravel key pathophysiological processes in multiple sclerosis (MS) has scarcely been investigated.

OBJECTIVE

To test, in patients with a clinically isolated syndrome (CIS), whether multi-shell imaging-derived connectivity metrics can differentiate patients from controls, correlate with clinical measures, and perform better than metrics obtained with conventional single-shell protocols.

METHODS

Nineteen patients within 3 months from the CIS and 12 healthy controls underwent anatomical and 53-direction multi-shell diffusion-weighted 3T images. Patients were cognitively assessed. Voxel-wise fibre orientation distribution functions were estimated and used to obtain network metrics. These were also calculated using a conventional single-shell diffusion protocol. Through linear regression, we obtained effect sizes and standardised regression coefficients.

RESULTS

Patients had lower mean nodal strength (p = 0.003) and greater network modularity than controls (p = 0.045). Greater modularity was associated with worse cognitive performance in patients, even after accounting for lesion load (p = 0.002). Multi-shell-derived metrics outperformed single-shell-derived ones.

CONCLUSION

Connectivity-based nodal strength and network modularity are abnormal in the CIS. Furthermore, the increased network modularity observed in patients, indicating microstructural damage, is clinically relevant. Connectivity analyses based on multi-shell imaging can detect potentially relevant network changes in early MS.

Automated computation and analysis of accuracy metrics in stereoencephalography

•

Automatic computation of SEEG accuracy metrics agree with those done manually.

•

The choice of image to generate a scalp model has an effect on entry point metrics.

•

Metrics have the lowest mean and variability when using an electrode bolt axis.

•

Lateral shift deviation should include a measure of insertion depth error.

Background

Implantation accuracy of electrodes during neurosurgical interventions is necessary to ensure safety and efficacy. Typically, metrics are computed by visual inspection which is tedious, prone to inter-/intra-observer variation, and difficult to replicate across sites.

New Method

We propose an automated approach for computing implantation metrics and investigate potential sources of error. We focus on accuracy metrics commonly reported in the literature to validate our approach against metrics computed manually including entry point (EP) and target point (TP) localisation errors and angle differences between planned and implanted trajectories in 15 patients with a total of 158 stereoelectroencephalography (SEEG) electrodes. We evaluate the effect of line-of-best-fit approaches, EP definition and lateral versus Euclidean distance on metrics to provide recommendations for reporting implantation accuracy metrics.

Results

We found no bias between manual and automated approaches for calculating accuracy metrics with limits of agreement of ±1 mm and ±1°. Automated metrics are robust to sources of errors including registration and electrode bending. We observe the highest error in EP deviations of μ = 0.25 mm when the post-implantation CT is used to define the point of entry.

Comparison with Existing Method(s)

We found no reports of automated approaches for quality assessment of SEEG electrode implantation. Neither the choice of metrics nor the possible errors that could occur have been investigated previously.

Conclusions

Our automated approach is useful to avoid human errors, unintentional bias and variation that may be introduced when manually computing metrics. Our work is relevant and timely to facilitate comparisons of studies reporting implantation accuracy.

Intrapapillary capillary loop classification in magnification endoscopy: open dataset and baseline methodology

PURPOSE

Early squamous cell neoplasia (ESCN) in the oesophagus is a highly treatable condition. Lesions confined to the mucosal layer can be curatively treated endoscopically. We build a computer-assisted detection system that can classify still images or video frames as normal or abnormal with high diagnostic accuracy.

METHODS

We present a new benchmark dataset containing 68K binary labelled frames extracted from 114 patient videos whose imaged areas have been resected and correlated to histopathology. Our novel convolutional network architecture solves the binary classification task and explains what features of the input domain drive the decision-making process of the network.

RESULTS

The proposed method achieved an average accuracy of 91.7% compared to the 94.7% achieved by a group of 12 senior clinicians. Our novel network architecture produces deeply supervised activation heatmaps that suggest the network is looking at intrapapillary capillary loop patterns when predicting abnormality.

CONCLUSION

We believe that this dataset and baseline method may serve as a reference for future benchmarks on both video frame classification and explainability in the context of ESCN detection. A future work path of high clinical relevance is the extension of the classification to ESCN types.

An automated framework for localization, segmentation and super-resolution reconstruction of fetal brain MRI

High-resolution volume reconstruction from multiple motion-corrupted stacks of 2D slices plays an increasing role for fetal brain Magnetic Resonance Imaging (MRI) studies. Currently existing reconstruction methods are time-consuming and often require user interactions to localize and extract the brain from several stacks of 2D slices. We propose a fully automatic framework for fetal brain reconstruction that consists of four stages: 1) fetal brain localization based on a coarse segmentation by a Convolutional Neural Network (CNN), 2) fine segmentation by another CNN trained with a multi-scale loss function, 3) novel, single-parameter outlier-robust super-resolution reconstruction, and 4) fast and automatic high-resolution visualization in standard anatomical space suitable for pathological brains. We validated our framework with images from fetuses with normal brains and with variable degrees of ventriculomegaly associated with open spina bifida, a congenital malformation affecting also the brain. Experiments show that each step of our proposed pipeline outperforms state-of-the-art methods in both segmentation and reconstruction comparisons including expert-reader quality assessments. The reconstruction results of our proposed method compare favorably with those obtained by manual, labor-intensive brain segmentation, which unlocks the potential use of automatic fetal brain reconstruction studies in clinical practice.

Comparison of manual and semi-automatic registration in augmented reality image-guided liver surgery: a clinical feasibility study

BACKGROUND

The laparoscopic approach to liver resection may reduce morbidity and hospital stay. However, uptake has been slow due to concerns about patient safety and oncological radicality. Image guidance systems may improve patient safety by enabling 3D visualisation of critical intra- and extrahepatic structures. Current systems suffer from non-intuitive visualisation and a complicated setup process. A novel image guidance system (SmartLiver), offering augmented reality visualisation and semi-automatic registration has been developed to address these issues. A clinical feasibility study evaluated the performance and usability of SmartLiver with either manual or semi-automatic registration.

METHODS

Intraoperative image guidance data were recorded and analysed in patients undergoing laparoscopic liver resection or cancer staging. Stereoscopic surface reconstruction and iterative closest point matching facilitated semi-automatic registration. The primary endpoint was defined as successful registration as determined by the operating surgeon. Secondary endpoints were system usability as assessed by a surgeon questionnaire and comparison of manual vs. semi-automatic registration accuracy. Since SmartLiver is still in development no attempt was made to evaluate its impact on perioperative outcomes.

RESULTS

The primary endpoint was achieved in 16 out of 18 patients. Initially semi-automatic registration failed because the IGS could not distinguish the liver surface from surrounding structures. Implementation of a deep learning algorithm enabled the IGS to overcome this issue and facilitate semi-automatic registration. Mean registration accuracy was 10.9 ± 4.2 mm (manual) vs. 13.9 ± 4.4 mm (semi-automatic) (Mean difference - 3 mm; p = 0.158). Surgeon feedback was positive about IGS handling and improved intraoperative orientation but also highlighted the need for a simpler setup process and better integration with laparoscopic ultrasound.

CONCLUSION

The technical feasibility of using SmartLiver intraoperatively has been demonstrated. With further improvements semi-automatic registration may enhance user friendliness and workflow of SmartLiver. Manual and semi-automatic registration accuracy were comparable but evaluation on a larger patient cohort is required to confirm these findings.

Automated postoperative muscle assessment of hip arthroplasty patients using multimodal imaging joint segmentation

BACKGROUND AND OBJECTIVE

In patients treated with hip arthroplasty, the muscular condition and presence of inflammatory reactions are assessed using magnetic resonance imaging (MRI). As MRI lacks contrast for bony structures, computed tomography (CT) is preferred for clinical evaluation of bone tissue and orthopaedic surgical planning. Combining the complementary information of MRI and CT could improve current clinical practice for diagnosis, monitoring and treatment planning. In particular, the different contrast of these modalities could help better quantify the presence of fatty infiltration to characterise muscular condition and assess implant failure. In this work, we combine CT and MRI for joint bone and muscle segmentation and we propose a novel Intramuscular Fat Fraction estimation method for the quantification of muscle atrophy.

METHODS

Our multimodal framework is able to segment healthy and pathological musculoskeletal structures as well as implants, and develops into three steps. First, input images are pre-processed to improve the low quality of clinically acquired images and to reduce the noise associated with metal artefact. Subsequently, CT and MRI are non-linearly aligned using a novel approach which imposes rigidity constraints on bony structures to ensure realistic deformation. Finally, taking advantage of a multimodal atlas we created for this task, a multi-atlas based segmentation delineates pelvic bones, abductor muscles and implants on both modalities jointly. From the obtained segmentation, a multimodal estimation of the Intramuscular Fat Fraction can be automatically derived.

RESULTS

Evaluation of the segmentation in a leave-one-out cross-validation study on 22 hip sides resulted in an average Dice score of 0.90 for skeletal and 0.84 for muscular structures. Our multimodal Intramuscular Fat Fraction was benchmarked on 27 different cases against a standard radiological score, showing stronger association than a single modality approach in a one-way ANOVA F-test analysis.

CONCLUSIONS

The proposed framework represents a promising tool to support image analysis in hip arthroplasty, being robust to the presence of implants and associated image artefacts. By allowing for the automated extraction of a muscle atrophy imaging biomarker, it could quantitatively inform the decision-making process about patient's management.

Cardiovascular Risk Factors and White Matter Hyperintensities: Difference in Susceptibility in South Asians Compared With Europeans

Background

Cardiovascular risk factors vary between ethnicities but little is known about their differential effects on white matter hyperintensities (WMH), an indicator of brain aging and burden of cerebrovascular disease.

Methods and Results

Brain magnetic resonance imaging scans from 213 people of South Asian and 256 of European ethnicity (total=469) were analyzed for global and regional WMH load. Associations with cardiovascular risk factors and a composite cardiovascular risk score (National Cholesterol Education Programme Adult Treatment Panel III) were compared by ethnicity, diabetes mellitus, smoking, and hypertension status. Distributional patterns of WMH were similar by ethnicity but the vulnerability to specific risk factors differed. Associations between WMH and age or National Cholesterol Education Programme Adult Treatment Panel III scores were stronger in South Asians compared with Europeans. For instance, a year of age led to an excess of 3.8% (confidence interval=[0.2, 7.6]; P=0.04) of WMH load in frontal regions in South Asians compared with Europeans. In the diabetic subgroup, South Asians had more WMH than Europeans (+63.3%, confidence interval=[14.1, 133.9]; P=0.007), particularly in the deeper regions (+102% confidence interval=[24, 329]; P=0.004). In the population as a whole, diabetes mellitus was not, or only weakly, related to an increase in WMH volume (12.4%, confidence interval=[−10.7, 41.3]; P=0.32), and diabetes mellitus duration was a positive predictor of frontal periventricular WMH load in Europeans but not in South Asians. In turn, diastolic blood pressure was positively associated with WMH volumes in South Asians but not in Europeans. Hypertension was not associated with WMH load (P=0.9).

Conclusions

Distribution patterns of WMH are similar in South Asians and Europeans but older age and higher cardiovascular risk are associated with more WMH in South Asians.