Artificial intelligence education for radiographers, an evaluation of a UK postgraduate educational intervention using participatory action research: a pilot study.

BACKGROUND: Artificial intelligence (AI)-enabled applications are increasingly being used in providing healthcare services, such as medical imaging support. Sufficient and appropriate education for medical imaging professionals is required for successful AI adoption. Although, currently, there are AI training programmes for radiologists, formal AI education for radiographers is lacking. Therefore, this study aimed to evaluate and discuss a postgraduate-level module on AI developed in the UK for radiographers. METHODOLOGY: A participatory action research methodology was applied, with participants recruited from the first cohort of students enrolled in this module and faculty members. Data were collected using online, semi-structured, individual interviews and focus group discussions. Textual data were processed using data-driven thematic analysis. RESULTS: Seven students and six faculty members participated in this evaluation. Results can be summarised in the following four themes: a. participants' professional and educational backgrounds influenced their experiences, b. participants found the learning experience meaningful concerning module design, organisation, and pedagogical approaches, c. some module design and delivery aspects were identified as barriers to learning, and d. participants suggested how the ideal AI course could look like based on their experiences. CONCLUSIONS: The findings of our work show that an AI module can assist educators/academics in developing similar AI education provisions for radiographers and other medical imaging and radiation sciences professionals. A blended learning delivery format, combined with customisable and contextualised content, using an interprofessional faculty approach is recommended for future similar courses.

Enhancing MR image segmentation with realistic adversarial data augmentation.

The success of neural networks on medical image segmentation tasks typically relies on large labeled datasets for model training. However, acquiring and manually labeling a large medical image set is resource-intensive, expensive, and sometimes impractical due to data sharing and privacy issues. To address this challenge, we propose AdvChain, a generic adversarial data augmentation framework, aiming at improving both the diversity and effectiveness of training data for medical image segmentation tasks. AdvChain augments data with dynamic data augmentation, generating randomly chained photo-metric and geometric transformations to resemble realistic yet challenging imaging variations to expand training data. By jointly optimizing the data augmentation model and a segmentation network during training, challenging examples are generated to enhance network generalizability for the downstream task. The proposed adversarial data augmentation does not rely on generative networks and can be used as a plug-in module in general segmentation networks. It is computationally efficient and applicable for both low-shot supervised and semi-supervised learning. We analyze and evaluate the method on two MR image segmentation tasks: cardiac segmentation and prostate segmentation with limited labeled data. Results show that the proposed approach can alleviate the need for labeled data while improving model generalization ability, indicating its practical value in medical imaging applications.

MOOD 2020: A Public Benchmark for Out-of-Distribution Detection and Localization on Medical Images.

Detecting Out-of-Distribution (OoD) data is one of the greatest challenges in safe and robust deployment of machine learning algorithms in medicine. When the algorithms encounter cases that deviate from the distribution of the training data, they often produce incorrect and over-confident predictions. OoD detection algorithms aim to catch erroneous predictions in advance by analysing the data distribution and detecting potential instances of failure. Moreover, flagging OoD cases may support human readers in identifying incidental findings. Due to the increased interest in OoD algorithms, benchmarks for different domains have recently been established. In the medical imaging domain, for which reliable predictions are often essential, an open benchmark has been missing. We introduce the Medical-Out-Of-Distribution-Analysis-Challenge (MOOD) as an open, fair, and unbiased benchmark for OoD methods in the medical imaging domain. The analysis of the submitted algorithms shows that performance has a strong positive correlation with the perceived difficulty, and that all algorithms show a high variance for different anomalies, making it yet hard to recommend them for clinical practice. We also see a strong correlation between challenge ranking and performance on a simple toy test set, indicating that this might be a valuable addition as a proxy dataset during anomaly detection algorithm development.

Anti-transfer learning for task invariance in convolutional neural networks for speech processing.

We introduce the novel concept of anti-transfer learning for speech processing with convolutional neural networks. While transfer learning assumes that the learning process for a target task will benefit from re-using representations learned for another task, anti-transfer avoids the learning of representations that have been learned for an orthogonal task, i.e., one that is not relevant and potentially confounding for the target task, such as speaker identity for speech recognition or speech content for emotion recognition. This extends the potential use of pre-trained models that have become increasingly available. In anti-transfer learning, we penalize similarity between activations of a network being trained on a target task and another one previously trained on an orthogonal task, which yields more suitable representations. This leads to better generalization and provides a degree of control over correlations that are spurious or undesirable, e.g. to avoid social bias. We have implemented anti-transfer for convolutional neural networks in different configurations with several similarity metrics and aggregation functions, which we evaluate and analyze with several speech and audio tasks and settings, using six datasets. We show that anti-transfer actually leads to the intended invariance to the orthogonal task and to more appropriate features for the target task at hand. Anti-transfer learning consistently improves classification accuracy in all test cases. While anti-transfer creates computation and memory cost at training time, there is relatively little computation cost when using pre-trained models for orthogonal tasks. Anti-transfer is widely applicable and particularly useful where a specific invariance is desirable or where labeled data for orthogonal tasks are difficult to obtain on a given dataset but pre-trained models are available.

A population-based phenome-wide association study of cardiac and aortic structure and function.

Differences in cardiac and aortic structure and function are associated with cardiovascular diseases and a wide range of other types of disease. Here we analyzed cardiovascular magnetic resonance images from a population-based study, the UK Biobank, using an automated machine-learning-based analysis pipeline. We report a comprehensive range of structural and functional phenotypes for the heart and aorta across 26,893 participants, and explore how these phenotypes vary according to sex, age and major cardiovascular risk factors. We extended this analysis with a phenome-wide association study, in which we tested for correlations of a wide range of non-imaging phenotypes of the participants with imaging phenotypes. We further explored the associations of imaging phenotypes with early-life factors, mental health and cognitive function using both observational analysis and Mendelian randomization. Our study illustrates how population-based cardiac and aortic imaging phenotypes can be used to better define cardiovascular disease risks as well as heart-brain health interactions, highlighting new opportunities for studying disease mechanisms and developing image-based biomarkers.

Deep Learning for Cardiac Image Segmentation: A Review.

Deep learning has become the most widely used approach for cardiac image segmentation in recent years. In this paper, we provide a review of over 100 cardiac image segmentation papers using deep learning, which covers common imaging modalities including magnetic resonance imaging (MRI), computed tomography (CT), and ultrasound and major anatomical structures of interest (ventricles, atria, and vessels). In addition, a summary of publicly available cardiac image datasets and code repositories are included to provide a base for encouraging reproducible research. Finally, we discuss the challenges and limitations with current deep learning-based approaches (scarcity of labels, model generalizability across different domains, interpretability) and suggest potential directions for future research.

Large-scale Quality Control of Cardiac Imaging in Population Studies: Application to UK Biobank.

In large population studies such as the UK Biobank (UKBB), quality control of the acquired images by visual assessment is unfeasible. In this paper, we apply a recently developed fully-automated quality control pipeline for cardiac MR (CMR) images to the first 19,265 short-axis (SA) cine stacks from the UKBB. We present the results for the three estimated quality metrics (heart coverage, inter-slice motion and image contrast in the cardiac region) as well as their potential associations with factors including acquisition details and subject-related phenotypes. Up to 14.2% of the analysed SA stacks had sub-optimal coverage (i.e. missing basal and/or apical slices), however most of them were limited to the first year of acquisition. Up to 16% of the stacks were affected by noticeable inter-slice motion (i.e. average inter-slice misalignment greater than 3.4 mm). Inter-slice motion was positively correlated with weight and body surface area. Only 2.1% of the stacks had an average end-diastolic cardiac image contrast below 30% of the dynamic range. These findings will be highly valuable for both the scientists involved in UKBB CMR acquisition and for the ones who use the dataset for research purposes.

Explainable Anatomical Shape Analysis Through Deep Hierarchical Generative Models.

Quantification of anatomical shape changes currently relies on scalar global indexes which are largely insensitive to regional or asymmetric modifications. Accurate assessment of pathology-driven anatomical remodeling is a crucial step for the diagnosis and treatment of many conditions. Deep learning approaches have recently achieved wide success in the analysis of medical images, but they lack interpretability in the feature extraction and decision processes. In this work, we propose a new interpretable deep learning model for shape analysis. In particular, we exploit deep generative networks to model a population of anatomical segmentations through a hierarchy of conditional latent variables. At the highest level of this hierarchy, a two-dimensional latent space is simultaneously optimised to discriminate distinct clinical conditions, enabling the direct visualisation of the classification space. Moreover, the anatomical variability encoded by this discriminative latent space can be visualised in the segmentation space thanks to the generative properties of the model, making the classification task transparent. This approach yielded high accuracy in the categorisation of healthy and remodelled left ventricles when tested on unseen segmentations from our own multi-centre dataset as well as in an external validation set, and on hippocampi from healthy controls and patients with Alzheimer's disease when tested on ADNI data. More importantly, it enabled the visualisation in three-dimensions of both global and regional anatomical features which better discriminate between the conditions under exam. The proposed approach scales effectively to large populations, facilitating high-throughput analysis of normal anatomy and pathology in large-scale studies of volumetric imaging.

Learning-Based Quality Control for Cardiac MR Images.

The effectiveness of a cardiovascular magnetic resonance (CMR) scan depends on the ability of the operator to correctly tune the acquisition parameters to the subject being scanned and on the potential occurrence of imaging artifacts, such as cardiac and respiratory motion. In the clinical practice, a quality control step is performed by visual assessment of the acquired images; however, this procedure is strongly operator-dependent, cumbersome, and sometimes incompatible with the time constraints in clinical settings and large-scale studies. We propose a fast, fully automated, and learning-based quality control pipeline for CMR images, specifically for short-axis image stacks. Our pipeline performs three important quality checks: 1) heart coverage estimation; 2) inter-slice motion detection; 3) image contrast estimation in the cardiac region. The pipeline uses a hybrid decision forest method-integrating both regression and structured classification models-to extract landmarks and probabilistic segmentation maps from both long- and short-axis images as a basis to perform the quality checks. The technique was tested on up to 3000 cases from the UK Biobank and on 100 cases from the UK Digital Heart Project and validated against manual annotations and visual inspections performed by expert interpreters. The results show the capability of the proposed pipeline to correctly detect incomplete or corrupted scans (e.g., on UK Biobank, sensitivity and specificity, respectively, 88% and 99% for heart coverage estimation and 85% and 95% for motion detection), allowing their exclusion from the analyzed dataset or the triggering of a new acquisition.

Automated cardiovascular magnetic resonance image analysis with fully convolutional networks.

BACKGROUND: Cardiovascular resonance (CMR) imaging is a standard imaging modality for assessing cardiovascular diseases (CVDs), the leading cause of death globally. CMR enables accurate quantification of the cardiac chamber volume, ejection fraction and myocardial mass, providing information for diagnosis and monitoring of CVDs. However, for years, clinicians have been relying on manual approaches for CMR image analysis, which is time consuming and prone to subjective errors. It is a major clinical challenge to automatically derive quantitative and clinically relevant information from CMR images. METHODS: Deep neural networks have shown a great potential in image pattern recognition and segmentation for a variety of tasks. Here we demonstrate an automated analysis method for CMR images, which is based on a fully convolutional network (FCN). The network is trained and evaluated on a large-scale dataset from the UK Biobank, consisting of 4,875 subjects with 93,500 pixelwise annotated images. The performance of the method has been evaluated using a number of technical metrics, including the Dice metric, mean contour distance and Hausdorff distance, as well as clinically relevant measures, including left ventricle (LV) end-diastolic volume (LVEDV) and end-systolic volume (LVESV), LV mass (LVM); right ventricle (RV) end-diastolic volume (RVEDV) and end-systolic volume (RVESV). RESULTS: By combining FCN with a large-scale annotated dataset, the proposed automated method achieves a high performance in segmenting the LV and RV on short-axis CMR images and the left atrium (LA) and right atrium (RA) on long-axis CMR images. On a short-axis image test set of 600 subjects, it achieves an average Dice metric of 0.94 for the LV cavity, 0.88 for the LV myocardium and 0.90 for the RV cavity. The mean absolute difference between automated measurement and manual measurement is 6.1 mL for LVEDV, 5.3 mL for LVESV, 6.9 gram for LVM, 8.5 mL for RVEDV and 7.2 mL for RVESV. On long-axis image test sets, the average Dice metric is 0.93 for the LA cavity (2-chamber view), 0.95 for the LA cavity (4-chamber view) and 0.96 for the RA cavity (4-chamber view). The performance is comparable to human inter-observer variability. CONCLUSIONS: We show that an automated method achieves a performance on par with human experts in analysing CMR images and deriving clinically relevant measures.

Large high-density lipoprotein particle number is independently associated with microvascular function in patients with well-controlled low-density lipoprotein concentration: A vasodilator stress magnetic resonance perfusion study.

BACKGROUND: Abnormalities in total cholesterol, high-density lipoprotein, low-density lipoprotein, and triglycerides are associated with microvascular dysfunction. Recent studies suggest that lipid subfractions better predict atherogenic burden than a routine lipid panel. We sought to determine, whether lipid subfractions are more strongly associated with microvascular function and subclinical atherosclerosis, than conventional lipid measurements using vasodilator stress cardiovascular magnetic resonance (CMR). METHODS: Twenty-four adults referred for risk stratification from a lipid clinic with low-density lipoprotein cholesterol (LDL-C) <100 mg/dL underwent vasodilator CMR. Time-intensity curves generated from stress and rest perfusion images were used to determine the area under the curve (AUC) for the mid-ventricular slice myocardium and the left ventricular (LV) cavity. Myocardial perfusion reserve index (MPRi) was defined as stress to rest ratio of mid-ventricular myocardium AUC, normalized to LV cavity AUC. Lipid panels that included subfractions of LDL and high-density lipoprotein (HDL) were measured using nuclear magnetic resonance testing. The association between MPRi and lipid parameters was examined using univariate linear regression; lipid components statistically correlated with MPRi (P < .05) were then subjected to multivariate analysis. RESULTS: Univariate regression analysis showed MPRi was associated with HDL-C, triglycerides, large HDL-P, and small LDL-P; no association was found between MPRi and total cholesterol, LDL-C, total LDL-P, or total HDL-P. Using multivariate analysis, large HDL-P was independently associated with MPRi. CONCLUSIONS: In patients with LDL-C <100 mg/dL, large HDL-P is independently associated with CMR-derived myocardial perfusion reserve, a surrogate for microvascular function and subclinical atherosclerosis. Further studies using lipid subfractions to better understand cardiovascular risks are warranted.

Fully-automated identification and segmentation of aortic lumen from fetal ultrasound images.

Intrauterine growth restriction (IUGR) is a fetal condition that has been linked to an increase in cardiovascular mortality in the adult life. IUGR induces cardiovascular remodeling, including a decrease in aortic intima-media thickness (aIMT) which can be evaluated using fetal ultrasound imaging, potentially improving IUGR assessment and cardiovascular risk management. A necessary step for aIMT quantification is the identification of a region-of-interest (ROI) containing the lumen. This step is usually performed manually, even within the few semi-automated approaches to aIMT estimation. The aims of this study were to develop and test a fully-automated technique for lumen identification and segmentation from ultrasound images as a basis for aIMT quantification. The technique relies on convolution with a set of discriminative kernels learned from a training dataset using an AdaBoost classifier followed by segmentation based on anisotropic filtering and level-set methods. This approach was tested on 50 images acquired from 5 subjects: automatically extracted mean lumen width values were compared to reference ones manually obtained by an experienced interpreter. Results (R = 0.97) show that the proposed technique is accurate, suggesting that it could serve as a basis for fully-automated approaches to aIMT quantification.

A supervised learning approach for the robust detection of heart beat in plethysmographic data.

Wearable devices equipped with photoplethysmography (PPG) sensors are gaining an increased interest in the context of biometric signal monitoring within clinical, e-health and fitness settings. When used in everyday life and during exercise, PPG traces are heavily affected by artifacts originating from motion and from a non constant positioning and contact of the PPG sensor with the skin. Many algorithms have been developed for the estimation of heart-rate from photoplethysmography signals. We remark that they were mainly conceived and tested in controlled settings and, in turn, do not provide robust performance, even during moderate exercise. Only a few of them have been designed for signals acquired at rest and during fitness. However, they provide the required resilience to motion artifacts at the cost of using computationally demanding signal processing tools. At variance with other methods from the literature, we propose a supervised learning approach, where a classifier is trained on a set of labelled data to detect the presence of heart beats at each position of a PPG signal, with only little preprocessing and postprocessing. We show that the results obtained on the TROIKA dataset using our approach are comparable with those shown in the original paper, providing a classification error of 14% in the detection of heart beat positions, that reduces to 2.86% on the heart-rate estimates after the postprocessing step.

Myocardial perfusion: near-automated evaluation from contrast-enhanced MR images obtained at rest and during vasodilator stress.

PURPOSE: To develop and validate a technique for near-automated definition of myocardial regions of interest suitable for perfusion evaluation during vasodilator stress cardiac magnetic resonance (MR) imaging. MATERIALS AND METHODS: The institutional review board approved the study protocol, and all patients provided informed consent. Image noise density distribution was used as a basis for endocardial and epicardial border detection combined with nonrigid registration. This method was tested in 42 patients undergoing contrast material-enhanced cardiac MR imaging (at 1.5 T) at rest and during vasodilator (adenosine or regadenoson) stress, including 15 subjects with normal myocardial perfusion and 27 patients referred for coronary angiography. Contrast enhancement-time curves were near-automatically generated and were used to calculate perfusion indexes. The results were compared with results of conventional manual analysis, using quantitative coronary angiography results as a reference for stenosis greater than 50%. Statistical analyses included the Student t test, linear regression, Bland-Altman analysis, and κ statistics. RESULTS: Analysis of one sequence required less than 1 minute and resulted in high-quality contrast enhancement curves both at rest and stress (mean signal-to-noise ratios, 17±7 [standard deviation] and 22±8, respectively), showing expected patterns of first-pass perfusion. Perfusion indexes accurately depicted stress-induced hyperemia (increased upslope, from 6.7 sec(-1)±2.3 to 15.6 sec(-1)±5.9; P<.0001). Measured segmental pixel intensities correlated highly with results of manual analysis (r=0.95). The derived perfusion indexes also correlated highly with (r up to 0.94) and showed the same diagnostic accuracy as manual analysis (area under the receiver operating characteristic curve, up to 0.72 vs 0.73). CONCLUSION: Despite the dynamic nature of contrast-enhanced image sequences and respiratory motion, fast near-automated detection of myocardial segments and accurate quantification of tissue contrast is feasible at rest and during vasodilator stress. This technique, shown to be as accurate as conventional manual analysis, allows detection of stress-induced perfusion abnormalities.

Prosthetic component segmentation with blur compensation: a fast method for 3D fluoroscopy.

A new method for prosthetic component segmentation from fluoroscopic images is presented. The hybrid approach we propose combines diffusion filtering, region growing and level-set techniques without exploiting any a priori knowledge of the analyzed geometry. The method was evaluated on a synthetic dataset including 270 images of knee and hip prosthesis merged to real fluoroscopic data simulating different conditions of blurring and illumination gradient. The performance of the method was assessed by comparing estimated contours to references using different metrics. Results showed that the segmentation procedure is fast, accurate, independent on the operator as well as on the specific geometrical characteristics of the prosthetic component, and able to compensate for amount of blurring and illumination gradient. Importantly, the method allows a strong reduction of required user interaction time when compared to traditional segmentation techniques. Its effectiveness and robustness in different image conditions, together with simplicity and fast implementation, make this prosthetic component segmentation procedure promising and suitable for multiple clinical applications including assessment of in vivo joint kinematics in a variety of cases.