Quality assurance of late gadolinium enhancement cardiac magnetic resonance images: a deep learning classifier for confidence in the presence or absence of abnormality with potential to prompt real-time image optimization.

BACKGROUND: Late gadolinium enhancement (LGE) of the myocardium has significant diagnostic and prognostic implications, with even small areas of enhancement being important. Distinguishing between definitely normal and definitely abnormal LGE images is usually straightforward, but diagnostic uncertainty arises when reporters are not sure whether the observed LGE is genuine or not. This uncertainty might be resolved by repetition (to remove artifact) or further acquisition of intersecting images, but this must take place before the scan finishes. Real-time quality assurance by humans is a complex task requiring training and experience, so being able to identify which images have an intermediate likelihood of LGE while the scan is ongoing, without the presence of an expert is of high value. This decision-support could prompt immediate image optimization or acquisition of supplementary images to confirm or refute the presence of genuine LGE. This could reduce ambiguity in reports. METHODS: Short-axis, phase-sensitive inversion recovery late gadolinium images were extracted from our clinical cardiac magnetic resonance (CMR) database and shuffled. Two, independent, blinded experts scored each individual slice for "LGE likelihood" on a visual analog scale, from 0 (absolute certainty of no LGE) to 100 (absolute certainty of LGE), with 50 representing clinical equipoise. The scored images were split into two classes-either "high certainty" of whether LGE was present or not, or "low certainty." The dataset was split into training, validation, and test sets (70:15:15). A deep learning binary classifier based on the EfficientNetV2 convolutional neural network architecture was trained to distinguish between these categories. Classifier performance on the test set was evaluated by calculating the accuracy, precision, recall, F1-score, and area under the receiver operating characteristics curve (ROC AUC). Performance was also evaluated on an external test set of images from a different center. RESULTS: One thousand six hundred and forty-five images (from 272 patients) were labeled and split at the patient level into training (1151 images), validation (247 images), and test (247 images) sets for the deep learning binary classifier. Of these, 1208 images were "high certainty" (255 for LGE, 953 for no LGE), and 437 were "low certainty". An external test comprising 247 images from 41 patients from another center was also employed. After 100 epochs, the performance on the internal test set was accuracy = 0.94, recall = 0.80, precision = 0.97, F1-score = 0.87, and ROC AUC = 0.94. The classifier also performed robustly on the external test set (accuracy = 0.91, recall = 0.73, precision = 0.93, F1-score = 0.82, and ROC AUC = 0.91). These results were benchmarked against a reference inter-expert accuracy of 0.86. CONCLUSION: Deep learning shows potential to automate quality control of late gadolinium imaging in CMR. The ability to identify short-axis images with intermediate LGE likelihood in real-time may serve as a useful decision-support tool. This approach has the potential to guide immediate further imaging while the patient is still in the scanner, thereby reducing the frequency of recalls and inconclusive reports due to diagnostic indecision.

The impact of artificial intelligence on clinical education: perceptions of postgraduate trainee doctors in London (UK) and recommendations for trainers.

BACKGROUND: Artificial intelligence (AI) technologies are increasingly used in clinical practice. Although there is robust evidence that AI innovations can improve patient care, reduce clinicians' workload and increase efficiency, their impact on medical training and education remains unclear. METHODS: A survey of trainee doctors' perceived impact of AI technologies on clinical training and education was conducted at UK NHS postgraduate centers in London between October and December 2020. Impact assessment mirrored domains in training curricula such as 'clinical judgement', 'practical skills' and 'research and quality improvement skills'. Significance between Likert-type data was analysed using Fisher's exact test. Response variations between clinical specialities were analysed using k-modes clustering. Free-text responses were analysed by thematic analysis. RESULTS: Two hundred ten doctors responded to the survey (response rate 72%). The majority (58%) perceived an overall positive impact of AI technologies on their training and education. Respondents agreed that AI would reduce clinical workload (62%) and improve research and audit training (68%). Trainees were skeptical that it would improve clinical judgement (46% agree, p = 0.12) and practical skills training (32% agree, p < 0.01). The majority reported insufficient AI training in their current curricula (92%), and supported having more formal AI training (81%). CONCLUSIONS: Trainee doctors have an overall positive perception of AI technologies' impact on clinical training. There is optimism that it will improve 'research and quality improvement' skills and facilitate 'curriculum mapping'. There is skepticism that it may reduce educational opportunities to develop 'clinical judgement' and 'practical skills'. Medical educators should be mindful that these domains are protected as AI develops. We recommend that 'Applied AI' topics are formalized in curricula and digital technologies leveraged to deliver clinical education.

Efficient labelling for efficient deep learning: the benefit of a multiple-image-ranking method to generate high volume training data applied to ventricular slice level classification in cardiac MRI.

Background: Getting the most value from expert clinicians' limited labelling time is a major challenge for artificial intelligence (AI) development in clinical imaging. We present a novel method for ground-truth labelling of cardiac magnetic resonance imaging (CMR) image data by leveraging multiple clinician experts ranking multiple images on a single ordinal axis, rather than manual labelling of one image at a time. We apply this strategy to train a deep learning (DL) model to classify the anatomical position of CMR images. This allows the automated removal of slices that do not contain the left ventricular (LV) myocardium. Methods: Anonymised LV short-axis slices from 300 random scans (3,552 individual images) were extracted. Each image's anatomical position relative to the LV was labelled using two different strategies performed for 5 hours each: (I) 'one-image-at-a-time': each image labelled according to its position: 'too basal', 'LV', or 'too apical' individually by one of three experts; and (II) 'multiple-image-ranking': three independent experts ordered slices according to their relative position from 'most-basal' to 'most apical' in batches of eight until each image had been viewed at least 3 times. Two convolutional neural networks were trained for a three-way classification task (each model using data from one labelling strategy). The models' performance was evaluated by accuracy, F1-score, and area under the receiver operating characteristics curve (ROC AUC). Results: After excluding images with artefact, 3,323 images were labelled by both strategies. The model trained using labels from the 'multiple-image-ranking strategy' performed better than the model using the 'one-image-at-a-time' labelling strategy (accuracy 86% vs. 72%, P=0.02; F1-score 0.86 vs. 0.75; ROC AUC 0.95 vs. 0.86). For expert clinicians performing this task manually the intra-observer variability was low (Cohen's κ=0.90), but the inter-observer variability was higher (Cohen's κ=0.77). Conclusions: We present proof of concept that, given the same clinician labelling effort, comparing multiple images side-by-side using a 'multiple-image-ranking' strategy achieves ground truth labels for DL more accurately than by classifying images individually. We demonstrate a potential clinical application: the automatic removal of unrequired CMR images. This leads to increased efficiency by focussing human and machine attention on images which are needed to answer clinical questions.

Kinematic-Model-Free Predictive Control for Robotic Manipulator Target Reaching With Obstacle Avoidance.

Model predictive control is a widely used optimal control method for robot path planning and obstacle avoidance. This control method, however, requires a system model to optimize control over a finite time horizon and possible trajectories. Certain types of robots, such as soft robots, continuum robots, and transforming robots, can be challenging to model, especially in unstructured or unknown environments. Kinematic-model-free control can overcome these challenges by learning local linear models online. This paper presents a novel perception-based robot motion controller, the kinematic-model-free predictive controller, that is capable of controlling robot manipulators without any prior knowledge of the robot's kinematic structure and dynamic parameters and is able to perform end-effector obstacle avoidance. Simulations and physical experiments were conducted to demonstrate the ability and adaptability of the controller to perform simultaneous target reaching and obstacle avoidance.

Towards Instant Calibration in Myoelectric Prosthetic Hands: A Highly Data-Efficient Controller Based on the Wasserstein Distance.

Prosthetic hand control research typically focuses on developing increasingly complex controllers to achieve diminishing returns in pattern recognition of muscle activity signals, making models less suitable for user calibration. Some works have investigated transfer learning to alleviate this, but such approaches increase model size dramatically-thus reducing their suitability for implementation on real prostheses. In this work, we propose a novel, non-parametric controller that uses the Wasserstein distance to compare the distribution of incoming signals to those of a set of reference distributions, with the intended action classified as the closest distribution. This controller requires only a single capture of data per reference distribution, making calibration almost instantaneous. Preliminary experiments building a reference library show that, in theory, users are able to produce up to 9 distinguishable muscle activity patterns. However, in practice, variation when repeating actions reduces this. Controller accuracy results show that 10 non-disabled and 1 disabled participant were able to use the controller with a maximum of two recalibrations to perform 6 actions at an average accuracy of 89.9% and 86.7% respectively. Practical experiments show that the controller allows users to complete all tasks of the Jebsen-Taylor Hand Function Test, although the task of picking and placing small common objects required on average more time than the protocol's maximum time.

Aerosolised fluorescein can quantify FFP mask faceseal leakage: a cost-effective adaptation to the existing point of care fit test.

BACKGROUND: A qualitative fit test using bitter-tasting aerosols is the commonest way to determine filtering face-piece (FFP) mask leakage. This taste test is subjective and biased by placebo. We propose a cheap, quantitative modification of the taste test by measuring the amount of fluorescein stained filter paper behind the mask using image analysis. METHODS: A bitter-tasting fluorescein solution was aerosolised during mask fit tests, with filter paper placed on masks' inner surfaces. Participants reported whether they could taste bitterness to determine taste test 'pass' or 'fail' results. Filter paper photographs were digitally analysed to quantify total fluorescence (TF). RESULTS: Fifty-six healthcare professionals were fit tested; 32 (57%) 'passed' the taste test. TF between the taste test 'pass' and 'fail' groups was significantly different (p<0.001). A cut-off (TF = 5.0 × 106 units) was determined at precision (78%) and recall (84%), resulting in 5/56 participants (9%) reclassified from 'pass' to 'fail' by the fluorescein test. Seven out of 56 (12%) reclassified from 'fail' to 'pass'. CONCLUSION: Fluorescein is detectable and sensitive at identifying FFP mask leaks. These low-cost adaptations can enhance exiting fit testing to determine 'pass' and 'fail' groups, protecting those who 'passed' the taste test but have high fluorescein leak, and reassuring those who 'failed' the taste test despite having little fluorescein leak.