Publications
352 results found
Glocker B, Jones C, Roschewitz M, et al., 2022, Risk of Bias in Chest Radiography Deep Learning Foundation Models
Purpose: To analyze a recently published chest radiography foundation modelfor the presence of biases that could lead to subgroup performance disparitiesacross biological sex and race. Materials and Methods: This retrospective study used 127,118 chestradiographs from 42,884 patients (mean age, 63 [SD] 17 years; 23,623 male,19,261 female) from the CheXpert dataset collected between October 2002 andJuly 2017. To determine the presence of bias in features generated by a chestradiography foundation model and baseline deep learning model, dimensionalityreduction methods together with two-sample Kolmogorov-Smirnov tests were usedto detect distribution shifts across sex and race. A comprehensive diseasedetection performance analysis was then performed to associate any biases inthe features to specific disparities in classification performance acrosspatient subgroups. Results: Ten out of twelve pairwise comparisons across biological sex andrace showed statistically significant differences in the studied foundationmodel, compared with four significant tests in the baseline model. Significantdifferences were found between male and female (P < .001) and Asian and Blackpatients (P < .001) in the feature projections that primarily capture disease.Compared with average model performance across all subgroups, classificationperformance on the 'no finding' label dropped between 6.8% and 7.8% for femalepatients, and performance in detecting 'pleural effusion' dropped between 10.7%and 11.6% for Black patients. Conclusion: The studied chest radiography foundation model demonstratedracial and sex-related bias leading to disparate performance across patientsubgroups and may be unsafe for clinical applications.
Ellis S, Manzanera OEM, Baltatzis V, et al., 2022, Evaluation of 3D GANs for lung tissue modelling in pulmonary CT, The Journal of Machine Learning for Biomedical Imaging, Vol: 1, Pages: 1-36
GANs are able to model accurately the distribution of complex,high-dimensional datasets, e.g. images. This makes high-quality GANs useful forunsupervised anomaly detection in medical imaging. However, differences intraining datasets such as output image dimensionality and appearance ofsemantically meaningful features mean that GAN models from the natural imagedomain may not work `out-of-the-box' for medical imaging, necessitatingre-implementation and re-evaluation. In this work we adapt and evaluate threeGAN models to the task of modelling 3D healthy image patches for pulmonary CT.To the best of our knowledge, this is the first time that such an evaluationhas been performed. The DCGAN, styleGAN and the bigGAN architectures wereinvestigated due to their ubiquity and high performance in natural imageprocessing. We train different variants of these methods and assess theirperformance using the FID score. In addition, the quality of the generatedimages was evaluated by a human observer study, the ability of the networks tomodel 3D domain-specific features was investigated, and the structure of theGAN latent spaces was analysed. Results show that the 3D styleGAN producesrealistic-looking images with meaningful 3D structure, but suffer from modecollapse which must be addressed during training to obtain samples diversity.Conversely, the 3D DCGAN models show a greater capacity for image variability,but at the cost of poor-quality images. The 3D bigGAN models provide anintermediate level of image quality, but most accurately model the distributionof selected semantically meaningful features. The results suggest that futuredevelopment is required to realise a 3D GAN with sufficient capacity forpatch-based lung CT anomaly detection and we offer recommendations for futureareas of research, such as experimenting with other architectures andincorporation of position-encoding.
Taylor-Phillips S, Seedat F, Kijauskaite G, et al., 2022, UK National Screening Committee's approach to reviewing evidence on artificial intelligence in breast cancer screening, The Lancet Digital Health, Vol: 4, Pages: e558-e565, ISSN: 2589-7500
Artificial intelligence (AI) could have the potential to accurately classify mammograms according to the presence or absence of radiological signs of breast cancer, replacing or supplementing human readers (radiologists). The UK National Screening Committee's assessments of the use of AI systems to examine screening mammograms continues to focus on maximising benefits and minimising harms to women screened, when deciding whether to recommend the implementation of AI into the Breast Screening Programme in the UK. Maintaining or improving programme specificity is important to minimise anxiety from false positive results. When considering cancer detection, AI test sensitivity alone is not sufficiently informative, and additional information on the spectrum of disease detected and interval cancers is crucial to better understand the benefits and harms of screening. Although large retrospective studies might provide useful evidence by directly comparing test accuracy and spectrum of disease detected between different AI systems and by population subgroup, most retrospective studies are biased due to differential verification (ie, the use of different reference standards to verify the target condition among study participants). Enriched, multiple-reader, multiple-case, test set laboratory studies are also biased due to the laboratory effect (ie, radiologists' performance in retrospective, laboratory, observer studies is substantially different to their performance in a clinical environment). Therefore, assessment of the effect of incorporating any AI system into the breast screening pathway in prospective studies is required as it will provide key evidence for the effect of the interaction of medical staff with AI, and the impact on women's outcomes.
Newcombe VFJ, Ashton NJ, Posti JP, et al., 2022, Post-acute blood biomarkers and disease progression in traumatic brain injury, BRAIN, Vol: 145, Pages: 2064-2076, ISSN: 0006-8950
- Author Web Link
- Cite
- Citations: 16
Maier-Hein L, Reinke A, Godau P, et al., 2022, Metrics reloaded: Recommendations for image analysis validation
Increasing evidence shows that flaws in machine learning (ML) algorithmvalidation are an underestimated global problem. Particularly in automaticbiomedical image analysis, chosen performance metrics often do not reflect thedomain interest, thus failing to adequately measure scientific progress andhindering translation of ML techniques into practice. To overcome this, ourlarge international expert consortium created Metrics Reloaded, a comprehensiveframework guiding researchers in the problem-aware selection of metrics.Following the convergence of ML methodology across application domains, MetricsReloaded fosters the convergence of validation methodology. The framework wasdeveloped in a multi-stage Delphi process and is based on the novel concept ofa problem fingerprint - a structured representation of the given problem thatcaptures all aspects that are relevant for metric selection, from the domaininterest to the properties of the target structure(s), data set and algorithmoutput. Based on the problem fingerprint, users are guided through the processof choosing and applying appropriate validation metrics while being made awareof potential pitfalls. Metrics Reloaded targets image analysis problems thatcan be interpreted as a classification task at image, object or pixel level,namely image-level classification, object detection, semantic segmentation, andinstance segmentation tasks. To improve the user experience, we implemented theframework in the Metrics Reloaded online tool, which also provides a point ofaccess to explore weaknesses, strengths and specific recommendations for themost common validation metrics. The broad applicability of our framework acrossdomains is demonstrated by an instantiation for various biological and medicalimage analysis use cases.
Bernhardt M, Jones C, Glocker B, 2022, Potential sources of dataset bias complicate investigation of underdiagnosis by machine learning algorithms, NATURE MEDICINE, Vol: 28, Pages: 1157-+, ISSN: 1078-8956
- Author Web Link
- Cite
- Citations: 7
Liu X, Glocker B, McCradden MM, et al., 2022, The medical algorithmic audit (vol 4, pg e384, 2022), LANCET DIGITAL HEALTH, Vol: 4, Pages: E405-E405
Grzech D, Azampour MF, Qiu H, et al., 2022, Uncertainty quantification in non-rigid image registration via stochastic gradient Markov chain Monte Carlo, Publisher: ArXiv
We develop a new Bayesian model for non-rigid registration ofthree-dimensional medical images, with a focus on uncertainty quantification.Probabilistic registration of large images with calibrated uncertaintyestimates is difficult for both computational and modelling reasons. To addressthe computational issues, we explore connections between the Markov chain MonteCarlo by backpropagation and the variational inference by backpropagationframeworks, in order to efficiently draw samples from the posteriordistribution of transformation parameters. To address the modelling issues, weformulate a Bayesian model for image registration that overcomes the existingbarriers when using a dense, high-dimensional, and diffeomorphic transformationparametrisation. This results in improved calibration of uncertainty estimates.We compare the model in terms of both image registration accuracy anduncertainty quantification to VoxelMorph, a state-of-the-art image registrationmodel based on deep learning.
Liu X, Glocker B, McCradden MM, et al., 2022, The medical algorithmic audit., The Lancet Digital Health, Vol: 4, Pages: e384-e397, ISSN: 2589-7500
Artificial intelligence systems for health care, like any other medical device, have the potential to fail. However, specific qualities of artificial intelligence systems, such as the tendency to learn spurious correlates in training data, poor generalisability to new deployment settings, and a paucity of reliable explainability mechanisms, mean they can yield unpredictable errors that might be entirely missed without proactive investigation. We propose a medical algorithmic audit framework that guides the auditor through a process of considering potential algorithmic errors in the context of a clinical task, mapping the components that might contribute to the occurrence of errors, and anticipating their potential consequences. We suggest several approaches for testing algorithmic errors, including exploratory error analysis, subgroup testing, and adversarial testing, and provide examples from our own work and previous studies. The medical algorithmic audit is a tool that can be used to better understand the weaknesses of an artificial intelligence system and put in place mechanisms to mitigate their impact. We propose that safety monitoring and medical algorithmic auditing should be a joint responsibility between users and developers, and encourage the use of feedback mechanisms between these groups to promote learning and maintain safe deployment of artificial intelligence systems.
Pati S, Baid U, Edwards B, et al., 2022, Federated learning enables big data for rare cancer boundary detection, Publisher: arXiv
Although machine learning (ML) has shown promise in numerous domains, there are concerns about generalizability to out-of-sample data. This is currently addressed by centrally sharing ample, and importantly diverse, data from multiple sites. However, such centralization is challenging to scale (or even not feasible) due to various limitations. Federated ML (FL) provides an alternative to train accurate and generalizable ML models, by only sharing numerical model updates. Here we present findings from the largest FL study to-date, involving data from 71 healthcare institutions across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, utilizing the largest dataset of such patients ever used in the literature (25,256 MRI scans from 6,314 patients). We demonstrate a 33% improvement over a publicly trained model to delineate the surgically targetable tumor, and 23% improvement over the tumor's entire extent. We anticipate our study to: 1) enable more studies in healthcare informed by large and diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further quantitative analyses for glioblastoma via performance optimization of our consensus model for eventual public release, and 3) demonstrate the effectiveness of FL at such scale and task complexity as a paradigm shift for multi-site collaborations, alleviating the need for data sharing.
Dou Q, So TY, Jiang M, et al., 2022, Author Correction: Federated deep learning for detecting COVID-19 lung abnormalities in CT: a privacy-preserving multinational validation study, npj Digital Medicine, Vol: 5, ISSN: 2398-6352
Correction to: npj Digital Medicine https://doi.org/10.1038/s41746-021-00431-6, published online 29 March 2021
Popescu SG, Sharp DJ, Cole JH, et al., 2022, Distributional Gaussian Processes Layers for Out-of-Distribution Detection, Journal of Machine Learning for Biomedical Imaging
Machine learning models deployed on medical imaging tasks must be equippedwith out-of-distribution detection capabilities in order to avoid erroneouspredictions. It is unsure whether out-of-distribution detection models relianton deep neural networks are suitable for detecting domain shifts in medicalimaging. Gaussian Processes can reliably separate in-distribution data pointsfrom out-of-distribution data points via their mathematical construction.Hence, we propose a parameter efficient Bayesian layer for hierarchicalconvolutional Gaussian Processes that incorporates Gaussian Processes operatingin Wasserstein-2 space to reliably propagate uncertainty. This directlyreplaces convolving Gaussian Processes with a distance-preserving affineoperator on distributions. Our experiments on brain tissue-segmentation showthat the resulting architecture approaches the performance of well-establisheddeterministic segmentation algorithms (U-Net), which has not been achieved withprevious hierarchical Gaussian Processes. Moreover, by applying the samesegmentation model to out-of-distribution data (i.e., images with pathologysuch as brain tumors), we show that our uncertainty estimates result inout-of-distribution detection that outperforms the capabilities of previousBayesian networks and reconstruction-based approaches that learn normativedistributions. To facilitate future work our code is publicly available.
Bernhardt M, Castro DC, Tanno R, et al., 2022, Active label cleaning for improved dataset quality under resource constraints, NATURE COMMUNICATIONS, Vol: 13
- Author Web Link
- Cite
- Citations: 7
Bernhardt M, Jones C, Glocker B, 2022, Potential sources of dataset bias complicate investigation of underdiagnosis by machine learning algorithms
An increasing number of reports raise concerns about the risk that machinelearning algorithms could amplify health disparities due to biases embedded inthe training data. Seyyed-Kalantari et al. find that models trained on threechest X-ray datasets yield disparities in false-positive rates (FPR) acrosssubgroups on the 'no-finding' label (indicating the absence of disease). Themodels consistently yield higher FPR on subgroups known to be historicallyunderserved, and the study concludes that the models exhibit and potentiallyeven amplify systematic underdiagnosis. We argue that the experimental setup inthe study is insufficient to study algorithmic underdiagnosis. In the absenceof specific knowledge (or assumptions) about the extent and nature of thedataset bias, it is difficult to investigate model bias. Importantly, their useof test data exhibiting the same bias as the training data (due to randomsplitting) severely complicates the interpretation of the reported disparities.
Kattau M, Scienti OP, Glocker B, et al., 2022, Deep Learning-based Tumour Delineation on Photon-counting CT Images
Treatment planning for stereotactic radiosurgery is based on manual delineation of tumour volumes and image registration of the CT and MRI patient scan, which can lead to uncertainties in the clinical workflow. These could be alleviated by implementing deep learning-based tumour contouring and a single modality imaging technique. This study investigates the performance of the deep learning models DeepMedic and nnU-Net for automatic tumour segmentation on two different data sets. In a first experiment, both models are trained on MRI data for vestibular schwannoma segmentation. A second experiment assesses the performance of a 2D nnU-Net model on a glioma data set including MRI and CT data. Additionally, the simulation framework GATE is used to simulate and reconstruct energy-integrating and photon-counting images of a simple brain phantom including a tumour volume. The 2D nnU-Net trained on the glioma CT data is applied to these images.All of the investigated networks achieve a high performance for vestibular schwannoma segmentation on MRI data. However, the performance achieved of the 2D nnU-Net model on the glioma data is low. A possible reason could be the heterogeneity of the data set. Investigation of this is part of ongoing work.The photon-counting image with a 50 keV threshold shows the best contrast to differentiate between tumour and brain tissue for our phantom. However, the 2D nnU-Net is not able to delineate the tumour on any of the reconstructed images which can be explained by the lack of higher-level brain features in the phantom.The results of this work show the need of incorporating a more realistic brain anatomy into the system and improvement of the network training and image reconstruction pipeline.
Rosnati M, De Sousa Ribeiro F, Monteiro M, et al., 2022, Analysing the effectiveness of a generative model for semi-supervised medical image segmentation
Image segmentation is important in medical imaging, providing valuable, quantitative information for clinical decision-making in diagnosis, therapy, and intervention. The state-of-The-Art in automated segmentation remains supervised learning, employing discriminative models such as U-Net. However, training these models requires access to large amounts of manually labelled data which is often difficult to obtain in real medical applications. In such settings, semi-supervised learning (SSL) attempts to leverage the abundance of unlabelled data to obtain more robust and reliable models. Recently, generative models have been proposed for semantic segmentation, as they make an attractive choice for SSL. Their ability to capture the joint distribution over input images and output label maps provides a natural way to incorporate information from unlabelled images. This paper analyses whether deep generative models such as the SemanticGAN are truly viable alternatives to tackle challenging medical image segmentation problems. To that end, we thoroughly evaluate the segmentation performance, robustness, and potential subgroup disparities of discriminative and generative segmentation methods when applied to large-scale, publicly available chest Xray datasets.
Shehata N, Bain W, Glocker B, 2022, A Comparative Study of Graph Neural Networks for Shape Classification in Neuroimaging, Pages: 160-171
Graph neural networks have emerged as a promising approach for the analysis of non-Euclidean data such as meshes. In medical imaging, mesh-like data plays an important role for modelling anatomical structures, and shape classification can be used in computer aided diagnosis and disease detection. However, with a plethora of options, the best architectural choices for medical shape analysis using GNNs remain unclear. We conduct a comparative analysis to provide practitioners with an overview of the current state-of-The-Art in geometric deep learning for shape classification in neuroimaging. Using biological sex classification as a proof-of-concept task, we find that using FPFH as node features substantially improves GNN performance and generalisation to out-of-distribution data; we compare the performance of three alternative convolutional layers; and we reinforce the importance of data augmentation for graph based learning. We then confirm these results hold for a clinically relevant task, using the classification of Alzheimer's disease.
Dorent R, Kujawa A, Ivory M, et al., 2022, CrossMoDA 2021 challenge: Benchmark of Cross-Modality Domain Adaptation techniques for Vestibular Schwnannoma and Cochlea Segmentation., CoRR, Vol: abs/2201.02831
Grzech D, Azampour MF, Glocker B, et al., 2022, A variational Bayesian method for similarity learning in non-rigid image registration, IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Publisher: IEEE COMPUTER SOC, Pages: 119-128, ISSN: 1063-6919
- Author Web Link
- Cite
- Citations: 1
Bernhardt M, Jones C, Glocker B, 2022, Investigating underdiagnosis of AI algorithms in the presence of multiple sources of dataset bias., CoRR, Vol: abs/2201.07856
Bernhardt M, Ribeiro FDS, Glocker B, 2022, Failure detection in medical image classification: a reality check and benchmarking testbed, Transactions on Machine Learning Research, Vol: 2022, ISSN: 2835-8856
Failure detection in automated image classification is a critical safeguard for clinical deployment. Detected failure cases can be referred to human assessment, ensuring patient safety in computer-aided clinical decision making. Despite its paramount importance, there is insufficient evidence about the ability of state-of-the-art confidence scoring methods to detect test-time failures of classification models in the context of medical imaging. This paper provides a reality check, establishing the performance of in-domain misclassification detection methods, benchmarking 9 widely used confidence scores on 6 medical imaging datasets with different imaging modalities, in multiclass and binary classification settings. Our experiments show that the problem of failure detection is far from being solved. We found that none of the benchmarked advanced methods proposed in the computer vision and machine learning literature can consistently outperform a simple softmax baseline, demonstrating that improved out-of-distribution detection or model calibration do not necessarily translate to improved in-domain misclassification detection. Our developed testbed facilitates future work in this important area.
Langley J, Monteiro M, Jones C, et al., 2022, Structured uncertainty in the observation space of variational autoencoders, Transactions on Machine Learning Research, Vol: 2022, ISSN: 2835-8856
Variational autoencoders (VAEs) are a popular class of deep generative models with many variants and a wide range of applications. Improvements upon the standard VAE mostly focus on the modelling of the posterior distribution over the latent space and the properties of the neural network decoder. In contrast, improving the model for the observational distribution is rarely considered and typically defaults to a pixel-wise independent categorical or normal distribution. In image synthesis, sampling from such distributions produces spatially-incoherent results with uncorrelated pixel noise, resulting in only the sample mean being somewhat useful as an output prediction. In this paper, we aim to stay true to VAE theory by improving the samples from the observational distribution. We propose SOS-VAE, an alternative model for the observation space, encoding spatial dependencies via a low-rank parameterisation. We demonstrate that this new observational distribution has the ability to capture relevant covariance between pixels, resulting in spatially-coherent samples. In contrast to pixel-wise independent distributions, our samples seem to contain semantically-meaningful variations from the mean allowing the prediction of multiple plausible outputs with a single forward pass.
Santhirasekaram A, Kori A, Winkler M, et al., 2022, Vector Quantisation for Robust Segmentation., CoRR, Vol: abs/2207.01919
Li Z, Kamnitsas K, Islam M, et al., 2022, Estimating Model Performance under Domain Shifts with Class-Specific Confidence Scores., CoRR, Vol: abs/2207.09957
Li Z, Kamnitsas K, Ouyang C, et al., 2022, Context Label Learning: Improving Background Class Representations in Semantic Segmentation., CoRR, Vol: abs/2212.08423
Popescu SG, Sharp DJ, Cole JH, et al., 2022, Distributional Gaussian Processes Layers for Out-of-Distribution Detection., CoRR, Vol: abs/2206.13346
Rosnati M, Ribeiro FDS, Monteiro M, et al., 2022, Analysing the effectiveness of a generative model for semi-supervised medical image segmentation., CoRR, Vol: abs/2211.01886
Sinclair M, Schuh A, Hahn K, et al., 2022, Atlas-ISTN: Joint segmentation, registration and atlas construction with image-and-spatial transformer networks., Medical Image Anal., Vol: 78, Pages: 102383-102383
Li Z, Kamnitsas K, Islam M, et al., 2022, Estimating Model Performance Under Domain Shifts with Class-Specific Confidence Scores., Publisher: Springer, Pages: 693-703
Santhirasekaram A, Kori A, Winkler M, et al., 2022, Vector Quantisation for Robust Segmentation., Publisher: Springer, Pages: 663-672
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.