RESUMO
Background: Inaccurate blood pressure (BP) classification results in inappropriate treatment. We tested whether machine learning (ML), using routine clinical data, can serve as a reliable alternative to ambulatory BP monitoring (ABPM) in classifying BP status. Methods: This study employed a multicentre approach involving 3 derivation cohorts from Glasgow, Gdansk, and Birmingham, and a fourth independent evaluation cohort. ML models were trained using office BP, ABPM, and clinical, laboratory, and demographic data, collected from patients referred for hypertension assessment. Seven ML algorithms were trained to classify patients into 5 groups, named as follows: Normal/Target; Hypertension-Masked; Normal/Target-White-Coat (WC); Hypertension-WC; and Hypertension. The 10-year cardiovascular outcomes and 27-year all-cause mortality risks were calculated for the ML-derived groups using the Cox proportional hazards model. Results: Overall, extreme gradient boosting (using XGBoost open source software) showed the highest area under the receiver operating characteristic curve of 0.85-0.88 across derivation cohorts, Glasgow (n = 923; 43% female; age 50.7 ± 16.3 years), Gdansk (n = 709; 46% female; age 54.4 ± 13 years), and Birmingham (n = 1222; 56% female; age 55.7 ± 14 years). But accuracy (0.57-0.72) and F1 (harmonic mean of precision and recall) scores (0.57-0.69) were low across the 3 patient cohorts. The evaluation cohort (n = 6213; 51% female; age 51.2 ± 10.8 years) indicated elevated 10-year risks of composite cardiovascular events in the Normal/Target-WC and the Hypertension-WC groups, with heightened 27-year all-cause mortality observed in all groups, except the Hypertension-Masked group, compared to the Normal/Target group. Conclusions: ML has limited potential in accurate BP classification when ABPM is unavailable. Larger studies including diverse patient groups and different resource settings are warranted.
Contexte: Les erreurs dans la classification des valeurs de la pression artérielle (PA) entraînent une inadéquation du traitement. Nous avons tâché de déterminer si l'apprentissage machine, à l'aide de données cliniques routinières, constituait une solution de rechange fiable à la surveillance ambulatoire de la PA pour définir le statut de la PA. Méthodologie: Cette étude a utilisé une approche multicentrique incluant trois cohortes de dérivation de Glasgow, Gdansk et Birmingham, et une quatrième cohorte d'évaluation indépendante. Les modèles d'apprentissage machine ont été développés en analysant les données démographiques, les valeurs de la PA mesurée au cabinet, les données relatives à la surveillance ambulatoire de la PA et aux épreuves de laboratoire recueillies auprès de patients adressés pour une évaluation de l'hypertension. Sept algorithmes d'apprentissage machine ont été appliqués pour classer les patients en cinq groupes : Normale/Cible; Hypertension-Masquée; Normal/Cible-Blouse blanche; Hypertension-Blouse blanche; Hypertension. Les événements cardiovasculaires sur 10 ans et le risque de mortalité toutes causes confondues sur 27 ans ont été calculés dans les groupes dérivés de l'apprentissage machine à l'aide d'un modèle de risques proportionnels de Cox. Résultats: D'une manière générale, l'amplification de gradient extrême (à l'aide du logiciel ouvert XGBoost) a mis en évidence l'aire sous la courbe de la fonction d'efficacité du récepteur (courbe ROC pour Receiver Operating Characteristic) la plus haute, soit 0,85 à 0,88, pour toutes les cohortes de dérivation : Glasgow (n = 923; 43 % de femmes; âge : 50,7 ± 16,3 ans); Gdansk (n = 709; 46 % de femmes; âge : 54,4 ± 13 ans); Birmingham (n = 1 222; 56 % de femmes; âge : 55,7 ± 14 ans). La précision (0,57 0,72) et le score F1 (moyenne harmonique de la précision et du rappel) (0,57 0,69) ont été faibles dans les trois cohortes de patients. La cohorte d'évaluation (n = 6 213; 51 % de femmes; âge : 51,2 ± 10,8 ans) a indiqué un risque d'événements cardiovasculaires composites sur 10 ans élevé dans les groupes Normale/Cible-Blouse blanche et Hypertension-Blouse blanche, tandis qu'une hausse de la mortalité toutes causes confondues sur 27 ans a été observée dans tous les groupes, sauf dans le groupe Hypertension-Masquée, comparativement au groupe Normale/Cible. Conclusions: Le potentiel d'exactitude de la classification de la PA à l'aide de l'apprentissage machine lorsque la surveillance ambulatoire de la PA n'est pas possible est limité. Des études de plus grande envergure portant sur des groupes de patients et des niveaux de ressources diversifiés s'imposent.
RESUMO
Deep learning models often need sufficient supervision (i.e., labelled data) in order to be trained effectively. By contrast, humans can swiftly learn to identify important anatomy in medical images like MRI and CT scans, with minimal guidance. This recognition capability easily generalises to new images from different medical facilities and to new tasks in different settings. This rapid and generalisable learning ability is largely due to the compositional structure of image patterns in the human brain, which are not well represented in current medical models. In this paper, we study the utilisation of compositionality in learning more interpretable and generalisable representations for medical image segmentation. Overall, we propose that the underlying generative factors that are used to generate the medical images satisfy compositional equivariance property, where each factor is compositional (e.g., corresponds to human anatomy) and also equivariant to the task. Hence, a good representation that approximates well the ground truth factor has to be compositionally equivariant. By modelling the compositional representations with learnable von-Mises-Fisher (vMF) kernels, we explore how different design and learning biases can be used to enforce the representations to be more compositionally equivariant under un-, weakly-, and semi-supervised settings. Extensive results show that our methods achieve the best performance over several strong baselines on the task of semi-supervised domain-generalised medical image segmentation. Code will be made publicly available upon acceptance at https://github.com/vios-s.
Assuntos
Algoritmos , Encéfalo , Aprendizado Profundo , Processamento de Imagem Assistida por Computador , Imageamento por Ressonância Magnética , Humanos , Encéfalo/diagnóstico por imagem , Imageamento por Ressonância Magnética/métodos , Processamento de Imagem Assistida por Computador/métodos , Tomografia Computadorizada por Raios X/métodosRESUMO
Background: Natural language processing (NLP) has the potential to automate the reading of radiology reports, but there is a need to demonstrate that NLP methods are adaptable and reliable for use in real-world clinical applications. Methods: We tested the F1 score, precision, and recall to compare NLP tools on a cohort from a study on delirium using images and radiology reports from NHS Fife and a population-based cohort (Generation Scotland) that spans multiple National Health Service health boards. We compared four off-the-shelf rule-based and neural NLP tools (namely, EdIE-R, ALARM+, ESPRESSO, and Sem-EHR) and reported on their performance for three cerebrovascular phenotypes, namely, ischaemic stroke, small vessel disease (SVD), and atrophy. Clinical experts from the EdIE-R team defined phenotypes using labelling techniques developed in the development of EdIE-R, in conjunction with an expert researcher who read underlying images. Results: EdIE-R obtained the highest F1 score in both cohorts for ischaemic stroke, ≥93%, followed by ALARM+, ≥87%. The F1 score of ESPRESSO was ≥74%, whilst that of Sem-EHR is ≥66%, although ESPRESSO had the highest precision in both cohorts, 90% and 98%. For F1 scores for SVD, EdIE-R scored ≥98% and ALARM+ ≥90%. ESPRESSO scored lowest with ≥77% and Sem-EHR ≥81%. In NHS Fife, F1 scores for atrophy by EdIE-R and ALARM+ were 99%, dropping in Generation Scotland to 96% for EdIE-R and 91% for ALARM+. Sem-EHR performed lowest for atrophy at 89% in NHS Fife and 73% in Generation Scotland. When comparing NLP tool output with brain image reads using F1 scores, ALARM+ scored 80%, outperforming EdIE-R at 66% in ischaemic stroke. For SVD, EdIE-R performed best, scoring 84%, with Sem-EHR 82%. For atrophy, EdIE-R and both ALARM+ versions were comparable at 80%. Conclusions: The four NLP tools show varying F1 (and precision/recall) scores across all three phenotypes, although more apparent for ischaemic stroke. If NLP tools are to be used in clinical settings, this cannot be performed "out of the box." It is essential to understand the context of their development to assess whether they are suitable for the task at hand or whether further training, re-training, or modification is required to adapt tools to the target task.
RESUMO
Pathological brain lesions exhibit diverse appearance in brain images, in terms of intensity, texture, shape, size, and location. Comprehensive sets of data and annotations are difficult to acquire. Therefore, unsupervised anomaly detection approaches have been proposed using only normal data for training, with the aim of detecting outlier anomalous voxels at test time. Denoising methods, for instance classical denoising autoencoders (DAEs) and more recently emerging diffusion models, are a promising approach, however naive application of pixelwise noise leads to poor anomaly detection performance. We show that optimization of the spatial resolution and magnitude of the noise improves the performance of different model training regimes, with similar noise parameter adjustments giving good performance for both DAEs and diffusion models. Visual inspection of the reconstructions suggests that the training noise influences the trade-off between the extent of the detail that is reconstructed and the extent of erasure of anomalies, both of which contribute to better anomaly detection performance. We validate our findings on two real-world datasets (tumor detection in brain MRI and hemorrhage/ischemia/tumor detection in brain CT), showing good detection on diverse anomaly appearances. Overall, we find that a DAE trained with coarse noise is a fast and simple method that gives state-of-the-art accuracy. Diffusion models applied to anomaly detection are as yet in their infancy and provide a promising avenue for further research. Code for our DAE model and coarse noise is provided at: https://github.com/AntanasKascenas/DenoisingAE.
RESUMO
Introduction: Thrombolysis treatment for acute ischaemic stroke can lead to better outcomes if administered early enough. However, contraindications exist which put the patient at greater risk of a bleed (e.g. recent major surgery, anticoagulant medication). Therefore, clinicians must check a patient's past medical history before proceeding with treatment. In this work we present a machine learning approach for accurate automatic detection of this information in unstructured text documents such as discharge letters or referral letters, to support the clinician in making a decision about whether to administer thrombolysis. Methods: We consulted local and national guidelines for thrombolysis eligibility, identifying 86 entities which are relevant to the thrombolysis decision. A total of 8,067 documents from 2,912 patients were manually annotated with these entities by medical students and clinicians. Using this data, we trained and validated several transformer-based named entity recognition (NER) models, focusing on transformer models which have been pre-trained on a biomedical corpus as these have shown most promise in the biomedical NER literature. Results: Our best model was a PubMedBERT-based approach, which obtained a lenient micro/macro F1 score of 0.829/0.723. Ensembling 5 variants of this model gave a significant boost to precision, obtaining micro/macro F1 of 0.846/0.734 which approaches the human annotator performance of 0.847/0.839. We further propose numeric definitions for the concepts of name regularity (similarity of all spans which refer to an entity) and context regularity (similarity of all context surrounding mentions of an entity), using these to analyse the types of errors made by the system and finding that the name regularity of an entity is a stronger predictor of model performance than raw training set frequency. Discussion: Overall, this work shows the potential of machine learning to provide clinical decision support (CDS) for the time-critical decision of thrombolysis administration in ischaemic stroke by quickly surfacing relevant information, leading to prompt treatment and hence to better patient outcomes.
RESUMO
Leakage of data from publicly available Machine Learning (ML) models is an area of growing significance since commercial and government applications of ML can draw on multiple sources of data, potentially including users' and clients' sensitive data. We provide a comprehensive survey of contemporary advances on several fronts, covering involuntary data leakage which is natural to ML models, potential malicious leakage which is caused by privacy attacks, and currently available defence mechanisms. We focus on inference-time leakage, as the most likely scenario for publicly available models. We first discuss what leakage is in the context of different data, tasks, and model architectures. We then propose a taxonomy across involuntary and malicious leakage, followed by description of currently available defences, assessment metrics, and applications. We conclude with outstanding challenges and open questions, outlining some promising directions for future research.
RESUMO
BACKGROUND: Information retrieval (IR) from the free text within electronic health records (EHRs) is time consuming and complex. We hypothesize that natural language processing (NLP)-enhanced search functionality for EHRs can make clinical workflows more efficient and reduce cognitive load for clinicians. OBJECTIVE: This study aimed to evaluate the efficacy of 3 levels of search functionality (no search, string search, and NLP-enhanced search) in supporting IR for clinical users from the free text of EHR documents in a simulated clinical environment. METHODS: A clinical environment was simulated by uploading 3 sets of patient notes into an EHR research software application and presenting these alongside 3 corresponding IR tasks. Tasks contained a mixture of multiple-choice and free-text questions. A prospective crossover study design was used, for which 3 groups of evaluators were recruited, which comprised doctors (n=19) and medical students (n=16). Evaluators performed the 3 tasks using each of the search functionalities in an order in accordance with their randomly assigned group. The speed and accuracy of task completion were measured and analyzed, and user perceptions of NLP-enhanced search were reviewed in a feedback survey. RESULTS: NLP-enhanced search facilitated more accurate task completion than both string search (5.14%; P=.02) and no search (5.13%; P=.08). NLP-enhanced search and string search facilitated similar task speeds, both showing an increase in speed compared to the no search function, by 11.5% (P=.008) and 16.0% (P=.007) respectively. Overall, 93% of evaluators agreed that NLP-enhanced search would make clinical workflows more efficient than string search, with qualitative feedback reporting that NLP-enhanced search reduced cognitive load. CONCLUSIONS: To the best of our knowledge, this study is the largest evaluation to date of different search functionalities for supporting target clinical users in realistic clinical workflows, with a 3-way prospective crossover study design. NLP-enhanced search improved both accuracy and speed of clinical EHR IR tasks compared to browsing clinical notes without search. NLP-enhanced search improved accuracy and reduced the number of searches required for clinical EHR IR tasks compared to direct search term matching.
RESUMO
Causal machine learning (CML) has experienced increasing popularity in healthcare. Beyond the inherent capabilities of adding domain knowledge into learning systems, CML provides a complete toolset for investigating how a system would react to an intervention (e.g. outcome given a treatment). Quantifying effects of interventions allows actionable decisions to be made while maintaining robustness in the presence of confounders. Here, we explore how causal inference can be incorporated into different aspects of clinical decision support systems by using recent advances in machine learning. Throughout this paper, we use Alzheimer's disease to create examples for illustrating how CML can be advantageous in clinical scenarios. Furthermore, we discuss important challenges present in healthcare applications such as processing high-dimensional and unstructured data, generalization to out-of-distribution samples and temporal relationships, that despite the great effort from the research community remain to be solved. Finally, we review lines of research within causal representation learning, causal discovery and causal reasoning which offer the potential towards addressing the aforementioned challenges.
RESUMO
Disentangled representation learning has been proposed as an approach to learning general representations even in the absence of, or with limited, supervision. A good general representation can be fine-tuned for new target tasks using modest amounts of data, or used directly in unseen domains achieving remarkable performance in the corresponding task. This alleviation of the data and annotation requirements offers tantalising prospects for applications in computer vision and healthcare. In this tutorial paper, we motivate the need for disentangled representations, revisit key concepts, and describe practical building blocks and criteria for learning such representations. We survey applications in medical imaging emphasising choices made in exemplar key works, and then discuss links to computer vision applications. We conclude by presenting limitations, challenges, and opportunities.
Assuntos
Aprendizagem , Aprendizado de Máquina , Humanos , SoftwareRESUMO
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.