Deep learning-based image quality assessment for optical coherence tomography macular scans: a multicentre study.

AIMS: To develop and externally test deep learning (DL) models for assessing the image quality of three-dimensional (3D) macular scans from Cirrus and Spectralis optical coherence tomography devices. METHODS: We retrospectively collected two data sets including 2277 Cirrus 3D scans and 1557 Spectralis 3D scans, respectively, for training (70%), fine-tuning (10%) and internal validation (20%) from electronic medical and research records at The Chinese University of Hong Kong Eye Centre and the Hong Kong Eye Hospital. Scans with various eye diseases (eg, diabetic macular oedema, age-related macular degeneration, polypoidal choroidal vasculopathy and pathological myopia), and scans of normal eyes from adults and children were included. Two graders labelled each 3D scan as gradable or ungradable, according to standardised criteria. We used a 3D version of the residual network (ResNet)-18 for Cirrus 3D scans and a multiple-instance learning pipline with ResNet-18 for Spectralis 3D scans. Two deep learning (DL) models were further tested via three unseen Cirrus data sets from Singapore and five unseen Spectralis data sets from India, Australia and Hong Kong, respectively. RESULTS: In the internal validation, the models achieved the area under curves (AUCs) of 0.930 (0.885-0.976) and 0.906 (0.863-0.948) for assessing the Cirrus 3D scans and Spectralis 3D scans, respectively. In the external testing, the models showed robust performance with AUCs ranging from 0.832 (0.730-0.934) to 0.930 (0.906-0.953) and 0.891 (0.836-0.945) to 0.962 (0.918-1.000), respectively. CONCLUSIONS: Our models could be used for filtering out ungradable 3D scans and further incorporated with a disease-detection DL model, allowing a fully automated eye disease detection workflow.

Association between resting myocardial work indices and stress myocardial perfusion in patients with angina and non-obstructive coronary artery disease.

Background: Myocardial work (MW) indices and longitudinal strain (LS) are sensitive markers of early left ventricular systolic dysfunction. Stress computed tomography myocardial perfusion imaging (CT-MPI) can assess early myocardial ischemia. The association between resting MW indices and stress myocardial perfusion remains unclear. This study compares resting MW indices with LS to assess stress myocardial perfusion in angina patients with non-obstructive coronary artery disease (CAD). Methods: Eighty-four patients who underwent resting echocardiography, coronary computed tomography angiography, and stress CT-MPI were reviewed. Seventeen myocardial segments were divided into three regions according to the epicardial coronary arteries. Global indices included global longitudinal strain (GLS), global work index (GWI), global constructive work (GCW), global wasted work (GWW), and global work efficiency (GWE). Regional indices included regional longitudinal strain (RLS), regional work index (RWI), and regional work efficiency (RWE). Reduced global perfusion was defined as an average stress myocardial blood flow (MBF) <116 mL/100 mL/min for the whole heart. Reduced regional perfusion was defined as an average stress MBF <116 mL/100 mL/min for the coronary territories. No patients demonstrated obstructions in the epicardial coronary arteries (stenosis diameter <50%). The MW indices and LS were compared. Receiver operating characteristic curves were constructed and logistic regression analyses were used to investigate the predictors of reduced myocardial perfusion. Results: Patients with reduced stress perfusion demonstrated reduced GLS, GWI, GCW, and GWE (P<0.05) and increased GWW (P<0.05). After adjustment for age and sex, GWE was still independently associated with reduced myocardial perfusion (odds ratio =0.386, 95% confidence interval: 0.214-0.697; P<0.05). Receiver operating characteristic curves reflected the good diagnostic ability of GWE and its superiority to GLS (area under the curve: 0.858 vs. 0.741). The optimal cutoff GWE value was 95% (sensitivity, 70%; specificity, 90%). Regions with lower stress perfusion showed lower RLS, RWI, and RWE (P<0.05). The optimal cutoff value of RWE for predicting reduced regional perfusion was 95%, with an area under the curve of 0.780, a sensitivity of 62%, and a specificity of 83%. Conclusions: Resting MW indices perform well in assessing global and regional stress myocardial perfusion in angina patients with non-obstructive CAD, and GWE is superior to GLS in the global evaluations.

The value of myocardial work in the estimation of left ventricular systolic function in patients with coronary microvascular disease: A study based on adenosine stress echocardiography.

Background: Coronary microvascular dysfunction (CMD) is associated with increased cardiovascular events in patients with angina with non-obstructive coronary (ANOCA), especially heart failure. Conventional echocardiography is difficult to identify early alterations in cardiac function due to CMD. Methods: We recruited 78 ANOCA patients. All patients underwent conventional echocardiography examination, adenosine stress echocardiography and examination of coronary flow reserve (CFR) by transthoracic echocardiography. Based on the CFR results, patients were divided into the CMD group (CFR < 2.5) and the non-CMD group (CFVR ≥ 2.5). Demographic data, conventional echocardiographic parameters, two-dimensional speckle-tracking echocardiography (2D-STE) parameters and myocardial work (MW) were compared between the two groups at rest and at stress. Logistic regression was used to analyze the factors associated with CMD. Results: There was no significant difference in conventional echocardiography parameters, 2D-STE related indices or MW at rest between the two groups. Global work index (GWI), global contractive work (GCW), and global work efficiency (GWE) were lower in the CMD group than in the non-CMD group at stress (p = 0.040, 0.044, <0.001, respectively), but global waste work (GWW) and peak strain dispersion (PSD) were higher (both p < 0.001). GWI and GCW were associated with systolic blood pressure, diastolic blood pressure, product of heart rate and blood pressure, GLS and coronary flow velocity. While GWW was mainly correlated with PSD, GWE was correlated with PSD and GLS. In the non-CMD group, the responses to adenosine was mainly manifested as an increase in GWI, GCW and GWE (p = 0.001, 0.001, 0.009, respectively) and a decrease in PSD and GWW (p = 0.001, 0.015, respectively). In the CMD group, the response to adenosine was mainly manifested as an increase in GWW and a decrease in GWE (p = 0.002, and 0.006, respectively). In the multivariate regression analysis, we found that ΔGWW (difference in GWW before vs. after adenosine stress) and ΔPSD (difference in PSD before vs. after adenosine stress) were independent factors associated with CMD. The ROC curves showed that the composite prediction model consisting of ΔGWW and ΔPSD had excellent diagnostic value for CMD (area under the curve = 0.913). Conclusion: In the present study, we found that CMD caused deterioration of myocardial work in ANOCA patients under adenosine stress, and that increased cardiac contraction asynchrony and wasted work may be the main changes caused by CMD.

DLTTA: Dynamic Learning Rate for Test-Time Adaptation on Cross-Domain Medical Images.

Test-time adaptation (TTA) has increasingly been an important topic to efficiently tackle the cross-domain distribution shift at test time for medical images from different institutions. Previous TTA methods have a common limitation of using a fixed learning rate for all the test samples. Such a practice would be sub-optimal for TTA, because test data may arrive sequentially therefore the scale of distribution shift would change frequently. To address this problem, we propose a novel dynamic learning rate adjustment method for test-time adaptation, called DLTTA, which dynamically modulates the amount of weights update for each test image to account for the differences in their distribution shift. Specifically, our DLTTA is equipped with a memory bank based estimation scheme to effectively measure the discrepancy of a given test sample. Based on this estimated discrepancy, a dynamic learning rate adjustment strategy is then developed to achieve a suitable degree of adaptation for each test sample. The effectiveness and general applicability of our DLTTA is extensively demonstrated on three tasks including retinal optical coherence tomography (OCT) segmentation, histopathological image classification, and prostate 3D MRI segmentation. Our method achieves effective and fast test-time adaptation with consistent performance improvement over current state-of-the-art test-time adaptation methods. Code is available at https://github.com/med-air/DLTTA.

Learning With Privileged Multimodal Knowledge for Unimodal Segmentation.

Multimodal learning usually requires a complete set of modalities during inference to maintain performance. Although training data can be well-prepared with high-quality multiple modalities, in many cases of clinical practice, only one modality can be acquired and important clinical evaluations have to be made based on the limited single modality information. In this work, we propose a privileged knowledge learning framework with the 'Teacher-Student' architecture, in which the complete multimodal knowledge that is only available in the training data (called privileged information) is transferred from a multimodal teacher network to a unimodal student network, via both a pixel-level and an image-level distillation scheme. Specifically, for the pixel-level distillation, we introduce a regularized knowledge distillation loss which encourages the student to mimic the teacher's softened outputs in a pixel-wise manner and incorporates a regularization factor to reduce the effect of incorrect predictions from the teacher. For the image-level distillation, we propose a contrastive knowledge distillation loss which encodes image-level structured information to enrich the knowledge encoding in combination with the pixel-level distillation. We extensively evaluate our method on two different multi-class segmentation tasks, i.e., cardiac substructure segmentation and brain tumor segmentation. Experimental results on both tasks demonstrate that our privileged knowledge learning is effective in improving unimodal segmentation and outperforms previous methods.

Impact of blood pressure changes on myocardial work indices in hypertensive patients in a day.

Evaluating left ventricular function through instantaneous left ventricular deformation parameters might not always be accurate for patients with high fluctuations in blood pressure value due to afterload dependence. Myocardial work (MW) is a more advanced tool that combines global myocardial longitudinal strain (GLS) with LV (left ventricular) systolic pressure. The purpose of this study was to investigate the effect of blood pressure changes on MW indices in the population with normal blood pressure and hypertension in a day. A total of 117 participants (34 control subjects and 83 hypertensive patients) underwent echocardiographic measurements at rest, twice a day. Simultaneously, the brachial blood pressure was also measured. LV pressure-strain loop (PSL) was used to calculate global work index (GWI), global constructive work (GCW), global wasted work (GWW), and global work efficiency (GWE). The differences in the GLS and MW indices between the groups were compared, and the correlation of blood pressure changes with the changes in GLS and MW indices were evaluated. Compared to the control group, the hypertensive group showed higher GWI, GCW, and GWW but lower GLS and GWE. Absolute changes in blood pressure, GLS, and MW indices in hypertensive patients were significantly higher than that of the control subjects. Blood pressure changes had significant univariate correlation with changes in GLS and MW indices. In conclusion, significant fluctuations in blood pressure could induce changes in MW indices to preserve left ventricular systolic function. Repeated assessment of MW indices is necessary for hypertensive patients with large blood pressure fluctuations.

Federated deep learning for detecting COVID-19 lung abnormalities in CT: a privacy-preserving multinational validation study.

Data privacy mechanisms are essential for rapidly scaling medical training databases to capture the heterogeneity of patient data distributions toward robust and generalizable machine learning systems. In the current COVID-19 pandemic, a major focus of artificial intelligence (AI) is interpreting chest CT, which can be readily used in the assessment and management of the disease. This paper demonstrates the feasibility of a federated learning method for detecting COVID-19 related CT abnormalities with external validation on patients from a multinational study. We recruited 132 patients from seven multinational different centers, with three internal hospitals from Hong Kong for training and testing, and four external, independent datasets from Mainland China and Germany, for validating model generalizability. We also conducted case studies on longitudinal scans for automated estimation of lesion burden for hospitalized COVID-19 patients. We explore the federated learning algorithms to develop a privacy-preserving AI model for COVID-19 medical image diagnosis with good generalization capability on unseen multinational datasets. Federated learning could provide an effective mechanism during pandemics to rapidly develop clinically useful AI across institutions and countries overcoming the burden of central aggregation of large amounts of sensitive data.

Contrastive Cross-Site Learning With Redesigned Net for COVID-19 CT Classification.

The pandemic of coronavirus disease 2019 (COVID-19) has lead to a global public health crisis spreading hundreds of countries. With the continuous growth of new infections, developing automated tools for COVID-19 identification with CT image is highly desired to assist the clinical diagnosis and reduce the tedious workload of image interpretation. To enlarge the datasets for developing machine learning methods, it is essentially helpful to aggregate the cases from different medical systems for learning robust and generalizable models. This paper proposes a novel joint learning framework to perform accurate COVID-19 identification by effectively learning with heterogeneous datasets with distribution discrepancy. We build a powerful backbone by redesigning the recently proposed COVID-Net in aspects of network architecture and learning strategy to improve the prediction accuracy and learning efficiency. On top of our improved backbone, we further explicitly tackle the cross-site domain shift by conducting separate feature normalization in latent space. Moreover, we propose to use a contrastive training objective to enhance the domain invariance of semantic embeddings for boosting the classification performance on each dataset. We develop and evaluate our method with two public large-scale COVID-19 diagnosis datasets made up of CT images. Extensive experiments show that our approach consistently improves the performanceson both datasets, outperforming the original COVID-Net trained on each dataset by 12.16% and 14.23% in AUC respectively, also exceeding existing state-of-the-art multi-site learning methods.

Semi-Supervised Medical Image Classification With Relation-Driven Self-Ensembling Model.

Training deep neural networks usually requires a large amount of labeled data to obtain good performance. However, in medical image analysis, obtaining high-quality labels for the data is laborious and expensive, as accurately annotating medical images demands expertise knowledge of the clinicians. In this paper, we present a novel relation-driven semi-supervised framework for medical image classification. It is a consistency-based method which exploits the unlabeled data by encouraging the prediction consistency of given input under perturbations, and leverages a self-ensembling model to produce high-quality consistency targets for the unlabeled data. Considering that human diagnosis often refers to previous analogous cases to make reliable decisions, we introduce a novel sample relation consistency (SRC) paradigm to effectively exploit unlabeled data by modeling the relationship information among different samples. Superior to existing consistency-based methods which simply enforce consistency of individual predictions, our framework explicitly enforces the consistency of semantic relation among different samples under perturbations, encouraging the model to explore extra semantic information from unlabeled data. We have conducted extensive experiments to evaluate our method on two public benchmark medical image classification datasets, i.e., skin lesion diagnosis with ISIC 2018 challenge and thorax disease classification with ChestX-ray14. Our method outperforms many state-of-the-art semi-supervised learning methods on both single-label and multi-label image classification scenarios.

Deep Mining External Imperfect Data for Chest X-Ray Disease Screening.

Deep learning approaches have demonstrated remarkable progress in automatic Chest X-ray analysis. The data-driven feature of deep models requires training data to cover a large distribution. Therefore, it is substantial to integrate knowledge from multiple datasets, especially for medical images. However, learning a disease classification model with extra Chest X-ray (CXR) data is yet challenging. Recent researches have demonstrated that performance bottleneck exists in joint training on different CXR datasets, and few made efforts to address the obstacle. In this paper, we argue that incorporating an external CXR dataset leads to imperfect training data, which raises the challenges. Specifically, the imperfect data is in two folds: domain discrepancy, as the image appearances vary across datasets; and label discrepancy, as different datasets are partially labeled. To this end, we formulate the multi-label thoracic disease classification problem as weighted independent binary tasks according to the categories. For common categories shared across domains, we adopt task-specific adversarial training to alleviate the feature differences. For categories existing in a single dataset, we present uncertainty-aware temporal ensembling of model predictions to mine the information from the missing labels further. In this way, our framework simultaneously models and tackles the domain and label discrepancies, enabling superior knowledge mining ability. We conduct extensive experiments on three datasets with more than 360,000 Chest X-ray images. Our method outperforms other competing models and sets state-of-the-art performance on the official NIH test set with 0.8349 AUC, demonstrating its effectiveness of utilizing the external dataset to improve the internal classification.

Unpaired Multi-Modal Segmentation via Knowledge Distillation.

Multi-modal learning is typically performed with network architectures containing modality-specific layers and shared layers, utilizing co-registered images of different modalities. We propose a novel learning scheme for unpaired cross-modality image segmentation, with a highly compact architecture achieving superior segmentation accuracy. In our method, we heavily reuse network parameters, by sharing all convolutional kernels across CT and MRI, and only employ modality-specific internal normalization layers which compute respective statistics. To effectively train such a highly compact model, we introduce a novel loss term inspired by knowledge distillation, by explicitly constraining the KL-divergence of our derived prediction distributions between modalities. We have extensively validated our approach on two multi-class segmentation problems: i) cardiac structure segmentation, and ii) abdominal organ segmentation. Different network settings, i.e., 2D dilated network and 3D U-net, are utilized to investigate our method's general efficacy. Experimental results on both tasks demonstrate that our novel multi-modal learning scheme consistently outperforms single-modal training and previous multi-modal approaches.

MS-Net: Multi-Site Network for Improving Prostate Segmentation With Heterogeneous MRI Data.

Automated prostate segmentation in MRI is highly demanded for computer-assisted diagnosis. Recently, a variety of deep learning methods have achieved remarkable progress in this task, usually relying on large amounts of training data. Due to the nature of scarcity for medical images, it is important to effectively aggregate data from multiple sites for robust model training, to alleviate the insufficiency of single-site samples. However, the prostate MRIs from different sites present heterogeneity due to the differences in scanners and imaging protocols, raising challenges for effective ways of aggregating multi-site data for network training. In this paper, we propose a novel multi-site network (MS-Net) for improving prostate segmentation by learning robust representations, leveraging multiple sources of data. To compensate for the inter-site heterogeneity of different MRI datasets, we develop Domain-Specific Batch Normalization layers in the network backbone, enabling the network to estimate statistics and perform feature normalization for each site separately. Considering the difficulty of capturing the shared knowledge from multiple datasets, a novel learning paradigm, i.e., Multi-site-guided Knowledge Transfer, is proposed to enhance the kernels to extract more generic representations from multi-site data. Extensive experiments on three heterogeneous prostate MRI datasets demonstrate that our MS-Net improves the performance across all datasets consistently, and outperforms state-of-the-art methods for multi-site learning.