TransMVAN: Multi-view Aggregation Network with Transformer for Pneumonia Diagnosis.

Automated and accurate classification of pneumonia plays a crucial role in improving the performance of computer-aided diagnosis systems for chest X-ray images. Nevertheless, it is a challenging task due to the difficulty of learning the complex structure information of lung abnormality from chest X-ray images. In this paper, we propose a multi-view aggregation network with Transformer (TransMVAN) for pneumonia classification in chest X-ray images. Specifically, we propose to incorporate the knowledge from glance and focus views to enrich the feature representation of lung abnormality. Moreover, to capture the complex relationships among different lung regions, we propose a bi-directional multi-scale vision Transformer (biMSVT), with which the informative messages between different lung regions are propagated through two directions. In addition, we also propose a gated multi-view aggregation (GMVA) to adaptively select the feature information from glance and focus views for further performance enhancement of pneumonia diagnosis. Our proposed method achieves AUCs of 0.9645 and 0.9550 for pneumonia classification on two different chest X-ray image datasets. In addition, it achieves an AUC of 0.9761 for evaluating positive and negative polymerase chain reaction (PCR). Furthermore, our proposed method also attains an AUC of 0.9741 for classifying non-COVID-19 pneumonia, COVID-19 pneumonia, and normal cases. Experimental results demonstrate the effectiveness of our method over other methods used for comparison in pneumonia diagnosis from chest X-ray images.

Diagnostic Performance of a Deep Learning Model Deployed at a National COVID-19 Screening Facility for Detection of Pneumonia on Frontal Chest Radiographs.

(1) Background: Chest radiographs are the mainstay of initial radiological investigation in this COVID-19 pandemic. A reliable and readily deployable artificial intelligence (AI) algorithm that detects pneumonia in COVID-19 suspects can be useful for screening or triage in a hospital setting. This study has a few objectives: first, to develop a model that accurately detects pneumonia in COVID-19 suspects; second, to assess its performance in a real-world clinical setting; and third, by integrating the model with the daily clinical workflow, to measure its impact on report turn-around time. (2) Methods: The model was developed from the NIH Chest-14 open-source dataset and fine-tuned using an internal dataset comprising more than 4000 CXRs acquired in our institution. Input from two senior radiologists provided the reference standard. The model was integrated into daily clinical workflow, prioritising abnormal CXRs for expedited reporting. Area under the receiver operating characteristic curve (AUC), F1 score, sensitivity, and specificity were calculated to characterise diagnostic performance. The average time taken by radiologists in reporting the CXRs was compared against the mean baseline time taken prior to implementation of the AI model. (3) Results: 9431 unique CXRs were included in the datasets, of which 1232 were ground truth-labelled positive for pneumonia. On the "live" dataset, the model achieved an AUC of 0.95 (95% confidence interval (CI): 0.92, 0.96) corresponding to a specificity of 97% (95% CI: 0.97, 0.98) and sensitivity of 79% (95% CI: 0.72, 0.84). No statistically significant degradation of diagnostic performance was encountered during clinical deployment, and report turn-around time was reduced by 22%. (4) Conclusion: In real-world clinical deployment, our model expedites reporting of pneumonia in COVID-19 suspects while preserving diagnostic performance without significant model drift.

Multi-level Attentive Skin Lesion Learning for Melanoma Classification.

Melanoma classification plays an important role in skin lesion diagnosis. Nevertheless, melanoma classification is a challenging task, due to the appearance variation of the skin lesions, and the interference of the noises from dermoscopic imaging. In this paper, we propose a multi-level attentive skin lesion learning (MASLL) network to enhance melanoma classification. Specifically, we design a local learning branch with a skin lesion localization (SLL) module to assist the network in learning the lesion features from the region of interest. In addition, we propose a weighted feature integration (WFI) module to fuse the lesion information from the global and local branches, which further enhances the feature discriminative capability of the skin lesions. Experimental results on ISIC 2017 dataset show the effectiveness of the proposed method on melanoma classification.

Coronary Artery Extraction from CT Coronary Angiography with Augmentation on Partially Labelled Data.

Coronary artery disease (CAD) is an important cause of morbidity and mortality. CT coronary angiography is considered as first-line of investigation in patients suspected of having CAD. Coronary artery centerline extraction is a challenging prerequisite for coronary artery stenosis evaluation. These challenges include the small and complex structure, variation of plaques and imaging noise. Deep learning methods often require adequate annotated data to build a good model. This work aims to adopt a dataset that has partial annotation to augment the data to train a Coronary Neural Network (CorNN) to track the coronary artery centerline. We combined a small training dataset with densely labelled centerline and radius, augmented with a larger dataset with only the centerline sparsely labelled to train networks to track centerlines from 3D computed tomography coronary angiography. The vessel orientation estimation is patch based, with or without additional radius prediction. The patch data are carefully positioned and sampled, which are tagged with the orientations computed based on the centerlines. Our experiment results show that, with the augmentation of the new data, although partially annotated, nearly 10% or more improvement has been achieved for the coronary artery extraction by the proposed approach.

Brain Functional Changes in Stroke Following Rehabilitation Using Brain-Computer Interface-Assisted Motor Imagery With and Without tDCS: A Pilot Study.

Brain-computer interface-assisted motor imagery (MI-BCI) or transcranial direct current stimulation (tDCS) has been proven effective in post-stroke motor function enhancement, yet whether the combination of MI-BCI and tDCS may further benefit the rehabilitation of motor functions remains unknown. This study investigated brain functional activity and connectivity changes after a 2 week MI-BCI and tDCS combined intervention in 19 chronic subcortical stroke patients. Patients were randomized into MI-BCI with tDCS group and MI-BCI only group who underwent 10 sessions of 20 min real or sham tDCS followed by 1 h MI-BCI training with robotic feedback. We derived amplitude of low-frequency fluctuation (ALFF), regional homogeneity (ReHo), and functional connectivity (FC) from resting-state functional magnetic resonance imaging (fMRI) data pre- and post-intervention. At baseline, stroke patients had lower ALFF in the ipsilesional somatomotor network (SMN), lower ReHo in the contralesional insula, and higher ALFF/Reho in the bilateral posterior default mode network (DMN) compared to age-matched healthy controls. After the intervention, the MI-BCI only group showed increased ALFF in contralesional SMN and decreased ALFF/Reho in the posterior DMN. In contrast, no post-intervention changes were detected in the MI-BCI + tDCS group. Furthermore, higher increases in ALFF/ReHo/FC measures were related to better motor function recovery (measured by the Fugl-Meyer Assessment scores) in the MI-BCI group while the opposite association was detected in the MI-BCI + tDCS group. Taken together, our findings suggest that brain functional re-normalization and network-specific compensation were found in the MI-BCI only group but not in the MI-BCI + tDCS group although both groups gained significant motor function improvement post-intervention with no group difference. MI-BCI and tDCS may exert differential or even opposing impact on brain functional reorganization during post-stroke motor rehabilitation; therefore, the integration of the two strategies requires further refinement to improve efficacy and effectiveness.

Differential Amplitude of Low-Frequency Fluctuations in brain networks after BCI Training with and without tDCS in Stroke.

Mapping the brain alterations post stroke and post intervention is important for rehabilitation therapy development. Previous work has shown changes in functional connectivity based on resting-state fMRI, structural connectivity derived from diffusion MRI and perfusion as a result of brain-computer interface-assisted motor imagery (MI-BCI) and transcranial direct current stimulation (tDCS) in upper-limb stroke rehabilitation. Besides functional connectivity, regional amplitude of local low-frequency fluctuations (ALFF) may provide complementary information on the underlying neural mechanism in disease. Yet, findings on spontaneous brain activity during resting-state in stroke patients after intervention are limited and inconsistent. Here, we sought to investigate the different brain alteration patterns induced by tDCS compared to MI-BCI for upper-limb rehabilitation in chronic stroke patients using resting-state fMRI-based ALFF method. Our results suggested that stroke patients have lower ALFF in the ipsilesional somatomotor network compared to controls at baseline. Increased ALFF at contralesional somatomotor network and alterations in higher-level cognitive networks such as the default mode network (DMN) and salience networks accompany motor recovery after intervention; though the MI-BCI alone group and MI-BCI combined with tDCS group exhibit differential patterns.

Modeling visual attention's modulatory aftereffects on visual sensitivity and quality evaluation.

With the fast development of visual noise-shaping related applications (visual compression, error resilience, watermarking, encryption, and display), there is an increasingly significant demand on incorporating perceptual characteristics into these applications for improved performance. In this paper, a very important mechanism of the human brain, visual attention, is introduced for visual sensitivity and visual quality evaluation. Based upon the analysis, a new numerical measure for visual attention's modulatory aftereffects, perceptual quality significance map (PQSM), is proposed. To a certain extent, the PQSM reflects the processing ability of the human brain on local visual contents statistically. The PQSM is generated with the integration of local perceptual stimuli from color contrast, texture contrast, motion, as well as cognitive features (skin color and face in this study). Experimental results with subjective viewing demonstrate the performance improvement on two PQSM-modulated visual sensitivity models and two PQSM-based visual quality metrics.

The L0 Regularized Mumford-Shah Model for Bias Correction and Segmentation of Medical Images.

We propose a new variant of the Mumford-Shah model for simultaneous bias correction and segmentation of images with intensity inhomogeneity. First, based on the model of images with intensity inhomogeneity, we introduce an L0 gradient regularizer to model the true intensity and a smooth regularizer to model the bias field. In addition, we derive a new data fidelity using the local intensity properties to allow the bias field to be influenced by its neighborhood. Second, we use a two-stage segmentation method, where the fast alternating direction method is implemented in the first stage for the recovery of true intensity and bias field and a simple thresholding is used in the second stage for segmentation. Different from most of the existing methods for simultaneous bias correction and segmentation, we estimate the bias field and true intensity without fixing either the number of the regions or their values in advance. Our method has been validated on medical images of various modalities with intensity inhomogeneity. Compared with the state-of-art approaches and the well-known brain software tools, our model is fast, accurate, and robust with initializations.

Combining EPI and motion correction for fMRI human brain images with big motion.

Motion correction is an important component in fMRI brain image analysis. Linear registration technique is mostly used in the process based on the assumption that there is not any shape changes of human brain during imaging process. Echo planar imaging (EPI) technique has been widely adapted in fMRI imaging to shorten encoding duration and increase temporal resolution. However, due to the magnetic field inhomogeneity caused by tissues, shape distortion and signal intensity lose are brought into fMRI images by the technique. On the other hand, subject's pose in scanner has a effect on magnetic field inhomogeneity, so the EPI distortions are subject to head movement, especially when the movement is big. As a result, most current motion correction techniques, which are based on rigid registration, cannot handle the problem. In this paper, a technique that combines EPI distortion correction and motion correction to handle the above-mentioned problem is proposed. Since it is almost impossible to obtain ground truth at present, a task-related fMRI BOLD time course image with big motion is selected as experimental material to test its performance. The image is pre-processed with the proposed EPI-motion correction scheme then analyzed by FSL feat tool. Compared with another process with only motion correction and FSL feat analysis, the experimental result using the proposed method has no false activation detection. It is suggested the proposed EPI-motion correction scheme has the ability to handle the fMRI human brain images with big motion.