Automatic Segmentation of Acute Ischemic Stroke From DWI Using 3-D Fully Convolutional DenseNets

Deep learning-based automatic segmentation of cerebral infarcts on diffusion MRI

We explored effects of (1) training with various sample sizes of multi-site vs. single-site training data, (2) cross-site domain adaptation, and (3) data sources and features on the performance of algorithms segmenting cerebral infarcts on Magnetic Resonance Imaging (MRI). We used 10,820 annotated diffusion-weighted images (DWIs) from 10 university hospitals. Algorithms based on 3D U-net were trained using progressively larger subsamples (ranging from 217 to 8661), while internal testing employed a distinct set of 2159 DWIs. External validation was conducted using three unrelated datasets (n = 2777, 50, and 250). For domain adaptation, we utilized 50 to 1000 subsamples from the 2777-image external target dataset. As the size of the multi-site training data increased from 217 to 1732, the Dice similarity coefficient (DSC) and average Hausdorff distance (AHD) improved from 0.58 to 0.65 and from 16.1 to 3.75 mm, respectively. Further increases in sample size to 4330 and 8661 led to marginal gains in DSC (to 0.68 and 0.70, respectively) and in AHD (to 2.92 and 1.73). Similar outcomes were observed in external testing. Notably, performance was relatively poor for segmenting brainstem or hyperacute (< 3 h) infarcts. Domain adaptation, even with a small subsample (n = 50) of external data, conditioned the algorithm trained with 217 images to perform comparably to an algorithm trained with 8661 images. In conclusion, the use of multi-site data (approximately 2000 DWIs) and domain adaptation significantly enhances the performance and generalizability of deep learning algorithms for infarct segmentation.

Multi-type stroke lesion segmentation: comparison of single-stage and hierarchical approach

Stroke, a major cause of death and disability worldwide, can be haemorrhagic or ischaemic depending on the type of bleeding in the brain. Rapid and accurate identification of stroke type and lesion segmentation is critical for timely and effective treatment. However, existing research primarily focuses on segmenting a single stroke type, potentially limiting their clinical applicability. This study addresses this gap by exploring multi-type stroke lesion segmentation using deep learning methods. Specifically, we investigate two distinct approaches: a single-stage approach that directly segments all tissue types in one model and a hierarchical approach that first classifies stroke types and then utilises specialised segmentation models for each subtype. Recognising the importance of accurate stroke classification for the hierarchical approach, we evaluate ResNet, ResNeXt and ViT networks, incorporating focal loss and oversampling techniques to mitigate the impact of class imbalance. We further explore the performance of U-Net, U-Net++ and DeepLabV3 models for segmentation within each approach. We use a comprehensive dataset of 6650 images provided by the Ministry of Health of the Republic of Türkiye. This dataset includes 1130 ischaemic strokes, 1093 haemorrhagic strokes and 4427 non-stroke cases. In our comparative experiments, we achieve an AUC score of 0.996 when classifying stroke and non-stroke slices. For lesion segmentation task, while the performance of different architectures is comparable, the hierarchical training approach outperforms the single-stage approach in terms of intersection over union (IoU). The performance of the U-Net model increased significantly from an IoU of 0.788 to 0.875 when the hierarchical approach is used. This comparative analysis aims to identify the most effective approach and deep learning model for multi-type stroke lesion segmentation in brain CT scans, potentially leading to improved clinical decision-making, treatment efficiency and outcomes.

Contrast quality control for segmentation task based on deep learning models-Application to stroke lesion in CT imaging

Introduction

Although medical imaging plays a crucial role in stroke management, machine learning (ML) has been increasingly used in this field, particularly in lesion segmentation. Despite advances in acquisition technologies and segmentation architectures, one of the main challenges of subacute stroke lesion segmentation in computed tomography (CT) imaging is image contrast.

Methods

To address this issue, we propose a method to assess the contrast quality of an image dataset with a ML trained model for segmentation. This method identifies the critical contrast level below which the medical-imaging model fails to learn meaningful content from images. Contrast measurement relies on the Fisher's ratio, estimating how well the stroke lesion is contrasted from the background. The critical contrast is found-thanks to the following three methods: Performance, graphical, and clustering analysis. Defining this threshold improves dataset design and accelerates training by excluding low-contrast images.

Results

Application of this method to brain lesion segmentation in CT imaging highlights a Fisher's ratio threshold value of 0.05, and training validation of a new model without these images confirms this with similar results with only 60% of the training data, resulting in an almost 30% reduction in initial training time. Moreover, the model trained without the low-contrast images performed equally well with all images when tested on another database.

Discussion

This study opens discussion with clinicians concerning the limitations, areas for improvement, and strategies for enhancing datasets and training models. While the methodology was only applied to stroke lesion segmentation in CT images, it has the potential to be adapted to other tasks.

Evaluation of Advanced Artificial Intelligence Algorithms' Diagnostic Efficacy in Acute Ischemic Stroke: A Comparative Analysis of ChatGPT-4o and Claude 3.5 Sonnet Models

Background/Objectives: Acute ischemic stroke (AIS) is a leading cause of mortality and disability worldwide, with early and accurate diagnosis being critical for timely intervention and improved patient outcomes. This retrospective study aimed to assess the diagnostic performance of two advanced artificial intelligence (AI) models, Chat Generative Pre-trained Transformer (ChatGPT-4o) and Claude 3.5 Sonnet, in identifying AIS from diffusion-weighted imaging (DWI). Methods: The DWI images of a total of 110 cases (AIS group: n = 55, healthy controls: n = 55) were provided to the AI models via standardized prompts. The models' responses were compared to radiologists' gold-standard evaluations, and performance metrics such as sensitivity, specificity, and diagnostic accuracy were calculated. Results: Both models exhibited a high sensitivity for AIS detection (ChatGPT-4o: 100%, Claude 3.5 Sonnet: 94.5%). However, ChatGPT-4o demonstrated a significantly lower specificity (3.6%) compared to Claude 3.5 Sonnet (74.5%). The agreement with radiologists was poor for ChatGPT-4o (κ = 0.036; %95 CI: -0.013, 0.085) but good for Claude 3.5 Sonnet (κ = 0.691; %95 CI: 0.558, 0.824). In terms of the AIS hemispheric localization accuracy, Claude 3.5 Sonnet (67.2%) outperformed ChatGPT-4o (32.7%). Similarly, for specific AIS localization, Claude 3.5 Sonnet (30.9%) showed greater accuracy than ChatGPT-4o (7.3%), with these differences being statistically significant (p < 0.05). Conclusions: This study highlights the superior diagnostic performance of Claude 3.5 Sonnet compared to ChatGPT-4o in identifying AIS from DWI. Despite its advantages, both models demonstrated notable limitations in accuracy, emphasizing the need for further development before achieving full clinical applicability. These findings underline the potential of AI tools in radiological diagnostics while acknowledging their current limitations.

DeSPPNet: A Multiscale Deep Learning Model for Cardiac Segmentation

BACKGROUND

Cardiac magnetic resonance imaging (MRI) plays a crucial role in monitoring disease progression and evaluating the effectiveness of treatment interventions. Cardiac MRI allows medical practitioners to assess cardiac function accurately by providing comprehensive and quantitative information about the structure and function, hence making it an indispensable tool for monitoring the disease and treatment response. Deep learning-based segmentation enables the precise delineation of cardiac structures including the myocardium, right ventricle, and left ventricle. The accurate segmentation of these structures helps in the diagnosis of heart failure, cardiac functional response to therapies, and understanding the state of the heart functions after treatment.

OBJECTIVES

The objective of this study is to develop a multiscale deep learning model to segment cardiac organs based on MRI imaging data. Good segmentation performance is difficult to achieve due to the complex nature of the cardiac structure, which includes a variety of chambers, arteries, and tissues. Furthermore, the human heart is also constantly beating, leading to motion artifacts that reduce image clarity and consistency. As a result, a multiscale method is explored to overcome various challenges in segmenting cardiac MRI images.

METHODS

This paper proposes DeSPPNet, a multiscale-based deep learning network. Its foundation follows encoder-decoder pair architecture that utilizes the Spatial Pyramid Pooling (SPP) layer to improve the performance of cardiac semantic segmentation. The SPP layer is designed to pool features from densely convolutional layers at different scales or sizes, which will be combined to maintain a set of spatial information. By processing features at different spatial resolutions, the multiscale densely connected layer in the form of the Pyramid Pooling Dense Module (PPDM) helps the network to capture both local and global context, preserving finer details of the cardiac structure while also capturing the broader context required to accurately segment larger cardiac structures. The PPDM is incorporated into the deeper layer of the encoder section of the deep learning network to allow it to recognize complex semantic features.

RESULTS

An analysis of multiple PPDM placement scenarios and structural variations revealed that the 3-path PPDM, positioned at the encoder layer 5, yielded optimal segmentation performance, achieving dice, intersection over union (IoU), and accuracy scores of 0.859, 0.800, and 0.993, respectively.

CONCLUSIONS

Different PPDM configurations produce a different effect on the network; as such, a shallower layer placement, like encoder layer 4, retains more spatial data that need more parallel paths to gather the optimal set of multiscale features. In contrast, deeper layers contain more informative features but at a lower spatial resolution, which reduces the number of parallel paths required to provide optimal multiscale context.

Robust Ensemble of Two Different Multimodal Approaches to Segment 3D Ischemic Stroke Segmentation Using Brain Tumor Representation Among Multiple Center Datasets

Ischemic stroke segmentation at an acute stage is vital in assessing the severity of patients' impairment and guiding therapeutic decision-making for reperfusion. Although many deep learning studies have shown attractive performance in medical segmentation, it is difficult to use these models trained on public data with private hospitals' datasets. Here, we demonstrate an ensemble model that employs two different multimodal approaches for generalization, a more effective way to perform on external datasets. First, after we jointly train a segmentation model on diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) MR modalities, the model is inferred on the DWI images. Second, a channel-wise segmentation model is trained by concatenating the DWI and ADC images as input, and then is inferred using both MR modalities. Before training with ischemic stroke data, we utilized BraTS 2021, a public brain tumor dataset, for transfer learning. An extensive ablation study evaluates which strategy learns better representations for ischemic stroke segmentation. In our study, nnU-Net well-known for robustness is selected as our baseline model. Our proposed method is evaluated on three different datasets: the Asan Medical Center (AMC) I and II, and the 2022 Ischemic Stroke Lesion Segmentation (ISLES). Our experiments are widely validated over a large, multi-center, and multi-scanner dataset with a huge amount of 846 scans. Not only stroke lesion models can benefit from transfer learning using brain tumor data, but combining the MR modalities using different training schemes also highly improves segmentation performance. The method achieved a top-1 ranking in the ongoing ISLES'22 challenge and performed particularly well on lesion-wise metrics of interest to neuroradiologists, achieving a Dice coefficient of 78.69% and a lesion-wise F1 score of 82.46%. Also, the method was relatively robust on the AMC I (Dice, 60.35%; lesion-wise F1, 68.30%) and II (Dice; 74.12%; lesion-wise F1, 67.53%) datasets in different settings. The high segmentation accuracy of our proposed method could improve radiologists' ability to detect ischemic stroke lesions in MRI images. Our model weights and inference code are available on https://github.com/MDOpx/ISLES22-model-inference .

A 25-Year Retrospective of the Use of AI for Diagnosing Acute Stroke: Systematic Review

BACKGROUND

Stroke is a leading cause of death and disability worldwide. Rapid and accurate diagnosis is crucial for minimizing brain damage and optimizing treatment plans.

OBJECTIVE

This review aims to summarize the methods of artificial intelligence (AI)-assisted stroke diagnosis over the past 25 years, providing an overview of performance metrics and algorithm development trends. It also delves into existing issues and future prospects, intending to offer a comprehensive reference for clinical practice.

METHODS

A total of 50 representative articles published between 1999 and 2024 on using AI technology for stroke prevention and diagnosis were systematically selected and analyzed in detail.

RESULTS

AI-assisted stroke diagnosis has made significant advances in stroke lesion segmentation and classification, stroke risk prediction, and stroke prognosis. Before 2012, research mainly focused on segmentation using traditional thresholding and heuristic techniques. From 2012 to 2016, the focus shifted to machine learning (ML)-based approaches. After 2016, the emphasis moved to deep learning (DL), which brought significant improvements in accuracy. In stroke lesion segmentation and classification as well as stroke risk prediction, DL has shown superiority over ML. In stroke prognosis, both DL and ML have shown good performance.

CONCLUSIONS

Over the past 25 years, AI technology has shown promising performance in stroke diagnosis.

FlexDTI: flexible diffusion gradient encoding scheme-based highly efficient diffusion tensor imaging using deep learning

Objective. Most deep neural network-based diffusion tensor imaging methods require the diffusion gradients' number and directions in the data to be reconstructed to match those in the training data. This work aims to develop and evaluate a novel dynamic-convolution-based method called FlexDTI for highly efficient diffusion tensor reconstruction with flexible diffusion encoding gradient scheme.Approach. FlexDTI was developed to achieve high-quality DTI parametric mapping with flexible number and directions of diffusion encoding gradients. The method used dynamic convolution kernels to embed diffusion gradient direction information into feature maps of the corresponding diffusion signal. Furthermore, it realized the generalization of a flexible number of diffusion gradient directions by setting the maximum number of input channels of the network. The network was trained and tested using datasets from the Human Connectome Project and local hospitals. Results from FlexDTI and other advanced tensor parameter estimation methods were compared.Main results. Compared to other methods, FlexDTI successfully achieves high-quality diffusion tensor-derived parameters even if the number and directions of diffusion encoding gradients change. It reduces normalized root mean squared error by about 50% on fractional anisotropy and 15% on mean diffusivity, compared with the state-of-the-art deep learning method with flexible diffusion encoding gradient scheme.Significance. FlexDTI can well learn diffusion gradient direction information to achieve generalized DTI reconstruction with flexible diffusion gradient scheme. Both flexibility and reconstruction quality can be taken into account in this network.

Global research evolution and frontier analysis of artificial intelligence in brain injury: A bibliometric analysis

Research on artificial intelligence for brain injury is currently a prominent area of scientific research. A significant amount of related literature has been accumulated in this field. This study aims to identify hotspots and clarify research resources by conducting literature metrology visualization analysis, providing valuable ideas and references for related fields. The research object of this paper consists of 3000 articles cited in the core database of Web of Science from 1998 to 2023. These articles are visualized and analyzed using VOSviewer and CiteSpace. The bibliometric analysis reveals a continuous increase in the number of articles published on this topic, particularly since 2016, indicating significant growth. The United States stands out as the leading country in artificial intelligence for brain injury, followed by China, which tends to catch up. The core research institutions are primarily universities in developed countries, but there is a lack of cooperation and communication between research groups. With the development of computer technology, the research in this field has shown strong wave characteristics, experiencing the early stage of applied research based on expert systems, the middle stage of prediction research based on machine learning, and the current phase of diversified research focused on deep learning. Artificial intelligence has innovative development prospects in brain injury, providing a new orientation for the treatment and auxiliary diagnosis in this field.

Automatic segmentation of brain glioma based on XY-Net

Glioma is a malignant primary brain tumor, which can easily lead to death if it is not detected in time. Magnetic resonance imaging is the most commonly used technique to diagnose gliomas, and precise outlining of tumor areas from magnetic resonance images (MRIs) is an important aid to physicians in understanding the patient's condition and formulating treatment plans. However, relying on radiologists to manually depict tumors is a tedious and laborious task, so it is clinically important to investigate an automated method for outlining glioma regions in MRIs. To liberate radiologists from the heavy task of outlining tumors, we propose a fully convolutional network, XY-Net, based on the most popular U-Net symmetric encoder-decoder structure to perform automatic segmentation of gliomas. We construct two symmetric sub-encoders for XY-Net and build interconnected X-shaped feature map transmission paths between the sub-encoders, while maintaining the feature map concatenation between each sub-encoder and the decoder. Moreover, a loss function composed of the balanced cross-entropy loss function and the dice loss function is used in the training task of XY-Net to solve the class unevenness problem of the medical image segmentation task. The experimental results show that the proposed XY-Net has a 2.16% improvement in dice coefficient (DC) compared to the network model with a single encoder structure, and compare with some state-of-the-art image segmentation methods, XY-Net achieves the best performance. The DC, HD, recall, and precision of our method on the test set are 74.49%, 10.89 mm, 78.06%, and 76.30%, respectively. The combination of sub-encoders and cross-transmission paths enables the model to perform better; based on this combination, the XY-Net achieves an end-to-end automatic segmentation of gliomas on 2D slices of MRIs, which can play a certain auxiliary role for doctors in grasping the state of illness.

Localization of early infarction on non-contrast CT images in acute ischemic stroke with deep learning approach

Localization of early infarction on first-line Non-contrast computed tomogram (NCCT) guides prompt treatment to improve stroke outcome. Our previous study has shown a good performance in the identification of ischemic injury on NCCT. In the present study, we developed a deep learning (DL) localization model to help localize the early infarction sign on NCCT. This retrospective study included consecutive 517 ischemic stroke (IS) patients who received NCCT within 12 h after stroke onset. A total of 21,436 infarction patches and 20,391 non-infarction patches were extracted from the slice pool of 1,634 NCCT according to brain symmetricity property. The generated patches were fed into different pretrained convolutional neural network (CNN) models such as Visual Geometry Group 16 (VGG16), GoogleNet, Residual Networks 50 (ResNet50), Inception-ResNet-v2 (IR-v2), Inception-v3 and Inception-v4. The selected VGG16 model could detect the early infarction in both supratentorial and infratentorial regions to achieve an average area under curve (AUC) 0.73 after extensive customization. The properly tuned-VGG16 model could identify the early infarction in the cortical, subcortical and cortical plus subcortical areas of supratentorial region with the mean AUC > 0.70. Further, the model could attain 95.6% of accuracy on recognizing infarction lesion in 494 out of 517 IS patients.

MSA-YOLOv5: Multi-scale attention-based YOLOv5 for automatic detection of acute ischemic stroke from multi-modality MRI images

BACKGROUND AND OBJECTIVE

Acute ischemic stroke (AIS) is a common neurological disorder characterized by the sudden onset of cerebral ischemia, leading to functional impairments. Swift and precise detection of AIS lesions is crucial for stroke diagnosis and treatment but poses a significant challenge. This study aims to leverage multimodal fusion technology to combine complementary information from various modalities, thereby enhancing the detection performance of AIS target detection models.

METHODS

In this retrospective study of AIS, we collected data from 316 AIS patients and created a multi-modality magnetic resonance imaging (MRI) dataset. We propose a Multi-Scale Attention-based YOLOv5 (MSA-YOLOv5), targeting challenges such as small lesion size and blurred borders at low resolutions. Specifically, we augment YOLOv5 with a prediction head to detect objects at various scales. Next, we replace the original prediction head with a Multi-Scale Swin Transformer Prediction Head (MS-STPH), which reduces computational complexity to linear levels and enhances the ability to detect small lesions. We incorporate a Second-Order channel attention (SOCA) module to adaptively rescale channel features by employing second-order feature statistics for more discriminative representations. Finally, we further validate the effectiveness of our method using the ISLES 2022 dataset.

RESULTS

On our in-house AIS dataset, MSA-YOLOv5 achieves a 79.0% mAP0.5, substantially surpassing other single-stage models. Compared to two-stage models, it maintains a comparable performance level while significantly reducing the number of parameters and resolution. On the ISLES 2022 dataset, MSA-YOLOv5 attains an 80.0% mAP0.5, outperforming other network models by a considerable margin. MS-STPH and SOCA modules can significantly increase mAP0.5 by 2.7% and 1.9%, respectively. Visualization interpretability results show that the proposed MSA-YOLOv5 restricts high attention in the small regions of AIS lesions.

CONCLUSIONS

The proposed MSA-YOLOv5 is capable of automatically and effectively detecting acute ischemic stroke lesions in multimodal images, particularly for small lesions and artifacts. Our enhanced model reduces the number of parameters while improving detection accuracy. This model can potentially assist radiologists in providing more accurate diagnosis, and enable clinicians to develop better treatment plans.

The role of input imaging combination and ADC threshold on segmentation of acute ischemic stroke lesion using U-Net

BACKGROUND

To evaluate the effect of the weighting of input imaging combo and ADC threshold on the performance of the U-Net and to find an optimized input imaging combo and ADC threshold in segmenting acute ischemic stroke (AIS) lesion.

METHODS

This study retrospectively enrolled a total of 212 patients having AIS. Four combos, including ADC-ADC-ADC (AAA), DWI-ADC-ADC (DAA), DWI-DWI-ADC (DDA), and DWI-DWI-DWI (DDD), were used as input images, respectively. Three ADC thresholds including 0.6, 0.8 and 1.8 × 10-3 mm2/s were applied. Dice similarity coefficient (DSC) was used to evaluate the segmentation performance of U-Nets. Nonparametric Kruskal-Wallis test with Tukey-Kramer post-hoc tests were used for comparison. A p < .05 was considered statistically significant.

RESULTS

The DSC significantly varied among different combos of images and different ADC thresholds. Hybrid U-Nets outperformed uniform U-Nets at ADC thresholds of 0.6 × 10-3 mm2/s and 0.8 × 10-3 mm2/s (p < .001). The U-Net with imaging combo of DDD had segmentation performance similar to hybrid U-Nets at an ADC threshold of 1.8 × 10-3 mm2/s (p = .062 to 1). The U-Net using the imaging combo of DAA at the ADC threshold of 0.6 × 10-3 mm2/s achieved the highest DSC in the segmentation of AIS lesion.

CONCLUSIONS

The segmentation performance of U-Net for AIS varies among the input imaging combos and ADC thresholds. The U-Net is optimized by choosing the imaging combo of DAA at an ADC threshold of 0.6 × 10-3 mm2/s in segmentating AIS lesion with highest DSC.

KEY POINTS

• Segmentation performance of U-Net for AIS differs among input imaging combos. • Segmentation performance of U-Net for AIS differs among ADC thresholds. • U-Net is optimized using DAA with ADC = 0.6 × 10-3 mm2/s.

A large public dataset of annotated clinical MRIs and metadata of patients with acute stroke

To extract meaningful and reproducible models of brain function from stroke images, for both clinical and research proposes, is a daunting task severely hindered by the great variability of lesion frequency and patterns. Large datasets are therefore imperative, as well as fully automated image post-processing tools to analyze them. The development of such tools, particularly with artificial intelligence, is highly dependent on the availability of large datasets to model training and testing. We present a public dataset of 2,888 multimodal clinical MRIs of patients with acute and early subacute stroke, with manual lesion segmentation, and metadata. The dataset provides high quality, large scale, human-supervised knowledge to feed artificial intelligence models and enable further development of tools to automate several tasks that currently rely on human labor, such as lesion segmentation, labeling, calculation of disease-relevant scores, and lesion-based studies relating function to frequency lesion maps.

Hybrid Contextual Semantic Network for Accurate Segmentation and Detection of Small-Size Stroke Lesions From MRI

Stroke is a cerebrovascular disease with high mortality and disability rates. The occurrence of the stroke typically produces lesions of different sizes, with the accurate segmentation and detection of small-size stroke lesions being closely related to the prognosis of patients. However, the large lesions are usually correctly identified, the small-size lesions are usually ignored. This article provides a hybrid contextual semantic network (HCSNet) that can accurately and simultaneously segment and detect small-size stroke lesions from magnetic resonance images. HCSNet inherits the advantages of the encoder-decoder architecture and applies a novel hybrid contextual semantic module that generates high-quality contextual semantic features from the spatial and channel contextual semantic features through the skip connection layer. Moreover, a mixing-loss function is proposed to optimize HCSNet for unbalanced small-size lesions. HCSNet is trained and evaluated on 2D magnetic resonance images produced from the Anatomical Tracings of Lesions After Stroke challenge (ATLAS R2.0). Extensive experiments demonstrate that HCSNet outperforms several other state-of-the-art methods in its ability to segment and detect small-size stroke lesions. Visualization and ablation experiments reveal that the hybrid semantic module improves the segmentation and detection performance of HCSNet.

Combining convolutional attention mechanism and residual deformable Transformer for infarct segmentation from CT scans of acute ischemic stroke patients

Background

Segmentation and evaluation of infarcts on medical images are essential for diagnosis and prognosis of acute ischemic stroke (AIS). Computed tomography (CT) is the first-choice examination for patients with AIS.

Methods

To accurately segment infarcts from the CT images of patients with AIS, we proposed an automated segmentation method combining the convolutional attention mechanism and residual Deformable Transformer in this article. The method used the encoder-decoder structure, where the encoders were employed for downsampling to obtain the feature of the images and the decoder was used for upsampling and segmentation. In addition, we further applied the convolutional attention mechanism and residual network structure to improve the effectiveness of feature extraction. Our code is available at: https://github.com/XZhiXiang/AIS-segmentation/tree/master.

Results

The proposed method was assessed on a public dataset containing 397 non-contrast CT (NCCT) images of AIS patients (AISD dataset). The symptom onset to CT time was less than 24 h. The experimental results illustrate that this work had a Dice coefficient (DC) of 58.66% for AIS infarct segmentation, which outperforms several existing methods. Furthermore, volumetric analysis of infarcts indicated a strong correlation (Pearson correlation coefficient = 0.948) between the AIS infarct volume obtained by the proposed method and manual segmentation.

Conclusion

The strong correlation between the infarct segmentation obtained via our method and the ground truth allows us to conclude that our method could accurately segment infarcts from NCCT images.

CarveMix: A simple data augmentation method for brain lesion segmentation

Brain lesion segmentation provides a valuable tool for clinical diagnosis and research, and convolutional neural networks (CNNs) have achieved unprecedented success in the segmentation task. Data augmentation is a widely used strategy to improve the training of CNNs. In particular, data augmentation approaches that mix pairs of annotated training images have been developed. These methods are easy to implement and have achieved promising results in various image processing tasks. However, existing data augmentation approaches based on image mixing are not designed for brain lesions and may not perform well for brain lesion segmentation. Thus, the design of this type of simple data augmentation method for brain lesion segmentation is still an open problem. In this work, we propose a simple yet effective data augmentation approach, dubbed as CarveMix, for CNN-based brain lesion segmentation. Like other mixing-based methods, CarveMix stochastically combines two existing annotated images (annotated for brain lesions only) to obtain new labeled samples. To make our method more suitable for brain lesion segmentation, CarveMix is lesion-aware, where the image combination is performed with a focus on the lesions and preserves the lesion information. Specifically, from one annotated image we carve a region of interest (ROI) according to the lesion location and geometry with a variable ROI size. The carved ROI then replaces the corresponding voxels in a second annotated image to synthesize new labeled images for network training, and additional harmonization steps are applied for heterogeneous data where the two annotated images can originate from different sources. Besides, we further propose to model the mass effect that is unique to whole brain tumor segmentation during image mixing. To evaluate the proposed method, experiments were performed on multiple publicly available or private datasets, and the results show that our method improves the accuracy of brain lesion segmentation. The code of the proposed method is available at https://github.com/ZhangxinruBIT/CarveMix.git.

Brain segmentation in patients with perinatal arterial ischemic stroke

BACKGROUND

Perinatal arterial ischemic stroke (PAIS) is associated with adverse neurological outcomes. Quantification of ischemic lesions and consequent brain development in newborn infants relies on labor-intensive manual assessment of brain tissues and ischemic lesions. Hence, we propose an automatic method utilizing convolutional neural networks (CNNs) to segment brain tissues and ischemic lesions in MRI scans of infants suffering from PAIS.

MATERIALS AND METHODS

This single-center retrospective study included 115 patients with PAIS that underwent MRI after the stroke onset (baseline) and after three months (follow-up). Nine baseline and 12 follow-up MRI scans were manually annotated to provide reference segmentations (white matter, gray matter, basal ganglia and thalami, brainstem, ventricles, extra-ventricular cerebrospinal fluid, and cerebellum, and additionally on the baseline scans the ischemic lesions). Two CNNs were trained to perform automatic segmentation on the baseline and follow-up MRIs, respectively. Automatic segmentations were quantitatively evaluated using the Dice coefficient (DC) and the mean surface distance (MSD). Volumetric agreement between segmentations that were manually and automatically obtained was computed. Moreover, the scan quality and automatic segmentations were qualitatively evaluated in a larger set of MRIs without manual annotation by two experts. In addition, the scan quality was qualitatively evaluated in these scans to establish its impact on the automatic segmentation performance.

RESULTS

Automatic brain tissue segmentation led to a DC and MSD between 0.78-0.92 and 0.18-1.08 mm for baseline, and between 0.88-0.95 and 0.10-0.58 mm for follow-up scans, respectively. For the ischemic lesions at baseline the DC and MSD were between 0.72-0.86 and 1.23-2.18 mm, respectively. Volumetric measurements indicated limited oversegmentation of the extra-ventricular cerebrospinal fluid in both the follow-up and baseline scans, oversegmentation of the ischemic lesions in the left hemisphere, and undersegmentation of the ischemic lesions in the right hemisphere. In scans without imaging artifacts, brain tissue segmentation was graded as excellent in more than 85% and 91% of cases, respectively for the baseline and follow-up scans. For the ischemic lesions at baseline, this was in 61% of cases.

CONCLUSIONS

Automatic segmentation of brain tissue and ischemic lesions in MRI scans of patients with PAIS is feasible. The method may allow evaluation of the brain development and efficacy of treatment in large datasets.

The multi-level classification network (MCN) with modified residual U-Net for ischemic stroke lesions segmentation from ATLAS

Ischemic and hemorrhagic strokes are two major types of internal brain injury. 3D brain MRI is suggested by neurologists to examine the brain. Manual examination of brain MRI is very sensitive and time-consuming task. However, automatic diagnosis can assist doctors in this regard. Anatomical Tracings of Lesions After Stroke (ATLAS) is publicly available dataset for research experiments. One of the major issues in medical imaging is class imbalance. Similarly, pixel representation of stroke lesion is less than 1% in ATLAS. Second major challenge in this dataset is inter-class similarity. A multi-level classification network (MCN) is proposed for segmentation of ischemic stroke lesions. MCN consists of three cascaded discrete networks. The first network designed to reduce the slice level class imbalance, where a classifier model is trained to extract the slices of stroke lesions from a whole brain MRI volume. The interclass similarity cause to produce false positives in segmented output. Therefore, all extracted stroke slices were divided into overlapping patches (64 × 64) and carried to the second network. The task associated with second network is to classify the patches comprises of stroke lesion. The third network is a 2D modified residual U-Net that segments out the stroke lesions from the patches extracted by the second network. MCN achieved 0.754 mean dice score on test dataset which is higher than the other state-of-the-art methods on the same dataset.

Automatic detection of stroke lesion from diffusion-weighted imaging via the improved YOLOv5

BACKGROUND AND OBJECTIVE

Stroke is the second most deadly disease globally and seriously endangers people's lives and health. The automatic detection of stroke lesions from diffusion-weighted imaging (DWI) can improve the diagnosis. Recently, automatic detection methods based on YOLOv5 have been utilized in medical images. However, most of them barely capture the stroke lesions because of their small size and fuzzy boundaries.

METHODS

To address this problem, a novel method for tracing the edge of the stroke lesion based on YOLOv5 (TE-YOLOv5) is proposed. Specifically, we constantly update the high-level features of the lesion using an aggregate pool (AP) module. Conversely, we feed the extracted feature into the reverse attention (RA) module to trace the edge relationship promptly. Overall, 1681 DWI images of 319 stroke patients have been collected, and experienced radiologists have marked the lesions. DWI images were randomly split into the training and test set at a ratio of 8:2. TE-YOLOv5 has been compared with the related models, and a detailed ablation analysis has been conducted to clarify the role of the RA and AP modules.

RESULTS

TE-YOLOv5 outperforms its counterparts and achieves competitive performance with a precision of 81.5%, a recall of 75.8%, and a mAP@0.5 of 80.7% (mean average precision while the intersection over union is 0.5) under the same backbone. At the patient level, the positive finding rate can reach 98.51%, while the confidence is set at 80.0%. After ablating RA, the mAP@0.5 decreases to 79.6%; after ablating RA and AP, the mAP@0.5 decreases to 78.1%.

CONCLUSIONS

The proposed TE-YOLOv5 can automatically and effectively detect stroke lesions from DWI images, especially for those with an extremely small size and blurred boundaries. AP and RA modules can aggregate multi-layer high-level features and concurrently track the edge relationship of stroke lesions. These detection methods might help radiologists improve stroke diagnosis and have great application potential in clinical practice.

Improving interobserver agreement and performance of deep learning models for segmenting acute ischemic stroke by combining DWI with optimized ADC thresholds

OBJECTIVES

To examine the role of ADC threshold on agreement across observers and deep learning models (DLMs) plus segmentation performance of DLMs for acute ischemic stroke (AIS).

METHODS

Twelve DLMs, which were trained on DWI-ADC-ADC combination from 76 patients with AIS using 6 different ADC thresholds with ground truth manually contoured by 2 observers, were tested by additional 67 patients in the same hospital and another 78 patients in another hospital. Agreement between observers and DLMs were evaluated by Bland-Altman plot and intraclass correlation coefficient (ICC). The similarity between ground truth (GT) defined by observers and between automatic segmentation performed by DLMs was evaluated by Dice similarity coefficient (DSC). Group comparison was performed using the Mann-Whitney U test. The relationship between the DSC and ADC threshold as well as AIS lesion size was evaluated by linear regression analysis. A p < .05 was considered statistically significant.

RESULTS

Excellent interobserver agreement and intraobserver repeatability in the manual segmentation (all ICC > 0.98, p < .001) were achieved. The 95% limit of agreement was reduced from 11.23 cm2 for GT on DWI to 0.59 cm2 for prediction at an ADC threshold of 0.6 × 10-3 mm2/s combined with DWI. The segmentation performance of DLMs was improved with an overall DSC from 0.738 ± 0.214 on DWI to 0.971 ± 0.021 on an ADC threshold of 0.6 × 10-3 mm2/s combined with DWI.

CONCLUSIONS

Combining an ADC threshold of 0.6 × 10-3 mm2/s with DWI reduces interobserver and inter-DLM difference and achieves best segmentation performance of AIS lesions using DLMs.

KEY POINTS

• Higher Dice similarity coefficient (DSC) in predicting acute ischemic stroke lesions was achieved by ADC thresholds combined with DWI than by DWI alone (all p < .05). • DSC had a negative association with the ADC threshold in most sizes, both hospitals, and both observers (most p < .05) and a positive association with the stroke size in all ADC thresholds, both hospitals, and both observers (all p < .001). • An ADC threshold of 0.6 × 10-3 mm2/s eliminated the difference of DSC at any stroke size between observers or between hospitals (p = .07 to > .99).

Deep Learning Applications for Acute Stroke Management

Brain imaging is essential to the clinical care of patients with stroke, a leading cause of disability and death worldwide. Whereas advanced neuroimaging techniques offer opportunities for aiding acute stroke management, several factors, including time delays, inter-clinician variability, and lack of systemic conglomeration of clinical information, hinder their maximal utility. Recent advances in deep machine learning (DL) offer new strategies for harnessing computational medical image analysis to inform decision making in acute stroke. We examine the current state of the field for DL models in stroke triage. First, we provide a brief, clinical practice-focused primer on DL. Next, we examine real-world examples of DL applications in pixel-wise labeling, volumetric lesion segmentation, stroke detection, and prediction of tissue fate postintervention. We evaluate recent deployments of deep neural networks and their ability to automatically select relevant clinical features for acute decision making, reduce inter-rater variability, and boost reliability in rapid neuroimaging assessments, and integrate neuroimaging with electronic medical record (EMR) data in order to support clinicians in routine and triage stroke management. Ultimately, we aim to provide a framework for critically evaluating existing automated approaches, thus equipping clinicians with the ability to understand and potentially apply DL approaches in order to address challenges in clinical practice. ANN NEUROL 2022.

Network impact score is an independent predictor of post-stroke cognitive impairment: A multicenter cohort study in 2341 patients with acute ischemic stroke

BACKGROUND

Post-stroke cognitive impairment (PSCI) is a common consequence of stroke. Accurate prediction of PSCI risk is challenging. The recently developed network impact score, which integrates information on infarct location and size with brain network topology, may improve PSCI risk prediction.

AIMS

To determine if the network impact score is an independent predictor of PSCI, and of cognitive recovery or decline.

METHODS

We pooled data from patients with acute ischemic stroke from 12 cohorts through the Meta VCI Map consortium. PSCI was defined as impairment in ≥ 1 cognitive domain on neuropsychological examination, or abnormal Montreal Cognitive Assessment. Cognitive recovery was defined as conversion from PSCI < 3 months post-stroke to no PSCI at follow-up, and cognitive decline as conversion from no PSCI to PSCI. The network impact score was related to serial measures of PSCI using Generalized Estimating Equations (GEE) models, and to PSCI stratified according to post-stroke interval (<3, 3-12, 12-24, >24 months) and cognitive recovery or decline using logistic regression. Models were adjusted for age, sex, education, prior stroke, infarct volume, and study site.

RESULTS

We included 2341 patients with 4657 cognitive assessments. PSCI was present in 398/844 patients (47%) <3 months, 709/1640 (43%) at 3-12 months, 243/853 (28%) at 12-24 months, and 208/522 (40%) >24 months. Cognitive recovery occurred in 64/181 (35%) patients and cognitive decline in 26/287 (9%). The network impact score predicted PSCI in the univariable (OR 1.50, 95%CI 1.34-1.68) and multivariable (OR 1.27, 95%CI 1.10-1.46) GEE model, with similar ORs in the logistic regression models for specified post-stroke intervals. The network impact score was not associated with cognitive recovery or decline.

CONCLUSIONS

The network impact score is an independent predictor of PSCI. As such, the network impact score may contribute to a more precise and individualized cognitive prognostication in patients with ischemic stroke. Future studies should address if multimodal prediction models, combining the network impact score with demographics, clinical characteristics and other advanced brain imaging biomarkers, will provide accurate individualized prediction of PSCI. A tool for calculating the network impact score is freely available at https://metavcimap.org/features/software-tools/lsm-viewer/.

Dense Convolutional Network and Its Application in Medical Image Analysis

Dense convolutional network (DenseNet) is a hot topic in deep learning research in recent years, which has good applications in medical image analysis. In this paper, DenseNet is summarized from the following aspects. First, the basic principle of DenseNet is introduced; second, the development of DenseNet is summarized and analyzed from five aspects: broaden DenseNet structure, lightweight DenseNet structure, dense unit, dense connection mode, and attention mechanism; finally, the application research of DenseNet in the field of medical image analysis is summarized from three aspects: pattern recognition, image segmentation, and object detection. The network structures of DenseNet are systematically summarized in this paper, which has certain positive significance for the research and development of DenseNet.

Quantification of infarct core signal using CT imaging in acute ischemic stroke

In stroke care, the extent of irreversible brain injury, termed infarct core, plays a key role in determining eligibility for acute treatments, such as intravenous thrombolysis and endovascular reperfusion therapies. Many of the pivotal randomized clinical trials testing those therapies used MRI Diffusion-Weighted Imaging (DWI) or CT Perfusion (CTP) to define infarct core. Unfortunately, these modalities are not available 24/7 outside of large stroke centers. As such, there is a need for accurate infarct core determination using faster and more widely available imaging modalities including Non-Contrast CT (NCCT) and CT Angiography (CTA). Prior studies have suggested that CTA provides improved predictions of infarct core relative to NCCT; however, this assertion has never been numerically quantified by automatic medical image computing pipelines using acquisition protocols not confounded by different scanner manufacturers, or other protocol settings such as exposure times, kilovoltage peak, or imprecision due to contrast bolus delays. In addition, single-phase CTA protocols are at present designed to optimize contrast opacification in the arterial phase. This approach works well to maximize the sensitivity to detect vessel occlusions, however, it may not be the ideal timing to enhance the ischemic infarct core signal (ICS). In this work, we propose an image analysis pipeline on CT-based images of 88 acute ischemic stroke (AIS) patients drawn from a single dynamic acquisition protocol acquired at the acute ischemic phase. We use the first scan at the time of the dynamic acquisition as a proxy for NCCT, and the rest of the scans as a proxy for CTA scans, with bolus imaged at different brain enhancement phases. Thus, we use the terms "NCCT" and "CTA" to refer to them. This pipeline enables us to answer the questions "Does the injection of bolus enhance the infarct core signal?" and "What is the ideal bolus timing to enhance the infarct core signal?" without being influenced by aforementioned factors such as scanner model, acquisition settings, contrast bolus delay, and human reader errors. We use reference MRI DWI images acquired after successful recanalization acting as our gold standard for infarct core. The ICS is quantified by calculating the difference in intensity distribution between the infarct core region and its symmetrical healthy counterpart on the contralateral hemisphere of the brain using a metric derived from information theory, the Kullback-Leibler divergence (KL divergence). We compare the ICS provided by NCCT and CTA and retrieve the optimal timing of CTA bolus to maximize the ICS. In our experiments, we numerically confirm that CTAs provide greater ICS compared to NCCT. Then, we find that, on average, the ideal CTA acquisition time to maximize the ICS is not the current target of standard CTA protocols, i.e., during the peak of arterial enhancement, but a few seconds afterward (median of 3 s; 95% CI [1.5, 3.0]). While there are other studies comparing the prediction potential of ischemic infarct core from NCCT and CTA images, to the best of our knowledge, this analysis is the first to perform a quantitative comparison of the ICS among CT based scans, with and without bolus injection, acquired using the same scanning sequence and a precise characterization of the bolus uptake, hence, reducing potential confounding factors.

Indirect Volume Estimation for Acute Ischemic Stroke from Diffusion Weighted Image Using Slice Image Segmentation

The accurate estimation of acute ischemic stroke (AIS) using diffusion-weighted imaging (DWI) is crucial for assessing patients and guiding treatment options. This study aimed to propose a method that estimates AIS volume in DWI objectively, quickly, and accurately. We used a dataset of DWI with AIS, including 2159 participants (1179 for internal validation and 980 for external validation) with various types of AIS. We constructed algorithms using 3D segmentation (direct estimation) and 2D segmentation (indirect estimation) and compared their performances with those annotated by neurologists. The proposed pretrained indirect model demonstrated higher segmentation performance than the direct model, with a sensitivity, specificity, F1-score, and Jaccard index of 75.0%, 77.9%, 76.0, and 62.1%, respectively, for internal validation, and 72.8%, 84.3%, 77.2, and 63.8%, respectively, for external validation. Volume estimation was more reliable for the indirect model, with 93.3% volume similarity (VS), 0.797 mean absolute error (MAE) for internal validation, VS of 89.2% and a MAE of 2.5% for external validation. These results suggest that the indirect model using 2D segmentation developed in this study can provide an accurate estimation of volume from DWI of AIS and may serve as a supporting tool to help physicians make crucial clinical decisions.

Computational Approaches for Acute Traumatic Brain Injury Image Recognition

In recent years, there have been major advances in deep learning algorithms for image recognition in traumatic brain injury (TBI). Interest in this area has increased due to the potential for greater objectivity, reduced interpretation times and, ultimately, higher accuracy. Triage algorithms that can re-order radiological reading queues have been developed, using classification to prioritize exams with suspected critical findings. Localization models move a step further to capture more granular information such as the location and, in some cases, size and subtype, of intracranial hematomas that could aid in neurosurgical management decisions. In addition to the potential to improve the clinical management of TBI patients, the use of algorithms for the interpretation of medical images may play a transformative role in enabling the integration of medical images into precision medicine. Acute TBI is one practical example that can illustrate the application of deep learning to medical imaging. This review provides an overview of computational approaches that have been proposed for the detection and characterization of acute TBI imaging abnormalities, including intracranial hemorrhage, skull fractures, intracranial mass effect, and stroke.

Multimode Gesture Recognition Algorithm Based on Convolutional Long Short-Term Memory Network

Gesture recognition utilizes deep learning network model to automatically extract deep features of data; however, traditional machine learning algorithms rely on manual feature extraction and poor model generalization ability. In this paper, a multimodal gesture recognition algorithm based on convolutional long-term memory network is proposed. First, a convolutional neural network (CNN) is employed to automatically extract the deeply hidden features of multimodal gesture data. Then, a time series model is constructed using a long short-term memory (LSTM) network to learn the long-term dependence of multimodal gesture features on the time series. On this basis, the classification of multimodal gestures is realized by the SoftMax classifier. Finally, the method is experimented and evaluated on two dynamic gesture datasets, VIVA and NVGesture. Experimental results indicate that the accuracy rates of the proposed method on the VIVA and NVGesture datasets are 92.55% and 87.38%, respectively, and its recognition accuracy and convergence performance are better than those of other comparison algorithms.

ISP-Net: Fusing features to predict ischemic stroke infarct core on CT perfusion maps

BACKGROUND

Acute ischemic stroke is one of the leading death causes. Delineating stoke infarct core in medical images plays a critical role in optimal stroke treatment selection. However, accurate estimation of infarct core still remains challenging because of 1) the large shape and location variation of infarct cores; 2) the complex relationships between perfusion parameters and final tissue outcome.

METHODS

We develop an encoder-decoder based semantic model, i.e., Ischemic Stroke Prediction Network (ISP-Net), to predict infarct core after thrombolysis treatment on CT perfusion (CTP) maps. Features of native CTP, CBF (Cerebral Blood Flow), CBV (Cerebral Blood Volume), MTT (Mean Transit Time), Tmax are generated and fused with five-path convolutions for comprehensive analysis. A multi-scale atrous convolution (MSAC) block is firstly put forward as the enriched high-level feature extractor in ISP-Net to improve prediction accuracy. A retrospective dataset which is collected from multiple stroke centers is used to evaluate the performance of ISP-Net. The gold standard infarct cores are delineated on the follow-up scans, i.e., non-contrast CT (NCCT) or MRI diffusion-weighted image (DWI).

RESULTS

In clinical dataset cross-validation, we achieve mean Dice Similarity Coefficient (DSC) of 0.801, precision of 81.3%, sensitivity of 79.5%, specificity of 99.5%, Area Under Curve (AUC) of 0.721. Our approach yields better outcomes than several advanced deep learning methods, i.e., Deeplab V3, U-Net++, CE-Net, X-Net and Non-local U-Net, demonstrating the promising performance in infarct core prediction. No significant difference of the prediction error is shown for the patients with follow-up NCCT and follow-up DWI (P >0.05).

CONCLUSION

This study provides an approach for fast and accurate stroke infarct core estimation. We anticipate the prediction results of ISP-Net could offer assistance to the physicians in the thrombolysis or thrombectomy therapy selection.

Development and clinical application of a deep learning model to identify acute infarct on magnetic resonance imaging

Stroke is a leading cause of death and disability. The ability to quickly identify the presence of acute infarct and quantify the volume on magnetic resonance imaging (MRI) has important treatment implications. We developed a machine learning model that used the apparent diffusion coefficient and diffusion weighted imaging series. It was trained on 6,657 MRI studies from Massachusetts General Hospital (MGH; Boston, USA). All studies were labelled positive or negative for infarct (classification annotation) with 377 having the region of interest outlined (segmentation annotation). The different annotation types facilitated training on more studies while not requiring the extensive time to manually segment every study. We initially validated the model on studies sequestered from the training set. We then tested the model on studies from three clinical scenarios: consecutive stroke team activations for 6-months at MGH, consecutive stroke team activations for 6-months at a hospital that did not provide training data (Brigham and Women’s Hospital [BWH]; Boston, USA), and an international site (Diagnósticos da América SA [DASA]; Brazil). The model results were compared to radiologist ground truth interpretations. The model performed better when trained on classification and segmentation annotations (area under the receiver operating curve [AUROC] 0.995 [95% CI 0.992–0.998] and median Dice coefficient for segmentation overlap of 0.797 [IQR 0.642–0.861]) compared to segmentation annotations alone (AUROC 0.982 [95% CI 0.972–0.990] and Dice coefficient 0.776 [IQR 0.584–0.857]). The model accurately identified infarcts for MGH stroke team activations (AUROC 0.964 [95% CI 0.943–0.982], 381 studies), BWH stroke team activations (AUROC 0.981 [95% CI 0.966–0.993], 247 studies), and at DASA (AUROC 0.998 [95% CI 0.993–1.000], 171 studies). The model accurately segmented infarcts with Pearson correlation comparing model output and ground truth volumes between 0.968 and 0.986 for the three scenarios. Acute infarct can be accurately detected and segmented on MRI in real-world clinical scenarios using a machine learning model.

A Fully Convolutional Network (FCN) based Automated Ischemic Stroke Segment Method using Chemical Exchange Saturation Transfer Imaging

BACKGROUND

Chemical exchange saturation transfer (CEST) MRI is a promising imaging modality in ischemic stroke detection for its sensitivity in sensing post-ischemic pH alteration. However, the accurate segmentation of pH-altered regions remains difficult due to the complicated sources in water signal changes of CEST MRI. Meanwhile, manual localization and quantification of stroke lesions are laborious and time-consuming, which cannot meet the urgent need for timely therapeutic interventions.

PURPOSE

The goal of this study was to develop an automatic lesion segmentation approach of ischemic region based on CEST MR images. A novel segmentation framework based on fully convolutional neural network was investigated for our task.

METHODS

Z-spectra from 10 rats were manually labeled as ground truth and split into two datasets, where the training dataset including 3 rats was used to generate a segmentation model, and the remaining rats were used as test datasets to evaluate the model's performance. Then a 1-D fully convolutional neural network equipped with bottleneck structures was set up, and a Grad-CAM approach was used to produce a coarse localization map, which can reflect the relevancy to the 'ischemia' class of each pixel.

RESULTS

As compared with the ground truth, the proposed network model achieved satisfying segmentation results with high values of evaluation metrics including specificity (SPE), sensitivity (SEN), accuracy (ACC), and Dice similarity coefficient (DSC), especially in some intractable situations where conventional MRI modalities and CEST quantitative method failed to distinguish between ischemic and normal tissues, and the model with augmentation was robust to input perturbations. The Grad-CAM maps performed clear tissue change distributions and interpreted the segmentations, and showed a strong correlation with the quantitative method, gave extended thinking to the function of networks.

CONCLUSIONS

The proposed method can segment ischemia region from CEST images, with the Grad-CAM maps give access to interpretative information about the segmentations, which demonstrates great potential in clinical routines. This article is protected by copyright. All rights reserved.

Automated Extraction of Cerebral Infarction Region in Head MR Image Using Pseudo Cerebral Infarction Image by CycleGAN

Since recognizing the location and extent of infarction is essential for diagnosis and treatment, many methods using deep learning have been reported. Generally, deep learning requires a large amount of training data. To overcome this problem, we generated pseudo patient images using CycleGAN, which performed image transformation without paired images. Then, we aimed to improve the extraction accuracy by using the generated images for the extraction of cerebral infarction regions. First, we used CycleGAN for data augmentation. Pseudo-cerebral infarction images were generated from healthy images using CycleGAN. Finally, U-Net was used to segment the cerebral infarction region using CycleGAN-generated images. Regarding the extraction accuracy, the Dice index was 0.553 for U-Net with CycleGAN, which was an improvement over U-Net without CycleGAN. Furthermore, the number of false positives per case was 3.75 for U-Net without CycleGAN and 1.23 for U-Net with CycleGAN, respectively. The number of false positives was reduced by approximately 67% by introducing the CycleGAN-generated images to training cases. These results indicate that utilizing CycleGAN-generated images was effective and facilitated the accurate extraction of the infarcted regions while maintaining the detection rate.

Ability of weakly supervised learning to detect acute ischemic stroke and hemorrhagic infarction lesions with diffusion-weighted imaging

BACKGROUND

Gradient-recalled echo (GRE) sequence is time-consuming and not routinely performed. Herein, we aimed to investigate the ability of weakly supervised learning to identify acute ischemic stroke (AIS) and concurrent hemorrhagic infarction based on diffusion-weighted imaging (DWI).

METHODS

First, we proposed spatially locating small stroke lesions in different positions and hemorrhagic infarction lesions by residual neural and visual geometry group networks using weakly supervised learning. Next, we compared the sensitivity and specificity for identifying automatically concurrent hemorrhagic infarction in stroke patients with the sensitivity and specificity of human readings of diffusion and b₀ images to evaluate the performance of the weakly supervised methods. Also, the labeling time of the weakly supervised approach was compared with that of the fully supervised approach.

RESULTS

Data from a total of 1,027 patients were analyzed. The residual neural network displayed a higher sensitivity than did the visual geometry group network in spatially locating the small stroke and hemorrhagic infarction lesions. The residual neural network had significantly greater patient-level sensitivity than did the human readers (98.4% versus 86.2%, P=0.008) in identifying concurrent hemorrhagic infarction with GRE as the reference standard; however, their specificities were comparable (95.4% versus 96.9%, P>0.99). Weak labeling of lesions required significantly less time than did full labeling of lesions (2.667 versus 10.115 minutes, P<0.001).

CONCLUSIONS

Weakly supervised learning was able to spatially locate small stroke lesions in different positions and showed more sensitivity than did human reading in identifying concurrent hemorrhagic infarction based on DWI. The proposed approach can reduce the labeling workload.

A Review on Computer Aided Diagnosis of Acute Brain Stroke

Amongst the most common causes of death globally, stroke is one of top three affecting over 100 million people worldwide annually. There are two classes of stroke, namely ischemic stroke (due to impairment of blood supply, accounting for ~70% of all strokes) and hemorrhagic stroke (due to bleeding), both of which can result, if untreated, in permanently damaged brain tissue. The discovery that the affected brain tissue (i.e., 'ischemic penumbra') can be salvaged from permanent damage and the bourgeoning growth in computer aided diagnosis has led to major advances in stroke management. Abiding to the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines, we have surveyed a total of 177 research papers published between 2010 and 2021 to highlight the current status and challenges faced by computer aided diagnosis (CAD), machine learning (ML) and deep learning (DL) based techniques for CT and MRI as prime modalities for stroke detection and lesion region segmentation. This work concludes by showcasing the current requirement of this domain, the preferred modality, and prospective research areas.

Co-optimization Learning Network for MRI Segmentation of Ischemic Penumbra Tissues

Convolutional neural networks (CNNs) have brought hope for the medical image auxiliary diagnosis. However, the shortfall of labeled medical image data is the bottleneck that limits the performance improvement of supervised CNN methods. In addition, annotating a large number of labeled medical image data is often expensive and time-consuming. In this study, we propose a co-optimization learning network (COL-Net) for Magnetic Resonance Imaging (MRI) segmentation of ischemic penumbra tissues. COL-Net base on the limited labeled samples and consists of an unsupervised reconstruction network (R), a supervised segmentation network (S), and a transfer block (T). The reconstruction network extracts the robust features from reconstructing pseudo unlabeled samples, which is the auxiliary branch of the segmentation network. The segmentation network is used to segment the target lesions under the limited labeled samples and the auxiliary of the reconstruction network. The transfer block is used to co-optimization the feature maps between the bottlenecks of the reconstruction network and segmentation network. We propose a mix loss function to optimize COL-Net. COL-Net is verified on the public ischemic penumbra segmentation challenge (SPES) with two dozen labeled samples. Results demonstrate that COL-Net has high predictive accuracy and generalization with the Dice coefficient of 0.79. The extended experiment also shows COL-Net outperforms most supervised segmentation methods. COL-Net is a meaningful attempt to alleviate the limited labeled sample problem in medical image segmentation.

Deep learning-based detection and segmentation of diffusion abnormalities in acute ischemic stroke

Background

Accessible tools to efficiently detect and segment diffusion abnormalities in acute strokes are highly anticipated by the clinical and research communities.

Methods

We developed a tool with deep learning networks trained and tested on a large dataset of 2,348 clinical diffusion weighted MRIs of patients with acute and sub-acute ischemic strokes, and further tested for generalization on 280 MRIs of an external dataset (STIR).

Results

Our proposed model outperforms generic networks and DeepMedic, particularly in small lesions, with lower false positive rate, balanced precision and sensitivity, and robustness to data perturbs (e.g., artefacts, low resolution, technical heterogeneity). The agreement with human delineation rivals the inter-evaluator agreement; the automated lesion quantification of volume and contrast has virtually total agreement with human quantification.

Conclusion

Our tool is fast, public, accessible to non-experts, with minimal computational requirements, to detect and segment lesions via a single command line. Therefore, it fulfills the conditions to perform large scale, reliable and reproducible clinical and translational research.

Automatic lung nodule segmentation and intra-nodular heterogeneity image generation

Automatic segmentation of lung nodules on computed tomography (CT) images is challenging owing to the variability of morphology, location, and intensity. In addition, few segmentation methods can capture intra-nodular heterogeneity to assist lung nodule diagnosis. In this study, we propose an end-to-end architecture to perform fully automated segmentation of multiple types of lung nodules and generate intra-nodular heterogeneity images for clinical use. To this end, a hybrid loss is considered by introducing a Faster R-CNN model based on generalized intersection over union loss in generative adversarial network. The Lung Image Database Consortium image collection dataset, comprising 2,635 lung nodules, was combined with 3,200 lung nodules from five hospitals for this study. Compared with manual segmentation by radiologists, the proposed model obtained an average dice coefficient (DC) of 82.05% on the test dataset. Compared with U-net, NoduleNet, nnU-net, and other three models, the proposed method achieved comparable performance on lung nodule segmentation and generated more vivid and valid intra-nodular heterogeneity images, which are beneficial in radiological diagnosis. In an external test of 91 patients from another hospital, the proposed model achieved an average DC of 81.61%. The proposed method effectively addresses the challenges of inevitable human interaction and additional pre-processing procedures in the existing solutions for lung nodule segmentation. In addition, the results show that the intra-nodular heterogeneity images generated by the proposed model are suitable to facilitate lung nodule diagnosis in radiology.

Temporal refinement of 3D CNN semantic segmentations on 4D time-series of undersampled tomograms using hidden Markov models

Recently, several convolutional neural networks have been proposed not only for 2D images, but also for 3D and 4D volume segmentation. Nevertheless, due to the large data size of the latter, acquiring a sufficient amount of training annotations is much more strenuous than in 2D images. For 4D time-series tomograms, this is usually handled by segmenting the constituent tomograms independently through time with 3D convolutional neural networks. Inter-volume information is therefore not utilized, potentially leading to temporal incoherence. In this paper, we attempt to resolve this by proposing two hidden Markov model variants that refine 4D segmentation labels made by 3D convolutional neural networks working on each time point. Our models utilize not only inter-volume information, but also the prediction confidence generated by the 3D segmentation convolutional neural networks themselves. To the best of our knowledge, this is the first attempt to refine 4D segmentations made by 3D convolutional neural networks using hidden Markov models. During experiments we test our models, qualitatively, quantitatively and behaviourally, using prespecified segmentations. We demonstrate in the domain of time series tomograms which are typically undersampled to allow more frequent capture; a particularly challenging problem. Finally, our dataset and code is publicly available.

A Novel Framework with Weighted Decision Map Based on Convolutional Neural Network for Cardiac MR Segmentation

For diagnosing cardiovascular disease, an accurate segmentation method is needed. There are several unre-solved issues in the complex field of cardiac magnetic resonance imaging, some of which have been partially addressed by using deep neural networks. To solve two problems of over-segmentation and under-segmentation of anatomical shapes in the short-axis view from different cardiac magnetic resonance sequences, we propose a novel two-stage framework with a weighted decision map based on convolutional neural networks to segment the myocardium (Myo), left ventricle (LV), and right ventricle (RV) simultaneously. The framework comprises a deci-sion map extractor and a cardiac segmenter. A cascaded U-Net++ is used as a decision map extractor to acquire the decision map that decides the category of each pixel. Cardiac segmenter is a multiscale dual-path feature ag-gregation network (MDFA-Net) which consists of a densely connected network and an asymmetric encoding and decoding network. The input to the cardiac seg-menter is derived from processed original images weighted by the output of the decision map extractor. We conducted experiments on two datasets of mul-ti-sequence cardiac magnetic resonance segmentation challenge 2019 (MS-CMRSeg 2019) and myocardial pa-thology segmentation challenge 2020 (MyoPS 2020). Test results obtained on MyoPS 2020 show that proposed method with average Dice coefficient of 84.70%, 86.00% and 86.31% in the segmentation task of Myo, LV, and RV, respectively.