Deep Learning-based Analysis Research Articles

Deep learning of echocardiography distinguishes between presence and absence of late gadolinium enhancement on cardiac magnetic resonance in patients with hypertrophic cardiomyopathy

BackgroundHypertrophic cardiomyopathy (HCM) can cause myocardial fibrosis, which can be a substrate for fatal ventricular arrhythmias and subsequent sudden cardiac death. Although late gadolinium enhancement (LGE) on cardiac magnetic resonance (CMR) represents myocardial fibrosis and is associated with sudden cardiac death in patients with HCM, CMR is resource-intensive, can carry an economic burden, and is sometimes contraindicated. In this study for patients with HCM, we aimed to distinguish between patients with positive and negative LGE on CMR using deep learning of echocardiographic images.MethodsIn the cross-sectional study of patients with HCM, we enrolled patients who underwent both echocardiography and CMR. The outcome was positive LGE on CMR. Among the 323 samples, we randomly selected 273 samples (training set) and employed deep convolutional neural network (DCNN) of echocardiographic 5-chamber view to discriminate positive LGE on CMR. We also developed a reference model using clinical parameters with significant differences between patients with positive and negative LGE. In the remaining 50 samples (test set), we compared the area under the receiver-operating-characteristic curve (AUC) between a combined model using the reference model plus the DCNN-derived probability and the reference model.ResultsAmong the 323 CMR studies, positive LGE was detected in 160 (50%). The reference model was constructed using the following 7 clinical parameters: family history of HCM, maximum left ventricular (LV) wall thickness, LV end-diastolic diameter, LV end-systolic volume, LV ejection fraction < 50%, left atrial diameter, and LV outflow tract pressure gradient at rest. The discriminant model combining the reference model with DCNN-derived probability significantly outperformed the reference model in the test set (AUC 0.86 [95% confidence interval 0.76–0.96] vs. 0.72 [0.57–0.86], P = 0.04). The sensitivity, specificity, positive predictive value, and negative predictive value of the combined model were 0.84, 0.76, 0.78, and 0.83, respectively.ConclusionCompared to the reference model solely based on clinical parameters, our new model integrating the reference model and deep learning-based analysis of echocardiographic images demonstrated superiority in distinguishing LGE on CMR in patients with HCM. The novel deep learning-based method can be used as an assistive technology to facilitate the decision-making process of performing CMR with gadolinium enhancement.

Deep learning-based analysis of EGFR mutation prevalence in lung adenocarcinoma H&E whole slide images.

EGFR mutations are a major prognostic factor in lung adenocarcinoma. However, current detection methods require sufficient samples and are costly. Deep learning is promising for mutation prediction in histopathological image analysis but has limitations in that it does not sufficiently reflect tumor heterogeneity and lacks interpretability. In this study, we developed a deep learning model to predict the presence of EGFR mutations by analyzing histopathological patterns in whole slide images (WSIs). We also introduced the EGFR mutation prevalence (EMP) score, which quantifies EGFR prevalence in WSIs based on patch-level predictions, and evaluated its interpretability and utility. Our model estimates the probability of EGFR prevalence in each patch by partitioning the WSI based on multiple-instance learning and predicts the presence of EGFR mutations at the slide level. We utilized a patch-masking scheduler training strategy to enable the model to learn various histopathological patterns of EGFR. This study included 868 WSI samples from lung adenocarcinoma patients collected from three medical institutions: Hallym University Medical Center, Inha University Hospital, and Chungnam National University Hospital. For the test dataset, 197 WSIs were collected from Ajou University Medical Center to evaluate the presence of EGFR mutations. Our model demonstrated prediction performance with an area under the receiver operating characteristic curve of 0.7680 (0.7607-0.7720) and an area under the precision-recall curve of 0.8391 (0.8326-0.8430). The EMP score showed Spearman correlation coefficients of 0.4705 (p = 0.0087) for p.L858R and 0.5918 (p = 0.0037) for exon 19 deletions in 64 samples subjected to next-generation sequencing analysis. Additionally, high EMP scores were associated with papillary and acinar patterns (p = 0.0038 and p = 0.0255, respectively), whereas low EMP scores were associated with solid patterns (p = 0.0001). These results validate the reliability of our model and suggest that it can provide crucial information for rapid screening and treatment plans.

Deep learning-based quantification of brain atrophy using 2D T1-weighted MRI for Alzheimer's disease classification.

Determining brain atrophy is crucial for the diagnosis of neurodegenerative diseases. Despite detailed brain atrophy assessments using three-dimensional (3D) T1-weighted magnetic resonance imaging, their practical utility is limited by cost and time. This study introduces deep learning algorithms for quantifying brain atrophy using a more accessible two-dimensional (2D) T1, aiming to achieve cost-effective differentiation of dementia of the Alzheimer's type (DAT) from cognitively unimpaired (CU), while maintaining or exceeding the performance obtained with T1-3D individuals and to accurately predict AD-specific atrophy similarity and atrophic changes [W-scores and Brain Age Index (BAI)]. Involving 924 participants (478 CU and 446 DAT), our deep learning models were trained on cerebrospinal fluid (CSF) volumes from 2D T1 images and compared with 3D T1 images. The performance of the models in differentiating DAT from CU was assessed using receiver operating characteristic analysis. Pearson's correlation analyses were used to evaluate the relations between 3D T1 and 2D T1 measurements of cortical thickness and CSF volumes, AD-specific atrophy similarity, W-scores, and BAIs. Our deep learning models demonstrated strong correlations between 2D and 3D T1-derived CSF volumes, with correlation coefficients r ranging from 0.805 to 0.971. The algorithms based on 2D T1 accurately distinguished DAT from CU with high accuracy (area under the curve values of 0.873), which were comparable to those of algorithms based on 3D T1. Algorithms based on 2D T1 image-derived CSF volumes showed high correlations in AD-specific atrophy similarity (r = 0.915), W-scores for brain atrophy (0.732 ≤ r ≤ 0.976), and BAIs (r = 0.821) compared with those based on 3D T1 images. Deep learning-based analysis of 2D T1 images is a feasible and accurate alternative for assessing brain atrophy, offering diagnostic precision comparable to that of 3D T1 imaging. This approach offers the advantage of the availability of T1-2D imaging, as well as reduced time and cost, while maintaining diagnostic precision comparable to T1-3D.

Disease Activity and Therapeutic Response to Pegcetacoplan for Geographic Atrophy Identified by Deep Learning-Based Analysis of OCT

PurposeTo quantify morphological changes of the photoreceptors (PR) and retinal pigment epithelium (RPE) layers under pegcetacoplan therapy in geographic atrophy (GA) using deep learning-based analysis of optical coherence tomography (OCT) images. DesignPost-hoc longitudinal image analysis SubjectsPatients with GA due to age-related macular degeneration from two prospective randomized phase III clinical trials (OAKS and DERBY) MethodsDeep learning-based segmentation of RPE loss and PR degeneration, defined as loss of the ellipsoid zone (EZ) layer on OCT, over 24 months on SD-OCT images Main outcome measuresChange in the mean area of RPE loss and EZ loss over time in the pooled sham arms and the monthly (PM)/every other month (PEOM) treatment arms Results897 eyes of 897 patients were included. There was a therapeutic reduction of RPE loss growth by 22%/20% in OAKS and 27%/21% in DERBY for PM/PEOM compared to sham, respectively, at 24 months. The reduction on the EZ level was significantly higher with 53%/46% in OAKS and 47%/46% in DERBY for PM/PEOM compared to sham at 24 months. The baseline EZ-RPE difference had an impact on disease activity and therapeutic response. The therapeutic benefit for RPE loss growth increased with larger EZ-RPE difference quartiles from 21.9%, 23.1%, 23.9% to 33.6% for PM vs. sham (all p<0.01) and from 13.6% (p=0.11), 23.8%, 23.8% to 20.0% for PEOM vs. sham (p<0.01) in quartiles 1,2,3 and 4, respectively, at 24 months. Regarding EZ layer maintenance, the therapeutic reduction of loss increased from 14.8% (p=0.09), 33.3%, 46.6% to 77.8% (p<0.0001) between PM and sham and from 15.9% (p=0.08), 33.8%, 52.0% to 64.9% (p<0.0001) between PEOM and sham for quartiles 1-4 at 24 months. ConclusionOCT-based AI analysis objectively identifies and quantifies PR and RPE degeneration in GA. Reductions in further PR degeneration consistent with EZ loss on OCT are even higher than the effect on RPE loss in phase 3 trials of pegcetacoplan treatment. The EZ-RPE difference has a strong impact on disease progression and therapeutic response. Identification of patients with higher EZ-RPE loss difference may become an important criterion for the management of GA secondary to AMD.

Mitochondrial Ultrastructure and Its Regulation Revealed by Deep Learning-based Analysis

Mitochondrial Ultrastructure and Its Regulation Revealed by Deep Learning-based Analysis

Deep Learning-Based Analysis of Emotional and Content Relevance Between Bullet Screens and Subtitles as Movie Narrative Medium

This study borrows elements of film text in the field of film studies to develop the potential of data analysis using AI algorithms, while aiming to explore the emotional and content relevances between the two narrative mediums in film, bullet screens and subtitles, as well as the narrative effect variations that they bring about. The film sample Forever Young was analyzed by deep learning, data analysis, and knowledge discovery algorithms and tools. The BERT model was used to quantify the intensity of emotions and conduct fine-grained emotional classification of film characters. Emotion curves are plotted based on time series data. Thematic words are extracted by the LDA algorithm, and text similarity is computed using Word2Vec. The study found differences in emotional expression between bullet screens and subtitles, and they also exhibit strong emotional and content relevances. Yet, this correlation is not only influenced by temporal deviation factors but also by specific emotional types, different stages of story development, and the collective emotions of online views. The conclusion suggests that the effective collection of plot metronomes in film narratives helps reveal crucial plot nodes and emotional trends in films that cause obvious emotional changes, and placing them appropriately is fundamental for continuously attracting and retaining viewers’ attention. The interaction between bullet comments and subtitles is not only about emotional expression but also enhances the effectiveness of the film narrative. Overall, this study demonstrates the importance of database logic and technology in understanding and expressing film emotions and meanings, and providing new ideas and methods for enhancing film narrative effectiveness.

Deep learning-based analysis of true triaxial DEM simulations: Role of fabric and particle aspect ratio

This study investigates the influence of micro-scale entities such as inherent and induced fabric anisotropy on the stress–strain behaviour of granular assemblies. In tandem with this exploration, our objective is to formulate a novel correlation that quantifies the evolution of fabric tensor across diverse loading paths. This correlation can be introduced to enhance the micro-mechanical insights in conventional constitutive models. Employing the Discrete Element Method (DEM), we simulate the drained and undrained responses of 680 transversely isotropic particulate assemblies with diverse initial fabrics and particle Aspect Ratio (AR) under true triaxial loading conditions. We consider a second-order fabric tensor based on inter-particle contact orientations to trace fabric evolution during loading. To account for fluid–solid interaction under undrained conditions, we adopted the DEM-Coupled Fluid Method (CFM). The simulation results highlight the significant influence of the Lode angle, particle shape and initial fabric on the stress–strain behaviour of granular materials, with fabric evolution primarily affected by the stress state, void ratio and particle AR. Lastly, we propose a new correlation for the quantification of the fabric tensor using a multi-layer feed-forward neural network. The satisfactory performance of the suggested correlation is demonstrated through a comparison between DEM data and predicted fabric tensor values.

Automated detection of type 1 ROP, type 2 ROP and A-ROP based on deep learning

PurposeTo provide automatic detection of Type 1 retinopathy of prematurity (ROP), Type 2 ROP, and A-ROP by deep learning-based analysis of fundus images obtained by clinical examination using convolutional neural networks.Material and methodsA total of 634 fundus images of 317 premature infants born at 23–34 weeks of gestation were evaluated. After image pre-processing, we obtained a rectangular region (ROI). RegNetY002 was used for algorithm training, and stratified 10-fold cross-validation was applied during training to evaluate and standardize our model. The model’s performance was reported as accuracy and specificity and described by the receiver operating characteristic (ROC) curve and area under the curve (AUC).ResultsThe model achieved 0.98 accuracy and 0.98 specificity in detecting Type 2 ROP versus Type 1 ROP and A-ROP. On the other hand, as a result of the analysis of ROI regions, the model achieved 0.90 accuracy and 0.95 specificity in detecting Stage 2 ROP versus Stage 3 ROP and 0.91 accuracy and 0.92 specificity in detecting A-ROP versus Type 1 ROP. The AUC scores were 0.98 for Type 2 ROP versus Type 1 ROP and A-ROP, 0.85 for Stage 2 ROP versus Stage 3 ROP, and 0.91 for A-ROP versus Type 1 ROP.ConclusionOur study demonstrated that ROP classification by DL-based analysis of fundus images can be distinguished with high accuracy and specificity. Integrating DL-based artificial intelligence algorithms into clinical practice may reduce the workload of ophthalmologists in the future and provide support in decision-making in the management of ROP.

NFS-23. ASSOCIATION OF LOCALIZED MAGNETIC RESONANCE IMAGING FEATURES IN ANTERIOR VISUAL PATHWAY WITH VISUAL ACUITY LOSS AMONG CHILDREN WITH NF1-OPG

Abstract BACKGROUND Half of children with optic pathway gliomas associated with neurofibromatosis type 1 (NF1-OPG) are at risk for vision acuity (VA) loss. NF1-OPGs manifest along the anterior visual pathway (AVP), where the overall shape and volume from MRI have been recently associated with VA loss. To investigate the added value of localized features within the optic nerves (ONs) and chiasm, we propose an automatic deep-learning framework for VA loss prediction from volumetric and radiomic analysis of ONs, chiasm, and AVP. METHODS We collected MRIs of 60 children with NF1-OPG from Children’s National Hospital (CNH, GE platform) and 75 from Children’s Hospital of Philadelphia (CHOP, Siemens platform). MRI sequences included T1-, T2-weighted and T2-FLAIR. Experts annotated the AVP to establish ground truth. Neuro-ophthalmic evaluation identified VA loss (LogMAR &amp;gt;= 0.2) in 52/135 children. Automatic AVP segmentation was performed using the Swin transformer network. ONs and chiasm were split through template-based registration. Brain volume, tumor location, child age, and 1,172 radiomic features were used to predict VA loss using support vector machines. Sequential feature selection associated with risk VA loss was done using sensitivity and univariate statistical tests (ANOVA). We compared predictions based on new ON and chiasm analysis with results based on AVP alone. RESULTS Balanced accuracy, sensitivity, specificity, and AUROC of VA loss prediction were 0.865, 0.882, 0.857, 0.899, respectively. Significant risk factors were maximal image intensity and gray level run length entropy in T2-FLAIR for ONs, and image intensity range in T2-FLAIR for chiasm. AUROC based on analysis of AVP alone dropped to 0.728 (p-value=0.028). CONCLUSIONS Deep learning-based analysis indicates an association of new localized MRI features in ONs and chiasm with VA loss, which show enhanced capability in predicting VA loss. This automated framework has the potential to accelerate and guide the treatment decisions for children with NF1-OPGs.

Toward robust and high-throughput detection of seed defects in X-ray images via deep learning

BackgroundThe detection of internal defects in seeds via non-destructive imaging techniques is a topic of high interest to optimize the quality of seed lots. In this context, X-ray imaging is especially suited. Recent studies have shown the feasibility of defect detection via deep learning models in 3D tomography images. We demonstrate the possibility of performing such deep learning-based analysis on 2D X-ray radiography for a faster yet robust method via the X-Robustifier pipeline proposed in this article.Results2D X-ray images of both defective and defect-free seeds were acquired. A deep learning model based on state-of-the-art object detection neural networks is proposed. Specific data augmentation techniques are introduced to compensate for the low ratio of defects and increase the robustness to variation of the physical parameters of the X-ray imaging systems. The seed defects were accurately detected (F1-score >90%), surpassing human performance in computation time and error rates. The robustness of these models against the principal distortions commonly found in actual agro-industrial conditions is demonstrated, in particular, the robustness to physical noise, dimensionality reduction and the presence of seed coating.ConclusionThis work provides a full pipeline to automatically detect common defects in seeds via 2D X-ray imaging. The method is illustrated on sugar beet and faba bean and could be efficiently extended to other species via the proposed generic X-ray data processing approach (X-Robustifier). Beyond a simple proof of feasibility, this constitutes important results toward the effective use in the routine of deep learning-based automatic detection of seed defects.

Toward confident prostate cancer detection using ultrasound: a multi-center study.

Deep learning-based analysis of micro-ultrasound images to detect cancerous lesions is a promising tool for improving prostate cancer (PCa) diagnosis. An ideal model should confidently identify cancer while responding with appropriate uncertainty when presented with out-of-distribution inputs that arise during deployment due to imaging artifacts and the biological heterogeneity of patients and prostatic tissue. Using micro-ultrasound data from 693 patients across 5 clinical centers who underwent micro-ultrasound guided prostate biopsy, we train and evaluate convolutional neural network models for PCa detection. To improve robustness to out-of-distribution inputs, we employ and comprehensively benchmark several state-of-the-art uncertainty estimation methods. PCa detection models achieve performance scores up to average AUROC with a 10-fold cross validation setup. Models with uncertainty estimation obtain expected calibration error scores as low as , indicating that confident predictions are very likely to be correct. Visualizations of the model output demonstrate that the model correctly identifies healthy versus malignant tissue. Deep learning models have been developed to confidently detect PCa lesions from micro-ultrasound. The performance of these models, determined from a large and diverse dataset, is competitive with visual analysis of magnetic resonance imaging, the clinical benchmark to identify PCa lesions for targeted biopsy. Deep learning with micro-ultrasound should be further studied as an avenue for targeted prostate biopsy.

Deep Machine Learning-based Analysis for Intelligent Phonetic Language Recognition

Modern speech generating systems can produce results that are almost as visually realistic as actual sounds. They still require further production management. This research presents a paradigm for managing prosodic output using explicit, unambiguous, and understandable parameters. We utilize this strategy to emphasize key words and provide a variety of architectural possibilities based on a richness of labelled resources. In an objective voice, we compare the options for producing data with or without labels. We assess them using listening tests that demonstrate our ability to retain the same level of naturalness while effectively attaining regulated concentration over a specific area.

Deep Learning-based marine species detection and classification framework for biomonitoring in the Arctic fjords, Svalbard

The effects of ongoing climate change have caused a poleward shift in the distribution of species due to the rapidly rising water temperatures. This calls for an immediate need to assess and document the extent of climate change-driven animal migrations occurring in the Arctic waters. However, the extreme climatic conditions and the remoteness of the region makes biomonitoring tedious in the Arctic ecosystem. The present study puts forward a deep learning-based analysis of a large underwater video dataset that was captured from the Arctic region. The dataset was acquired using underwater cameras mounted on custom-made stainless-steel frames. The video footages were collected over a period of 26 days from the Kongsfjorden- Krossfjorden twin Arctic fjords in Svalbard, Norway. The collected data sets were used to train YOLO-based object detection framework (You Only Look Once) for an automated detection of the organisms. The YOLO model employed for the study was found to be very efficient in classifying the underwater images captured from the region. The object detection framework could detect images of Comb jelly, Echinoderm, Sea Anemone and Ulke (Shorthorn sculpin) from the underwater images. The model attained a superior value of Mean Average Precision (mAP), precision, and recall of 99.5%, 99.2%, and 97.4%, respectively.

Deep learning-based analysis of basins of attraction.

This research addresses the challenge of characterizing the complexity and unpredictability of basins within various dynamical systems. The main focus is on demonstrating the efficiency of convolutional neural networks (CNNs) in this field. Conventional methods become computationally demanding when analyzing multiple basins of attraction across different parameters of dynamical systems. Our research presents an innovative approach that employs CNN architectures for this purpose, showcasing their superior performance in comparison to conventional methods. We conduct a comparative analysis of various CNN models, highlighting the effectiveness of our proposed characterization method while acknowledging the validity of prior approaches. The findings not only showcase the potential of CNNs but also emphasize their significance in advancing the exploration of diverse behaviors within dynamical systems.

Crossroads Traffic Data Analysis Using Drone-Based Deep Learning

This study proposes a method for monitoring traffic conditions via the collection of road traffic data using drones and deep learning-based analysis. Utilizing the YOLOv8 object detection model, we analyzed traffic videos captured at 45 and 90 degrees at intersections, which revealed differences in object detection capabilities based on the shooting angle. To address issues related to occluded areas during object tracking, we optimized the shooting angles, and achieved a final model with an average of 96.9% and 93.7% path and vehicle type recognition rate, respectively. This validates the effective utilization of drone-based traffic data collection and deep learning analysis in road traffic management. Further research is necessary for the development and application of more sophisticated traffic information systems in the future.

Deep-learning image analysis for high-throughput screening of opsono-phagocytosis-promoting monoclonal antibodies against Neisseria gonorrhoeae

Antimicrobial resistance (AMR) is nowadays a global health concern as bacterial pathogens are increasingly developing resistance to antibiotics. Monoclonal antibodies (mAbs) represent a powerful tool for addressing AMR thanks to their high specificity for pathogenic bacteria which allows sparing the microbiota, kill bacteria through complement deposition, enhance phagocytosis or inhibit bacterial adhesion to epithelial cells. Here we describe a visual opsono-phagocytosis assay which relies on confocal microscopy to measure the impact of mAbs on phagocytosis of the bacterium Neisseria gonorrhoeae by macrophages. With respect to traditional CFU-based assays, generated images can be automatically analysed by convolutional neural networks. Our results demonstrate that confocal microscopy and deep learning-based analysis allow screening for phagocytosis-promoting mAbs against N. gonorrhoeae, even when mAbs are not purified and are expressed at low concentration. Ultimately, the flexibility of the staining protocol and of the deep-learning approach make the assay suitable for other bacterial species and cell lines where mAb activity needs to be investigated.

Prediction of Biliary Complications After Human Liver Transplantation Using Hyperspectral Imaging and Convolutional Neural Networks: A Proof-of-concept Study.

Biliary complications (BCs) negatively impact the outcome after liver transplantation. We herein tested whether hyperspectral imaging (HSI) generated data from bile ducts (BD) on reperfusion and machine learning techniques for data readout may serve as a novel approach for predicting BC. Tissue-specific data from 136 HSI liver images were integrated into a convolutional neural network (CNN). Fourteen patients undergoing liver transplantation after normothermic machine preservation served as a validation cohort. Assessment of oxygen saturation, organ hemoglobin, and tissue water levels through HSI was performed after completing the biliary anastomosis. Resected BD segments were analyzed by immunohistochemistry and real-time confocal microscopy. Immunohistochemistry and real-time confocal microscopy revealed mild (grade I: 1%-40%) BD damage in 8 patients and moderate (grade II: 40%-80%) injury in 1 patient. Donor and recipient data alone had no predictive capacity toward BC. Deep learning-based analysis of HSI data resulted in >90% accuracy of automated detection of BD. The CNN-based analysis yielded a correct classification in 72% and 69% for BC/no BC. The combination of HSI with donor and recipient factors showed 94% accuracy in predicting BC. Deep learning-based modeling using CNN of HSI-based tissue property data represents a noninvasive technique for predicting postoperative BC.

Deep Learning-Based Analysis of Aortic Morphology From Three-Dimensional MRI.

Quantification of aortic morphology plays an important role in the evaluation and follow-up assessment of patients with aortic diseases, but often requires labor-intensive and operator-dependent measurements. Automatic solutions would help enhance their quality and reproducibility. To design a deep learning (DL)-based automated approach for aortic landmarks and lumen detection derived from three-dimensional (3D) MRI. Retrospective. Three hundred ninety-one individuals (female: 47%, age = 51.9 ± 18.4) from three sites, including healthy subjects and patients (hypertension, aortic dilation, Turner syndrome), randomly divided into training/validation/test datasets (N = 236/77/78). Twenty-five subjects were randomly selected and analyzed by three operators with different levels of expertise. 1.5-T and 3-T, 3D spoiled gradient-recalled or steady-state free precession sequences. Reinforcement learning and a two-stage network trained using reference landmarks and segmentation from an existing semi-automatic software were used for aortic landmark detection and segmentation from sinotubular junction to coeliac trunk. Aortic segments were defined using the detected landmarks while the aortic centerline was extracted from the segmentation and morphological indices (length, aortic diameter, and volume) were computed for both the reference and the proposed segmentations. Segmentation: Dice similarity coefficient (DSC), Hausdorff distance (HD), average symmetrical surface distance (ASSD); landmark detection: Euclidian distance (ED); model robustness: Spearman correlation, Bland-Altman analysis, Kruskal-Wallis test for comparisons between reference and DL-derived aortic indices; inter-observer study: Williams index (WI). A WI 95% confidence interval (CI) lower bound >1 indicates that the method is within the inter-observer variability. A P-value <0.05 was considered statistically significant. DSC was 0.90 ± 0.05, HD was 12.11 ± 7.79 mm, and ASSD was 1.07 ± 0.63 mm. ED was 5.0 ± 6.1 mm. A good agreement was found between all DL-derived and reference aortic indices (r >0.95, mean bias <7%). Our segmentation and landmark detection performances were within the inter-observer variability except the sinotubular junction landmark (CI = 0.96;1.04). A DL-based aortic segmentation and anatomical landmark detection approach was developed and applied to 3D MRI data for achieve aortic morphology evaluation. 3 TECHNICAL EFFICACY: Stage 2.

Network response of brain microvasculature to neuronal stimulation

Neurovascular coupling (NVC), or the adjustment of blood flow in response to local increases in neuronal activity is a hallmark of healthy brain function, and the physiological foundation for functional magnetic resonance imaging (fMRI). However, it remains only partly understood due to the high complexity of the structure and function of the cerebrovascular network. Here we set out to understand NVC at the network level, i.e. map cerebrovascular network reactivity to activation of neighbouring neurons within a 500×500×500 μm3 cortical volume (∼30 high-resolution 3-nL fMRI voxels). Using 3D two-photon fluorescence microscopy data, we quantified blood volume and flow changes in the brain vessels in response to spatially targeted optogenetic activation of cortical pyramidal neurons. We registered the vessels in a series of image stacks acquired before and after stimulations and applied a deep learning pipeline to segment the microvascular network from each time frame acquired. We then performed image analysis to extract the microvascular graphs, and graph analysis to identify the branch order of each vessel in the network, enabling the stratification of vessels by their branch order, designating branches 1–3 as precapillary arterioles and branches 4+ as capillaries. Forty-five percent of all vessels showed significant calibre changes; with 85 % of responses being dilations. The largest absolute CBV change was in the capillaries; the smallest, in the venules. Capillary CBV change was also the largest fraction of the total CBV change, but normalized to the baseline volume, arterioles and precapillary arterioles showed the biggest relative CBV change. From linescans along arteriole-venule microvascular paths, we measured red blood cell velocities and hematocrit, allowing for estimation of pressure and local resistance along these paths. While diameter changes following neuronal activation gradually declined along the paths; the pressure drops from arterioles to venules increased despite decreasing resistance: blood flow thus increased more than local resistance decreases would predict. By leveraging functional volumetric imaging and high throughput deep learning-based analysis, our study revealed distinct hemodynamic responses across the vessel types comprising the microvascular network. Our findings underscore the need for large, dense sampling of brain vessels for characterization of neurovascular coupling at the network level in health and disease.

Bearing fault diagnosis with parallel CNN and LSTM.

Intelligent diagnosis of bearing faults is fundamental to machinery automation and their intelligent operation. Deep learning-based analysis of bearing vibration data has emerged as one research mainstream for fault diagnosis. To enhance the quality of feature extraction from bearing vibration signals and the robustness of the model, we construct a fault diagnostic model based on convolutional neural network (CNN) and long short-term memory (LSTM) parallel network to extract their temporal and spatial features from two perspectives. First, via resampling, vibration signal is split into equal-sized slices which are then converted into time-frequency images by continuous wavelet transform (CWT). Second, LSTM extracts the time-correlation features of 1D signals as one path, and 2D-CNN extracts the local frequency distribution features of time-frequency images as another path. Third, 1D-CNN further extracts integrated features from the fusion features yielded by former parallel paths. Finally, these categories are calculated through the softmax function. According to experimental results, the proposed model has satisfactory diagnostic accuracy and robustness in different contexts on two different datasets.