A dynamic brain network decomposition method discovers effective brain hemodynamic sub-networks for Parkinson's disease.

Objective.Dopaminergic treatment is effective for Parkinson's disease (PD). Nevertheless, the conventional treatment assessment mainly focuses on human-administered behavior examination while the underlying functional improvements have not been well explored. This paper aims to investigate brain functional variations of PD patients after dopaminergic therapy.Approach.This paper proposed a dynamic brain network decomposition method and discovered brain hemodynamic sub-networks that well characterized the efficacy of dopaminergic treatment in PD. Firstly, a clinical walking procedure with functional near-infrared spectroscopy was developed, and brain activations during the procedure from fifty PD patients under the OFF and ON states (without and with dopaminergic medication) were captured. Then, dynamic brain networks were constructed with sliding-window analysis of phase lag index and integrated time-varying functional networks across all patients. Afterwards, an aggregated network decomposition algorithm was formulated based on aggregated effectiveness optimization of functional networks in spanning network topology and cross-validation network variations, and utilized to unveil effective brain hemodynamic sub-networks for PD patients. Further, dynamic sub-network features were constructed to characterize the brain flexibility and dynamics according to the temporal switching and activation variations of discovered sub-networks, and their correlations with differential treatment-induced gait alterations were analyzed.Results.The results demonstrated that PD patients exhibited significantly enhanced flexibility after dopaminergic therapy within a sub-network related to the improvement of motor functions. Other sub-networks were significantly correlated with trunk-related axial symptoms and exhibited no significant treatment-induced dynamic interactions.Significance.The proposed method promises a quantified and objective approach for dopaminergic treatment evaluation. Moreover, the findings suggest that the gait of PD patients comprises distinct motor domains, and the corresponding neural controls are selectively responsive to dopaminergic treatment.

Brain Temporal-Spectral Functional Variability Reveals Neural Improvements of DBS Treatment for Disorders of Consciousness.

Deep brain stimulation (DBS) is establishing itself as a promising treatment for disorders of consciousness (DOC). Measuring consciousness changes is crucial in the optimization of DBS therapy for DOC patients. However, conventional measures use subjective metrics that limit the investigations of treatment-induced neural improvements. The focus of this study is to analyze the regulatory effects of DBS and explain the regulatory mechanism at the brain functional level for DOC patients. Specifically, this paper proposed a dynamic brain temporal-spectral analysis method to quantify DBS-induced brain functional variations in DOC patients. Functional near-infrared spectroscopy (fNIRS) that promised to evaluate consciousness levels was used to monitor brain variations of DOC patients. Specifically, a fNIRS-based experimental procedure with auditory stimuli was developed, and the brain activities during the procedure from thirteen DOC patients before and after the DBS treatment were recorded. Then, dynamic brain functional networks were formulated with a sliding-window correlation analysis of phase lag index. Afterwards, with respect to the temporal variations of global and regional networks, the variability of global efficiency, local efficiency, and clustering coefficient were extracted. Further, dynamic networks were converted into spectral representations by graph Fourier transform, and graph energy and diversity were formulated to assess the spectral global and regional variability. The results showed that DOC patients under DBS treatment exhibited increased global and regional functional variability that was significantly associated with consciousness improvements. Moreover, the functional variability in the right brain regions had a stronger correlation with consciousness enhancements than that in the left brain regions. Therefore, the proposed method well signifies DBS-induced brain functional variations in DOC patients, and the functional variability may serve as promising biomarkers for consciousness evaluations in DOC patients.

Subject-Specific Modeling of EEG-fNIRS Neurovascular Coupling by Task-Related Tensor Decomposition.

Neurovascular coupling (NVC) connects neural activity with hemodynamics and plays a vital role in sustaining brain function. Combining electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) is a promising way to explore the NVC. However, the high-order property of EEG data and variability of hemodynamic response function (HRF) across subjects have not been well considered in existing NVC studies. In this study, we proposed a novel framework to enhance the subject-specific parametric modeling of NVC from simultaneous EEG-fNIRS measurement. Specifically, task-related tensor decomposition of high-order EEG data was performed to extract the underlying connections in the temporal-spectral-spatial structures of EEG activities and identify the most relevant temporal signature within multiple trials. Subject-specific HRFs were estimated by parameters optimization of a double gamma function model. A canonical motor task experiment was designed to induce neural activity and validate the effectiveness of the proposed framework. The results indicated that the proposed framework significantly improves the reproducibility of EEG components and the correlation between the predicted hemodynamic activities and the real fNIRS signal. Moreover, the estimated parameters characterized the NVC differences in the task with two speeds. Therefore, the proposed framework provides a feasible solution for the quantitative assessment of the NVC function.

fNIRS-based graph frequency analysis to identify mild cognitive impairment in Parkinson's disease.

BACKGROUND: Early identification of mild cognitive impairment (MCI) is essential for its treatment and the prevention of dementia in Parkinson's disease (PD). Existing approaches are mostly based on neuropsychological assessments, while brain activation and connection have not been well considered. NEW METHOD: This paper presents a neuroimaging-based graph frequency analysis method and the generated features to quantify the brain functional neurodegeneration and distinguish between PD-MCI patients and healthy controls. The Stroop color-word experiment was conducted with 20 PD-MCI patients and 34 healthy controls, and the brain activation was recorded with functional near-infrared spectroscopy (fNIRS). Then, the functional brain network was constructed based on Pearson's correlation coefficient calculation between every two fNIRS channels. Next, the functional brain network was represented as a graph and decomposed in the graph frequency domain through the graph Fourier transform (GFT) to obtain the eigenvector matrix. Total variation and weighted zero crossings of eigenvectors were defined and integrated to quantify functional interaction between brain regions and the spatial variability of the brain network in specific graph frequency ranges, respectively. After that, the features were employed in training a support vector machine (SVM) classifier. RESULTS: The presented method achieved a classification accuracy of 0.833 and an F1 score of 0.877, significantly outperforming existing methods and features. COMPARISON WITH EXISTING METHODS: Our method provided improved classification performance in the identification of PD-MCI. CONCLUSION: The results suggest that the presented graph frequency analysis method well identify PD-MCI patients and the generated features promise functional brain biomarkers for PD-MCI diagnosis.

Effective DBS treatment improves neural information transmission of patients with disorders of consciousness: an fNIRS study.

Objective.Deep brain stimulation (DBS) is a potential treatment that promotes the recovery of patients with disorders of consciousness (DOC). This study quantified the changes in consciousness and the neuromodulation effect of DBS on patients with DOC.Approach.Eleven patients were recruited for this study which consists of three conditions: 'Pre' (two days before DBS surgery), 'Post-On' (one month after surgery with stimulation), and 'Post-Off' (one month after surgery without stimulation). Functional near-infrared spectroscopy (fNIRS) was recorded from the frontal lobe, parietal lobe, and occipital lobe of patients during the experiment of auditory stimuli paradigm, in parallel with the coma recovery scale-revised (CRS-R) assessment. The brain hemodynamic states were defined and state transition acceleration was taken to quantify the information transmission strength of the brain network. Linear regression analysis was conducted between the changes in regional and global indicators and the changes in the CRS-R index.Main results.Significant correlation was observed between the changes in the global transition acceleration indicator and the changes in the CRS-R index (slope = 55.910,p< 0.001,R2= 0.732). For the regional indicators, similar correlations were found between the changes in the frontal lobe and parietal lobe indicators and the changes in the CRS-R index (slope = 46.612,p< 0.01,R2= 0.694; slope = 47.491,p< 0.01,R2= 0.676).Significance.Our study suggests that fNIRS-based brain hemodynamics transition analysis can signify the neuromodulation effect of DBS treatment on patients with DOC, and the transition acceleration indicator is a promising brain functional marker for DOC.

Coarse and Fine Two-Stage Calibration Method for Enhancing the Accuracy of Inverse Finite Element Method.

The inverse finite element method (iFEM) is a novel method for reconstructing the full-field displacement of structures by discrete measurement strain. In practical engineering applications, the accuracy of iFEM is reduced due to the positional offset of strain sensors during installation and errors in structural installation. Therefore, a coarse and fine two-stage calibration (CFTSC) method is proposed to enhance the accuracy of the reconstruction of structures. Firstly, the coarse calibration is based on a single-objective particle swarm optimization algorithm (SOPSO) to optimize the displacement-strain transformation matrix related to the sensor position. Secondly, as selecting different training data can affect the training effect of self-constructed fuzzy networks (SCFN), this paper proposes to screen the appropriate training data based on residual analysis. Finally, the experiments of the wing-integrated antenna structure verify the efficiency of the method on the reconstruction accuracy of the structural body displacement field.

An EEG-fNIRS neurovascular coupling analysis method to investigate cognitive-motor interference.

BACKGROUND AND OBJECTIVE: The simultaneous execution of a motor and cognitive dual task may lead to the deterioration of task performance in one or both tasks due to cognitive-motor interference (CMI). Neuroimaging techniques are promising ways to reveal the underlying neural mechanism of CMI. However, existing studies have only explored CMI from a single neuroimaging modality, which lack built-in validation and comparison of analysis results. This work is aimed to establish an effective analysis framework to comprehensively investigate the CMI by exploring the electrophysiological and hemodynamic activities as well as their neurovascular coupling. METHODS: Experiments including an upper limb single motor task, single cognitive task, and cognitive-motor dual task were designed and performed with 16 healthy young participants. Bimodal signals of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) were recorded simultaneously during the experiments. A novel bimodal signal analysis framework was proposed to extract the task-related components for EEG and fNIRS signals respectively and analyze their correlation. Indicators including within-class similarity and between-class distance were utilized to validate the effectiveness of the proposed analysis framework compared to the canonical channel-averaged method. Statistical analysis was performed to investigate the difference in the behavior and neural correlates between the single and dual tasks. RESULTS: Our results revealed that the extra cognitive interference caused divided attention in the dual task, which led to the decreased neurovascular coupling between fNIRS and EEG in all theta, alpha, and beta rhythms. The proposed framework was demonstrated to have a better ability in characterizing the neural patterns than the canonical channel-averaged method with significantly higher within-class similarity and between-class distance indicators. CONCLUSIONS: This study proposed a method to investigate CMI by exploring the task-related electrophysiological and hemodynamic activities as well as their neurovascular coupling. Our concurrent EEG-fNIRS study provides new insight into the EEG-fNIRS correlation analysis and novel evidence for the mechanism of neurovascular coupling in the CMI.

An fNIRS-Based Dynamic Functional Connectivity Analysis Method to Signify Functional Neurodegeneration of Parkinson's Disease.

Parkinson's disease (PD) is a prevalent brain disorder, and PD diagnosis is crucial for treatment. Existing methods for PD diagnosis are mainly focused on behavior analysis, while the functional neurodegeneration of PD has not been well investigated. This paper proposes a method to signify functional neurodegeneration of PD with dynamic functional connectivity analysis. A functional near-infrared spectroscopy (fNIRS)-based experimental paradigm was designed to capture brain activation from 50 PD patients and 41 age-matched healthy controls in clinical walking tests. Dynamic functional connectivity was constructed with sliding-window correlation analysis, and k-means clustering was applied to generate the key brain connectivity states. Dynamic state features including state occurrence probability, state transition percentage and state statistical features were extracted to quantify the variations of brain functional networks. A support vector machine was trained to classify PD patients and healthy controls. Statistical analysis was conducted to investigate the difference between PD patients and healthy controls as well as the relationship between dynamic state features and the MDS-UPDRS sub-score of gait. The results showed that PD patients had a higher probability of transiting to brain connectivity states with high levels of information transmission compared with healthy controls. The MDS-UPDRS sub-score of gait and the dynamics state features showed a significant correlation. Moreover, the proposed method had better classification performances than the available fNIRS-based methods in terms of accuracy and F1 score. Thus, the proposed method well signified functional neurodegeneration of PD, and the dynamic state features may serve as promising functional biomarkers for PD diagnosis.

Increased Effective Connectivity of the Left Parietal Lobe During Walking Tasks in Parkinson's Disease.

BACKGROUND: In Parkinson's disease (PD), walking may depend on the activation of the cerebral cortex. Understanding the patterns of interaction between cortical regions during walking tasks is of great importance. OBJECTIVE: This study investigated differences in the effective connectivity (EC) of the cerebral cortex during walking tasks in individuals with PD and healthy controls. METHODS: We evaluated 30 individuals with PD (62.4±7.2 years) and 22 age-matched healthy controls (61.0±6.4 years). A mobile functional near-infrared spectroscopy (fNIRS) was used to record cerebral oxygenation signals in the left prefrontal cortex (LPFC), right prefrontal cortex (RPFC), left parietal lobe (LPL), and right parietal lobe (RPL) and analyze the EC of the cerebral cortex. A wireless movement monitor was used to measure the gait parameters. RESULTS: Individuals with PD demonstrated a primary coupling direction from LPL to LPFC during walking tasks, whereas healthy controls did not demonstrate any main coupling direction. Compared with healthy controls, individuals with PD showed statistically significantly increased EC coupling strength from LPL to LPFC, from LPL to RPFC, and from LPL to RPL. Individuals with PD showed decreased gait speed and stride length and increased variability in speed and stride length. The EC coupling strength from LPL to RPFC negatively correlated with speed and positively correlated with speed variability in individuals with PD. CONCLUSION: In individuals with PD, the left prefrontal cortex may be regulated by the left parietal lobe during walking. This may be the result of functional compensation in the left parietal lobe.

fNIRS-based functional connectivity signifies recovery in patients with disorders of consciousness after DBS treatment.

OBJECTIVE: While deep brain stimulation (DBS) has proved effective for certain patients with disorders of consciousness (DOC), the working neural mechanism is not clear, the response varies for patients, and the assessment is inadequate. This paper aims to quantify the DBS-induced changes of consciousness in DOC patients at the neural functional level. METHODS: Ten DOC patients were included for DBS surgery. The DBS target was the right centromedian-parafascicular (CM-pf) nuclei for four patients and the bilateral CM-pf nuclei for six patients. Functional near-infrared spectroscopy (fNIRS) was taken to measure the neural activation of patients, in parallel with Coma Recovery Scale-Revised (CRS-R), before the DBS surgery and one month after. The fNIRS signals were recorded from the frontal, parietal, and occipital lobes. Functional connectivity analysis quantified the communication between brain regions, area communication strength, and global communication efficiency. Linear regression analysis was conducted between the changes of indices based on functional connectivity analysis and the changes of the CRS-R index. RESULTS: Patients with trauma (n = 4) exhibited a greater increase of CRS-R scores after DBS treatment compared with patients with hemorrhage (n = 4) and brainstem infarction (n = 2). Global communication efficiency changed consistently with the CRS-R index (slope = 57.384, p < 0.05, R2=0.483). No significant relationship was found between the changes of area communication strength of six brain lobes and the changes of the CRS-R index. CONCLUSIONS: The cause of DOC is essential for the outcome of DBS treatment, and brain communication efficiency is a promising functional marker for DOC recovery. SIGNIFICANCE: fNIRS-based functional connectivity analysis on brain network signifies changes of consciousness in DOC patients after DBS treatment.

Fronto-parietal cortex activation during walking in patients with Parkinson's disease adopting different postural strategies.

Background: Cortical activation patterns in patients with Parkinson's disease (PD) may be influenced by postural strategies, but the underlying neural mechanisms remain unclear. Our aim is to examine the role of the fronto-parietal lobes in patients with PD adopting different postural strategies and the effect of dual task (DT) on fronto-parietal activation. Methods: Two groups of patients with PD adopting either the posture first strategy (PD-PF) or the posture second strategy (PD-PS) were examined respectively when in the "OFF" state while single-walking task (SW) and DT. Frontal and parietal lobe activity was assessed by functional near infrared spectroscopy (fNIRS) and measuring gait parameters. Linear mixed models were used for analyses. Results: Patients with PD who adopted PS had greater cortical activation than those who adopted PF, and there was no difference between PF and PS in the behavioral parameters. For oxyhemoglobin levels, the task condition (SW vs. DT) had a main effect in fronto-parietal lobes. Postural strategy (PD-PF vs. PD-PS) a main effect in the left prefrontal cortex (LPFC), left parietal lobe (LPL), and right parietal lobe (RPL) regions. In the task of walking with and without the cognitive task, patients with PD adopting PS had higher activation in the LPL than those adopting PF. In DT, only PD patients who adopted PS had elevated oxyhemoglobin levels in the LPFC, right prefrontal cortex (RPFC), and LPL compared with the SW, whereas patients with PD who adopted PF showed no differences in any region. Conclusion: Different patterns of fronto-parietal activation exist between PD-PF and PD-PS. This may be because PD-PS require greater cortical functional compensation than those adopting PF.

Estimation of knee joint movement using single-channel sEMG signals with a feature-guided convolutional neural network.

Estimating human motion intention, such as intent joint torque and movement, plays a crucial role in assistive robotics for ensuring efficient and safe human-robot interaction. For coupled human-robot systems, surface electromyography (sEMG) signal has been proven as an effective means for estimating human's intended movements. Usually, joint movement estimation uses sEMG signals measured from multiple muscles and needs many sEMG sensors placed on the human body, which may cause discomfort or result in mechanical/signal interference from wearable robots/environment during long-term routine use. Although the muscle synergy principle implies that it is possible to estimate human motion using sEMG signals from even one signal muscle, few studies investigated the feasibility of continuous motion estimation based on single-channel sEMG. In this study, a feature-guided convolutional neural network (FG-CNN) has been proposed to estimate human knee joint movement using single-channel sEMG. In the proposed FG-CNN, several handcrafted features have been fused into a CNN model to guide CNN feature extraction, and both handcrafted and CNN-extracted features were applied to a regression model, i.e., random forest regression, to estimate knee joint movements. Experiments with 8 healthy subjects were carried out, and sEMG signals measured from 6 muscles, i.e., vastus lateralis, vastus medialis, biceps femoris, semitendinosus, lateral or medial gastrocnemius (LG or MG), were separately evaluated for knee joint estimation using the proposed method. The experimental results demonstrated that the proposed FG-CNN method with single-channel sEMG signals from LG or MG can effectively estimate human knee joint movements. The average correlation coefficient between the measured and the estimated knee joint movements is 0.858 ± 0.085 for LG and 0.856 ± 0.057 for MG. Meanwhile, comparative studies showed that the combined handcrafted-CNN features outperform either the handcrafted features or the CNN features; the performance of the proposed signal-channel sEMG-based FG-CNN method is comparable to those of the traditional multi-channel sEMG-based methods. The outcomes of this study enable the possibility of developing a single-channel sEMG-based human-robot interface for knee joint movement estimation, which can facilitate the routine use of assistive robots.

fNIRS-based brain state transition features to signify functional degeneration after Parkinson's disease.

Objective.Parkinson's disease (PD) is a common neurodegenerative brain disorder, and early diagnosis is of vital importance for treatment. Existing methods are mainly focused on behavior examination, while the functional neurodegeneration after PD has not been well explored. This paper aims to investigate the brain functional variation of PD patients in comparison with healthy controls.Approach.In this work, we propose brain hemodynamic states and state transition features to signify functional degeneration after PD. Firstly, a functional near-infrared spectroscopy (fNIRS)-based experimental paradigm was designed to capture brain activation during dual-task walking from PD patients and healthy controls. Then, three brain states, named expansion, contraction, and intermediate states, were defined with respect to the oxyhemoglobin and deoxyhemoglobin responses. After that, two features were designed from a constructed transition factor and concurrent variations of oxy- and deoxy-hemoglobin over time, to quantify the transitions of brain states. Further, a support vector machine classifier was trained with the proposed features to distinguish PD patients and healthy controls.Main results.Experimental results showed that our method with the proposed brain state transition features achieved classification accuracy of 0.8200 andFscore of 0.9091, and outperformed existing fNIRS-based methods. Compared with healthy controls, PD patients had significantly smaller transition acceleration and transition angle.Significance.The proposed brain state transition features well signify functional degeneration of PD patients and may serve as promising functional biomarkers for PD diagnosis.

A Functional Region Decomposition Method to Enhance fNIRS Classification of Mental States.

Functional near-infrared spectroscopy (fNIRS) classification of mental states is of important significance in many neuroscience and clinical applications. Existing classification algorithms use all signal-collected brain regions as a whole, and brain sub-region contributions have not been well investigated. This paper proposes a functional region decomposition (FRD) method to incorporate brain sub-region contributions and enhance fNIRS classification of mental states. Specifically, the method iteratively decomposes the brain region into multiple sub-regions to maximize their contributions with respect to the validation accuracy and coverage of brain sub-regions. Then for the fNIRS data in brain sub-regions, features are extracted and classified to output the predictions. The final predictions are determined by fusing predictions from multiple brain sub-regions with stacking. Experiments on a publicly available fNIRS dataset showed that the proposed functional region decomposition method led to 9.01% and 10.58% increase of classification accuracy for the methods related to slope-based features and mean concentration change features, respectively. Therefore, the proposed method can decompose the brain region into sub-regions with respect to their functional contributions and fundamentally enhance the performance of mental state classification.

Plantar pressure-based temporal analysis of gait disturbance in idiopathic normal pressure hydrocephalus: Indications from a pilot longitudinal study.

BACKGROUND AND OBJECTIVE: Idiopathic normal pressure hydrocephalus (iNPH) is a common yet potentially reversible neurodegenerative disease, and gait disturbance is a major symptom. Lots of methodological and clinical work has been conducted on gait disturbance analysis for differential diagnosis, presurgical test, and postsurgery assessment of iNPH. Nevertheless, the verification analysis was mostly lacking for surgery response, and the temporal characteristics of ground reaction force has been rarely investigated. METHODS: In this work, we propose that plantar pressure features fundamentally signifies iNPH gait disturbance and improvement after cerebrospinal fluid (CSF) drainage by lumbar puncture tap test as well as surgical shunt implantation. The plantar pressure signals of six iNPH patients and eight healthy controls were collected, and an online database of sixteen healthy controls were used. For patients, data were collected in five periods, which are the baseline before the tap test, 8, 24, and 72 hours after the tap test, and one month after the shunt implantation surgery, respectively. Fast dynamic time warping (DTW) with an improved DTW barycenter averaging (DBA) method was proposed for temporal analysis with the measured and online plantar pressure data. An plantar-pressure variation index (PPVI) was formulated to characterize the temporal dynamic stability of walking. RESULTS: The PPVI based on temporal analysis of plantar pressure well discriminated the impaired gait (baseline, 24 and 72 hours after tap test) with the improved gait (8 hours after tap test and follow up after surgery) of the patients. Further, the PPVI was close for the improved gait of the patients and the healthy gait measured in our study as well as in the online database. CONCLUSIONS: Plantar pressure-based temporal features are promisingly effective for clinical examination and treatment of iNPH.

Quantified assessment of deep brain stimulation on Parkinson's patients with task fNIRS measurements and functional connectivity analysis: a pilot study.

BACKGROUND: Deep brain stimulation (DBS) has proved effective for Parkinson's disease (PD), but the identification of stimulation parameters relies on doctors' subjective judgment on patient behavior. METHODS: Five PD patients performed 10-meter walking tasks under different brain stimulation frequencies. During walking tests, a wearable functional near-infrared spectroscopy (fNIRS) system was used to measure the concentration change of oxygenated hemoglobin (△HbO2) in prefrontal cortex, parietal lobe and occipital lobe. Brain functional connectivity and global efficiency were calculated to quantify the brain activities. RESULTS: We discovered that both the global and regional brain efficiency of all patients varied with stimulation parameters, and the DBS pattern enabling the highest brain efficiency was optimal for each patient, in accordance with the clinical assessments and DBS treatment decision made by the doctors. CONCLUSIONS: Task fNIRS assessments and brain functional connectivity analysis promise a quantified and objective solution for patient-specific optimization of DBS treatment. TRIAL REGISTRATION: Name: Accurate treatment under the multidisciplinary cooperative diagnosis and treatment model of Parkinson's disease. Registration number is ChiCTR1900022715. Date of registration is April 23, 2019.

Detection of Referable Horizontal Strabismus in Children's Primary Gaze Photographs Using Deep Learning.

Purpose: This study implements and demonstrates a deep learning (DL) approach for screening referable horizontal strabismus based on primary gaze photographs using clinical assessments as a reference. The purpose of this study was to develop and evaluate deep learning algorithms that screen referable horizontal strabismus in children's primary gaze photographs. Methods: DL algorithms were developed and trained using primary gaze photographs from two tertiary hospitals of children with primary horizontal strabismus who underwent surgery as well as orthotropic children who underwent routine refractive tests. A total of 7026 images (3829 non-strabismus from 3021 orthoptics [healthy] subjects and 3197 strabismus images from 2772 subjects) were used to develop the DL algorithms. The DL model was evaluated by 5-fold cross-validation and tested on an independent validation data set of 277 images. The diagnostic performance of the DL algorithm was assessed by calculating the accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results: Using 5-fold cross-validation during training, the average AUCs of the DL models were approximately 0.99. In the external validation data set, the DL algorithm achieved an AUC of 0.99 with a sensitivity of 94.0% and a specificity of 99.3%. The DL algorithm's performance (with an accuracy of 0.95) in diagnosing referable horizontal strabismus was better than that of the resident ophthalmologists (with accuracy ranging from 0.81 to 0.85). Conclusions: We developed and evaluated a DL model to automatically identify referable horizontal strabismus using primary gaze photographs. The diagnostic performance of the DL model is comparable to or better than that of ophthalmologists. Translational Relevance: DL methods that automate the detection of referable horizontal strabismus can facilitate clinical assessment and screening for children at risk of strabismus.

Detecting glaucoma based on spectral domain optical coherence tomography imaging of peripapillary retinal nerve fiber layer: a comparison study between hand-crafted features and deep learning model.

PURPOSE: To develop a deep learning (DL) model for automated detection of glaucoma and to compare diagnostic capability against hand-craft features (HCFs) based on spectral domain optical coherence tomography (SD-OCT) peripapillary retinal nerve fiber layer (pRNFL) images. METHODS: A DL model with pre-trained convolutional neural network (CNN) based was trained using a retrospective training set of 1501 pRNFL OCT images, which included 690 images from 153 glaucoma patients and 811 images from 394 normal subjects. The DL model was further tested in an independent test set of 50 images from 50 glaucoma patients and 52 images from 52 normal subjects. A customized software was used to extract and measure HCFs including pRNFL thickness in average and four different sectors. Area under the receiver operator characteristics (AROC) curves was calculated to compare the diagnostic capability between DL model and hand-crafted pRNFL parameters. RESULTS: In this study, the DL model achieved an AROC of 0.99 [CI: 0.97 to 1.00] which was significantly larger than the AROC values of all other HCFs (AROCs 0.661 with 95% CI 0.549 to 0.772 for temporal sector, AROCs 0.696 with 95% CI 0.549 to 0.799 for nasal sector, AROCs 0.913 with 95% CI 0.855 to 0.970 for superior sector, AROCs 0.938 with 95% CI 0.894 to 0.982 for inferior sector, and AROCs 0.895 with 95% CI 0.832 to 0.957 for average). CONCLUSION: Our study demonstrated that DL models based on pre-trained CNN are capable of identifying glaucoma with high sensitivity and specificity based on SD-OCT pRNFL images.

A Hierarchical Image Matting Model for Blood Vessel Segmentation in Fundus Images.

In this paper, a hierarchical image matting model is proposed to extract blood vessels from fundus images. More specifically, a hierarchical strategy is integrated into the image matting model for blood vessel segmentation. Normally the matting models require a user specified trimap, which separates the input image into three regions: the foreground, background and unknown regions. However, creating a user specified trimap is laborious for vessel segmentation tasks. In this paper, we propose a method that first generates trimap automatically by utilizing region features of blood vessels, then applies a hierarchical image matting model to extract the vessel pixels from the unknown regions. The proposed method has low calculation time and outperforms many other state-of-art supervised and unsupervised methods. It achieves a vessel segmentation accuracy of 96.0%, 95.7% and 95.1% in an average time of 10.72s, 15.74s and 50.71s on images from three publicly available fundus image datasets DRIVE, STARE, and CHASE DB1, respectively.

Optic Disk Detection in Fundus Image Based on Structured Learning.

Automated optic disk (OD) detection plays an important role in developing a computer aided system for eye diseases. In this paper, we propose an algorithm for the OD detection based on structured learning. A classifier model is trained based on structured learning. Then, we use the model to achieve the edge map of OD. Thresholding is performed on the edge map, thus a binary image of the OD is obtained. Finally, circle Hough transform is carried out to approximate the boundary of OD by a circle. The proposed algorithm has been evaluated on three public datasets and obtained promising results. The results (an area overlap and Dices coefficients of 0.8605 and 0.9181, respectively, an accuracy of 0.9777, and a true positive and false positive fraction of 0.9183 and 0.0102) show that the proposed method is very competitive with the state-of-the-art methods and is a reliable tool for the segmentation of OD.