The role of cortical structural variance in deep learning-based prediction of fetal brain age.

Background: Deep-learning-based brain age estimation using magnetic resonance imaging data has been proposed to identify abnormalities in brain development and the risk of adverse developmental outcomes in the fetal brain. Although saliency and attention activation maps have been used to understand the contribution of different brain regions in determining brain age, there has been no attempt to explain the influence of shape-related cortical structural features on the variance of predicted fetal brain age. Methods: We examined the association between the predicted brain age difference (PAD: predicted brain age-chronological age) from our convolution neural networks-based model and global and regional cortical structural measures, such as cortical volume, surface area, curvature, gyrification index, and folding depth, using regression analysis. Results: Our results showed that global brain volume and surface area were positively correlated with PAD. Additionally, higher cortical surface curvature and folding depth led to a significant increase in PAD in specific regions, including the perisylvian areas, where dramatic agerelated changes in folding structures were observed in the late second trimester. Furthermore, PAD decreased with disorganized sulcal area patterns, suggesting that the interrelated arrangement and areal patterning of the sulcal folds also significantly affected the prediction of fetal brain age. Conclusion: These results allow us to better understand the variance in deep learning-based fetal brain age and provide insight into the mechanism of the fetal brain age prediction model.

Automatic cortical surface parcellation in the fetal brain using attention-gated spherical U-net.

Cortical surface parcellation for fetal brains is essential for the understanding of neurodevelopmental trajectories during gestations with regional analyses of brain structures and functions. This study proposes the attention-gated spherical U-net, a novel deep-learning model designed for automatic cortical surface parcellation of the fetal brain. We trained and validated the model using MRIs from 55 typically developing fetuses [gestational weeks: 32.9 ± 3.3 (mean ± SD), 27.4-38.7]. The proposed model was compared with the surface registration-based method, SPHARM-net, and the original spherical U-net. Our model demonstrated significantly higher accuracy in parcellation performance compared to previous methods, achieving an overall Dice coefficient of 0.899 ± 0.020. It also showed the lowest error in terms of the median boundary distance, 2.47 ± 1.322 (mm), and mean absolute percent error in surface area measurement, 10.40 ± 2.64 (%). In this study, we showed the efficacy of the attention gates in capturing the subtle but important information in fetal cortical surface parcellation. Our precise automatic parcellation model could increase sensitivity in detecting regional cortical anomalies and lead to the potential for early detection of neurodevelopmental disorders in fetuses.

Exploring Autonomic Alterations during Seizures in Temporal Lobe Epilepsy: Insights from a Heart-Rate Variability Analysis.

Epilepsy's impact on cardiovascular function and autonomic regulation, including heart-rate variability, is complex and may contribute to sudden unexpected death in epilepsy (SUDEP). Lateralization of autonomic control in the brain remains the subject of debate; nevertheless, ultra-short-term heart-rate variability (HRV) analysis is a useful tool for understanding the pathophysiology of autonomic dysfunction in epilepsy patients. A retrospective study reviewed medical records of patients with temporal lobe epilepsy who underwent presurgical evaluations. Data from 75 patients were analyzed and HRV indices were extracted from electrocardiogram recordings of preictal, ictal, and postictal intervals. Various HRV indices were calculated, including time domain, frequency domain, and nonlinear indices, to assess autonomic function during different seizure intervals. The study found significant differences in HRV indices based on hemispheric laterality, language dominancy, hippocampal atrophy, amygdala enlargement, sustained theta activity, and seizure frequency. HRV indices such as the root mean square of successive differences between heartbeats, pNN50, normalized low-frequency, normalized high-frequency, and the low-frequency/high-frequency ratio exhibited significant differences during the ictal period. Language dominancy, hippocampal atrophy, amygdala enlargement, and sustained theta activity were also found to affect HRV. Seizure frequency was correlated with HRV indices, suggesting a potential relationship with the risk of SUDEP.

Semi-supervised automatic seizure detection using personalized anomaly detecting variational autoencoder with behind-the-ear EEG.

BACKGROUND AND OBJECTIVE: Epilepsy is one of the most common neurologic diseases worldwide, and 30% of the patients live with uncontrolled seizures. For the safety of patients with epilepsy, an automatic seizure detection algorithm for continuous seizure monitoring in daily life is important to reduce risks related to seizures, including sudden unexpected death. Previous researchers applied machine learning to detect seizures with EEG, but the epileptic EEG waveform contains subtle changes that are difficult to identify. Furthermore, the imbalance problem due to the small proportion of ictal events caused poor prediction performance in supervised learning approaches. This study aimed to present a personalized deep learning-based anomaly detection algorithm for seizure monitoring with behind-the-ear electroencephalogram (EEG) signals. METHODS: We collected behind-the-ear EEG signals from 16 patients with epilepsy in the hospital and used them to develop and evaluate seizure detection algorithms. We modified the variational autoencoder network to learn the latent representation of normal EEG signals and performed seizure detection by measuring the anomalies in EEG signals using the trained network. To personalize the algorithm, we also proposed a method to calibrate the anomaly score for each patient by comparing the representations in the latent space. RESULTS: Our proposed algorithm showed a sensitivity of 90.4% with a false alarm rate of 0.83 per hour without personal calibration. On the other hand, the one-class support vector machine only showed a sensitivity of 84.6% with a false alarm rate of 2.17 per hour. Furthermore, our proposed model with personal calibration achieved 94.2% sensitivity with a false alarm rate of 0.29 while detecting 49 of 52 ictal events. CONCLUSIONS: We proposed a novel seizure detection algorithm with behind-the-ear EEG signals via semi-supervised learning of an anomaly detecting variational autoencoder and personalization method of anomaly scoring by comparing latent representations. Our approach achieved improved seizure detection with high sensitivity and a lower false alarm rate.

Analysis of Gait Characteristics Using Hip-Knee Cyclograms in Patients with Hemiplegic Stroke.

Gait disturbance is a common sequela of stroke. Conventional gait analysis has limitations in simultaneously assessing multiple joints. Therefore, we investigated the gait characteristics in stroke patients using hip-knee cyclograms, which have the advantage of simultaneously visualizing the gait kinematics of multiple joints. Stroke patients (n = 47) were categorized into two groups according to stroke severity, and healthy controls (n = 32) were recruited. An inertial measurement unit sensor-based gait analysis system, which requires placing seven sensors on the dorsum of both feet, the shafts of both tibias, the middle of both femurs, and the lower abdomen, was used for the gait analysis. Then, the hip-knee cyclogram parameters (range of motion, perimeter, and area) were obtained from the collected data. The coefficient of variance of the cyclogram parameters was obtained to evaluate gait variability. The cyclogram parameters differed between the stroke patients and healthy controls, and differences according to stroke severity were also observed. The gait variability parameters mainly differed in patients with more severe stroke, and specific visualized gait patterns of stroke patients were obtained through cyclograms. In conclusion, the hip-knee cyclograms, which show inter-joint coordination and visualized gait cycle in stroke patients, are clinically significant.

Comparing Ictal Cardiac Autonomic Changes in Patients with Frontal Lobe Epilepsy and Temporal Lobe Epilepsy by Ultra-Short-Term Heart Rate Variability Analysis.

Background and Objectives: Abnormal epileptic discharges in the brain can affect the central brain regions that regulate autonomic activity and produce cardiac symptoms, either at onset or during propagation of a seizure. These autonomic alterations are related to cardiorespiratory disturbances, such as sudden unexpected death in epilepsy. This study aims to investigate the differences in cardiac autonomic function between patients with temporal lobe epilepsy (TLE) and frontal lobe epilepsy (FLE) using ultra-short-term heart rate variability (HRV) analysis around seizures. Materials and Methods: We analyzed electrocardiogram (ECG) data recorded during 309 seizures in 58 patients with epilepsy. Twelve patients with FLE and 46 patients with TLE were included in this study. We extracted the HRV parameters from the ECG signal before, during and after the ictal interval with ultra-short-term HRV analysis. We statistically compared the HRV parameters using an independent t-test in each interval to compare the differences between groups, and repeated measures analysis of variance was used to test the group differences in longitudinal changes in the HRV parameters. We performed the Tukey-Kramer multiple comparisons procedure as the post hoc test. Results: Among the HRV parameters, the mean interval between heartbeats (RRi), normalized low-frequency band power (LF) and LF/HF ratio were statistically different between the interval and epilepsy types in the t-test. Repeated measures ANOVA showed that the mean RRi and RMSSD were significantly different by epilepsy type, and the normalized LF and LF/HF ratio significantly interacted with the epilepsy type and interval. Conclusions: During the pre-ictal interval, TLE patients showed an elevation in sympathetic activity, while the FLE patients showed an apparent increase and decrease in sympathetic activity when entering and ending the ictal period, respectively. The TLE patients showed a maintained elevation of sympathetic and vagal activity in the pos-ictal interval. These differences in autonomic cardiac characteristics between FLE and TLE might be relevant to the ictal symptoms which eventually result in SUDEP.

Unsupervised automatic seizure detection for focal-onset seizures recorded with behind-the-ear EEG using an anomaly-detecting generative adversarial network.

BACKGROUND AND OBJECTIVE: Epilepsy is a neurological disorder of the brain, which involves recurrent seizures. An encephalogram (EEG) is a gold standard method in the detection and analysis of epileptic seizures. However, the standard EEG recording system is too obstructive to be used in daily life. Behind-the-ear EEG is an alternative approach to record EEG conveniently. Previous researchers applied machine learning to automatically detect seizures with EEG, but the epileptic EEG waveform contains subtle changes that are difficult to be identified. Furthermore, the extremely small proportion of ictal events in the long-term monitoring may cause the imbalance problem and, consequently, poor prediction performance in supervised learning approaches. In this study, we present an automatic seizure detection algorithm with a generative adversarial network (GAN) trained by unsupervised learning and evaluated it with behind-the-ear EEG. METHODS: We recorded behind-the-ear EEGs from 12 patients who have various types of epilepsy. Data were reviewed separately by two epileptologists, who determined the onsets and ends of seizures. First, we conducted unsupervised learning with the normal records for the GAN to learn the representation of normal states. Second, we performed automatic seizure detection with the trained GAN as an anomaly detector. Last, we combined the Gram matrix with other anomaly losses to improve detection performance. RESULTS: The proposed approach achieved detection performance with an area under the receiver operating curve of 0.939 and sensitivity of 96.3% with a false alarm rate of 0.14 per hour in the test dataset. In addition, we confirmed distinguishability with the distribution of the anomaly scores in terms of EEG frequency bands. CONCLUSIONS: It is expected that the proposed anomaly detection via GAN with the behind-the-ear EEG can be effectively used for long-term seizure monitoring in daily life.

Development of an automatic muscle atrophy measuring algorithm to calculate the ratio of supraspinatus in supraspinous fossa using deep learning.

BACKGROUND AND OBJECTIVE: Rotator cuff muscle tear is one of the most frequent reason of operations in orthopedic surgery. There are several clinical indicators such as Goutallier grade and occupation ratio in the diagnosis and surgery of these diseases, but subjective intervention of the diagnosis is an obstacle in accurately detecting the correct region. METHODS: Therefore, in this paper, we propose a fully convolutional deep learning algorithm to quantitatively detect the fossa and muscle region by measuring the occupation ratio of supraspinatus in the supraspinous fossa. In the development and performance evaluation of the algorithm, 240 patients MRI dataset with various disease severities were included. RESULTS: As a result, the pixel-wise accuracy of the developed algorithm is 0.9984 ± 0.073 in the fossa region and 0.9988 ± 0.065 in the muscle region. The dice coefficient is 0.9718 ± 0.012 in the fossa region and 0.9463 ± 0.047 in the muscle region. CONCLUSIONS: We expect that the proposed convolutional neural network can improve the efficiency and objectiveness of diagnosis by quantifying the index used in the orthopedic rotator cuff tear.

A data-driven artificial intelligence model for remote triage in the prehospital environment.

In a mass casualty incident, the factors that determine the survival rate of injured patients are diverse, but one of the key factors is the time for triage. Additionally, the main factor that determines the time of triage is the number of medical personnel. However, when relying on a small number of medical personnel, the ability to increase survivability is limited. Therefore, developing a classification model for survival prediction that can quickly and precisely triage via wearable devices without medical personnel is important. In this study, we designed a consciousness index to substitute the factor by manpower and improved the classification accuracy by applying a machine learning algorithm. First, logistic regression analysis using vital signs and a consciousness index capable of remote monitoring through wearable devices confirmed the high efficiency of the consciousness index. We then developed a classification model with high accuracy which corresponds to existing injury severity scoring systems through the machine learning algorithms. We extracted 460,865 cases which met our criteria for developing the survival prediction from the national sample project in the national trauma databank which contains 408,316 cases of blunt injury and 52,549 cases of penetrating injury. Among the dataset, 17,918 (3.9%) cases died while the other survived. The AUCs with 95% confidence intervals (CIs) for the different models with the proposed simplified consciousness score as follows: RTS (as baseline), 0.78 (95% CI = 0.775 to 0.785); logistic regression, 0.87 (95% CI = 0.862 to 0.870); random forest, 0.87 (95% CI = 0.862 to 0.872); deep neural network, 0.89 (95% CI = 0.882 to 0.890). As a result, we confirmed the possibility of remote triage using a wearable device. It is expected that the time required for triage can be effectively reduced by using the developed classification model of survival prediction.