Neuroscience informatics最新文献

The accurate segmentation of brain stroke lesions in medical images are critical for early diagnosis, treatment planning, and monitoring of stroke patients. In recent years, deep learning-based approaches have shown great potential for brain stroke segmentation in both MRI and CT scans. However, it is not clear which modality is superior for this task. This paper provides a comprehensive review of recent advancements in the use of deep learning for stroke lesion segmentation in both MRI and CT scans. We compare the performance of various deep learning-based approaches and highlight the advantages and limitations of each modality. The deep learning models for ischemic segmentation task are evaluated using segmentation metrics including Dice, Jaccard, Sensitivity, and Specificity.

Intraoperative neurophysiological monitoring (IONM) provides data on the state of neurological functionality. However, the current state of technology impedes the reliable and timely extraction and communication of relevant information. Advanced signal processing and machine learning (ML) technologies can develop a robust surveillance system that can reliably monitor the current state of a patient's nervous system and promptly alert the surgeons of any imminent risk. Various ML and signal processing tools can be utilized to develop a real-time, objective, multi-modal IONM based-alert system for spine surgery. Next generation systems should be able to obtain inputs from anesthesiologists on vital sign disturbances and pharmacological changes, as well as being capable of adapting patient baseline and model parameters for patient variability in age, gender, and health. It is anticipated that the application of automated decision guiding of checklist strategies in response to warning criteria can reduce human work-burden, improve accuracy, and minimize errors.

Seizure detection from EEG signals is crucial for diagnosing and treating neurological disorders. However, accurately detecting seizures is challenging due to the complexity and variability of EEG signals. This paper proposes a deep learning model, called Hybrid Cross Layer Attention Based Convolutional Bidirectional Gated Recurrent Unit (HCLA_CBiGRU), which combines convolutional neural networks and recurrent neural networks to capture spatial and temporal features in EEG signals. A combinational EEG dataset was created by merging publicly available datasets and applying a preprocessing pipeline to remove noise and artifacts. The dataset was then segmented and split into training and testing sets. The HCLA_CBiGRU model was trained on the training set and evaluated on the testing set, achieving an impressive accuracy of 98.5%, surpassing existing state-of-the-art methods. Sensitivity and specificity, critical metrics in clinical practice, were also assessed, with the model demonstrating a sensitivity of 98.5% and a specificity of 98.9%, highlighting its effectiveness in seizure detection. Visualization techniques were used to analyze the learned features, showing the model's ability to capture distinguishing seizure-related characteristics. In conclusion, the proposed CBiGRU model outperforms existing methods in terms of accuracy, sensitivity, and specificity for seizure detection from EEG signals. Its integration with EEG signal analysis has significant implications for improving the diagnosis and treatment of neurological disorders, potentially leading to better patient outcomes.

Background

The identification of seizure and its complex waveforms in electroencephalography (EEG) through manual examination is time consuming, tedious, and susceptible to human mistakes. These issues have prompted the design of an automated seizure detection system that can assist the neurophysiologists by providing a fast and accurate analysis.

Methods

Existing automated seizure detection systems are either machine learning based or deep learning based. Machine learning based algorithms employ handcrafted features with sophisticated feature selection approaches. As a result of which their performance varies with the choice of the feature extraction and selection techniques employed. On the other hand, deep learning-based methods automatically deduce the best subset of features required for the categorization task but they are computationally expensive and lacks generalization on clinical EEG datasets. To address the above stated limitations and motivated by the advantage of continuous wavelet transform's (CWT) in elucidating the non-stationary nature of the EEG signals in a better way, we propose an approach based on EEG image representations (constructed via applying WT at different scale and time intervals) and transfer learning for seizure detection. Firstly, the pre-trained model is fine-tuned on the EEG image representations and thereafter features are extracted from the trained model by performing activations on different layers of the network. Subsequently, the features are passed through a Support Vector Machine (SVM) for categorization using a 10-fold data partitioning scheme.

Results and comparison with existing methods

The proposed mechanism results in a ceiling level of classification performance (accuracy=99.50/98.67, sensitivity=100/100 & specificity=99/96) for both the standard and the clinical dataset that are better than the existing state-of-the art works.

Conclusion

The rapid advancement in the field of deep learning has created a paradigm shift in automated diagnosis of epilepsy. The proposed tool has effectually marked the relevant EEG segments for the clinician to review thereby reducing the time burden in scanning the long duration EEG records.

Evaluation of access routes and shunting points plays a crucial role in the treatment of cavernous sinus dural arteriovenous fistulas (CS-dAVF). Generally, these evaluations are performed using three-dimensional rotation angiography. However, assessing access routes becomes challenging in cases lacking anterior or posterior drainage routes. Zero TE magnetic resonance imaging (MRI) is an innovative technique enabling the visualization of cortical bone. By merging fusion images of zero TE and contrast-enhanced T1 weighted imaging (CE-T1WI), enhanced arteries can be visualized, resembling cranial bone-like three-dimensional rotation angiography. To determine the usefulness of fusion images in evaluating access routes and shunting points for dural arteriovenous fistulas, a comparison was made between these fusion images and three-dimensional rotation angiography in the same case. This report describes the application of fusion images in evaluating access routes and shunting points.

As a change in the electroencephalogram (EEG) during motor tasks, the phenomenon in the sensorimotor area (SM1) is called event-related desynchronization (ERD). Motor commands are discharged from the primary motor area (M1) to the muscle through the corticospinal pathway and feedback to the primary somatosensory area (S1). This sensory input from the peripheral nerve stimulation to the central nervous system is attenuated during motor tasks by motor commands. This phenomenon is known as movement gating and is observed not only in S1, but also in non-primary motor areas. However, the brain circuits that trigger these motor-related changes and how the brain circuit modulates them as a controller remain unsolved. In this study, we evaluated the effects of spontaneous EEG changes and movement gating of somatosensory evoked potentials (SEPs) during motor execution by modulating cortical excitability with low-frequency repetitive transcranial magnetic stimulation (rTMS) over the PMc. Low frequency rTMS is known as an application where cortical excitability is suppressed after the stimulation. After rTMS, not only the previously known ERD, but also the newly gating of SEPs N30 and corticocortical spontaneous EEG changes were evaluated by Granger causality, which indicates that the time-varying causal relationship from the frontal to parietal area was significantly attenuated among eight healthy participants. These results suggest that spontaneous changes in EEG on SM1 and cortico-cortical connectivity during motor tasks are related to sensory feedback suppression of the frontal cortex.

Background and objective

3D rotational angiography (3DRA) provides high quality images of the cerebral arteriovenous malformation (AVM) nidus that can be reconstructed in 3D. However, these reconstructions are limited to only 3D visualization without possible interactive exploration of geometric characteristics of cerebral structures. Refined understanding of the AVM angioarchitecture prior to treatment is mandatory and vascular segmentation is an important preliminary step that allow physicians analyze the complex vascular networks and can help guide microcatheters navigation and embolization of AVM.

Methods

A deep learning method was developed for the segmentation of 3DRA images of AVM patients. The method uses a fully convolutional neural network with a U-Net-like architecture and a DenseNet backbone. A compound loss function, combining Cross Entropy and Focal Tversky, is employed for robust segmentation. Binary masks automatically generated from region-growing segmentation have been used to train and validate our model.

Results

The developed network was able to achieve the segmentation of the vessels and the malformation and significantly outperformed the region-growing algorithm. Our experiments were performed on 9 AVM patients. The trained network achieved a Dice Similarity Coefficient (DSC) of 80.43%, surpassing other U-Net like architectures and the region-growing algorithm on the manually approved test set by physicians.

Conclusions

This work demonstrates the potential of a learning-based segmentation method for characterizing very complex and tiny vascular structures even when the training phase is performed with the results of an automatic or a semi-automatic method. The proposed method can contribute to the planning and guidance of endovascular procedures.

Background

Large vessel occlusion (LVO) stroke research is limited regarding high-risk patient groups for complications including cerebral edema. Large, well-phenotyped cohorts hold potential insights, but identifying cohorts and manually extracting outcomes is impractical. Natural language processing (NLP) software has previously extracted stroke characteristics from radiology reports, but there has not been an integrated extraction of both LVO classification and acute stroke outcomes.

Methods

We constructed a rules-based NLP pipeline that extracted presence/location of arterial occlusion and core/penumbral volumes from multimodal CT reports, along with presence of edema and midline shift on follow-up CTs. The algorithm flagged inconsistent reports for manual adjudication. We validated performance over two cohorts and analyzed the associations between NLP-extracted variables and clinical edema outcomes.

Results

The algorithm identified occlusions in the development ($n=577$) and test cohorts ($n=442$) with 94% and 85% recall, increasing to 97% and 93% after review of flagged reports. It could distinguish proximal ICA/M1 from distal occlusions with 96% recall and correctly extracted 98% of core/penumbral volumes. NLP recall was 93% and 86% for identifying edema and midline shift from follow-up reports of 213 patients with ICA/MCA occlusions. NLP-extracted radiographic edema captured 89% of those who developed clinical cerebral edema, which was more likely in those with NLP-identified proximal vs distal occlusions and associated with significantly higher core/penumbral volumes.

Conclusion

A rules-based NLP pipeline can accurately identify and phenotype an LVO cohort, yielding clinical associations with stroke research implications.

This paper investigates the development of an intelligent system method to address completely locked-in-syndrome (CLIS) that is caused by some illnesses such as Amyotrophic Lateral Sclerosis (ALS) as the most predominant type of Motor Neuron Disease (MND). In the last stages of ALS and despite the limitations in body movements, patients however will have a fully functional brain and cognitive capabilities and able to feel pain but fail to communicate. This paper aims to address the CLIS problem by utilizing EEG signals that human brain generates when thinking about a specific feeling or imagination as a way to communicate. The aim is to develop a low-cost and affordable system for patients to use to communicate with carers and family members. In this paper, the novel implementation of the ASPS (Automated Sensor and Signal Processing Selection) approach for feature extraction of EEG is presented to select the most suitable Sensory Characteristic Features (SCFs) to detect human thoughts and imaginations. Artificial Neural Networks (ANN) are used to verify the results. The findings show that EEG signals are able to capture imagination information that can be used as a means of communication; and the ASPS approach allows the selection of the most important features for reliable communication. This paper explains the implementation and validation of ASPS approach in brain signal classification for bespoke arrangement. Hence, future work will present the results of relatively high number of volunteers, sensors and signal processing methods.