A new method to predict anomaly in brain network based on graph deep learning

Applications of generative adversarial networks in neuroimaging and clinical neuroscience

Generative adversarial networks (GANs) are one powerful type of deep learning models that have been successfully utilized in numerous fields. They belong to the broader family of generative methods, which learn to generate realistic data with a probabilistic model by learning distributions from real samples. In the clinical context, GANs have shown enhanced capabilities in capturing spatially complex, nonlinear, and potentially subtle disease effects compared to traditional generative methods. This review critically appraises the existing literature on the applications of GANs in imaging studies of various neurological conditions, including Alzheimer's disease, brain tumors, brain aging, and multiple sclerosis. We provide an intuitive explanation of various GAN methods for each application and further discuss the main challenges, open questions, and promising future directions of leveraging GANs in neuroimaging. We aim to bridge the gap between advanced deep learning methods and neurology research by highlighting how GANs can be leveraged to support clinical decision making and contribute to a better understanding of the structural and functional patterns of brain diseases.

Evolution of brain functional plasticity associated with increasing symptom severity in degenerative cervical myelopathy

BACKGROUND

Advanced imaging modalities have helped elucidate the cerebral alterations associated with neurological impairment caused by degenerative cervical myelopathy (DCM), but it remains unknown how brain functional network changes at different stages of myelopathy severity in DCM patients, and if patterns in network connectivity can be used to predict transition to more myelopathic stages of DCM.

METHODS

This pilot cross-sectional study, which involves the collection of resting-state functional MRI (rs-fMRI) images and the modified Japanese Orthopedic Association (mJOA) score, enrolled 116 participants (99 patients and 17 healthy controls) from 2016 to 2021. The patient cohort included 21patients with asymptomatic spinal cord compression, 48 mild DCM patients, and 20 moderate or severe DCM patients. Functional connectivity networks were quantified for all participants, and the transition matrices were quantified to determine the differences in network connectivity through increasingly myelopathic stages of DCM. Additionally, a link prediction model was used to determine whether more severe stages of DCM can be predicted from less symptomatic stages using the transition matrices.

FINDINGS

Results indicated interruptions in most connections within the sensorimotor network in conjunction with spinal cord compression, while compensatory connectivity was observed within and between primary and secondary sensorimotor regions, subcortical regions, visuospatial regions including the cuneus, as well as the brainstem and cerebellum. A link prediction model achieved an excellent predictive performance in estimating connectivity of more severe myelopathic stages of DCM, with the highest area under the receiver operator curve (AUC) of 0.927 for predicting mild DCM from patients with asymptomatic spinal cord compression.

INTERPRETATION

A series of predictable changes in functional connectivity occur throughout the stages of DCM pathogenesis. The brainstem and cerebellum appear highly influential in optimizing sensorimotor function during worsening myelopathy. The link predication model can inclusively estimate brain alterations associated with myelopathy severity.

FUNDING

NIH/NINDS grants (1R01NS078494-01A1, and 2R01NS078494).

Multimodal Orbital Angular Momentum Data Model Based on Mechanically Reconfigurable Arrays and Neural Networks

Multimodal orbital angular momentum is a research hotspot in the field of electromagnetic wave communication. How to accurately detect and identify multimodal orbital angular momentum data is a current academic problem. Based on the theory of mechanically reconfigurable arrays and neural networks, the purity, detection method, and transmission and reception of orbital angular momentum vortex waves are modeled in this paper. Through the network identification of the dynamic model of the three-degree-of-freedom reconfigurable manipulator, the paper takes the identification result and the control input of the single neuron PID as the input of the system control torque of the manipulator and realizes the reconfigurable manipulator. In the simulation process, the local approximation effect of the nonlinear control system used is very ideal. The single neuron PID controller overcomes the shortcomings of time-consuming and unsatisfactory control accuracy caused by the constant parameter of the traditional PID controller and realizes the circular loop. On the other hand, at the point of interest of the human eye, its resolution value is the largest, and its value gradually decreases as the distance from the pit increases. The experimental results show that the three-transmitting and three-receiving orbital angular momentum vortex wave transceiver system based on the mechanically reconfigurable array and neural network theory is relatively complete, and the transmission coefficient between the same modes reaches 0.827, which is much higher than that between different modes. On this basis, the modal purity, detection method, and reception of orbital angular momentum are studied accordingly. At the same time, the damage to the microscopic particles can be greatly reduced. At the same time, the information delay is reduced to 8.25%, which effectively improves the isolation characteristics of different modal orbital angular momentum channels and promotes the communication transmission of multimodal signals.

Developing Pulmonary Rehabilitation for COVID-19: Are We Linked with the Present Literature? A Lexical and Geographical Evaluation Study Based on the Graph Theory

The Coronavirus Disease 2019 (COVID-19) pandemic is a severe ongoing global emergency. Despite high rates of asymptomatic patients, in many cases, the infection causes a rapid decline in pulmonary function due to an acute respiratory distress-like syndrome, leading to multi-organ failure and death. To date, recommendations about rehabilitation on COVID-19 are based on clinical data derived from other similar lung diseases. Rehabilitation literature lacks a standard taxonomy, limiting a proper evaluation of the most effective treatments for patients after COVID-19 infection. In this study, we assessed the clinical and rehabilitative associations and the geographical area involved in interstitial lung diseases (ILD) and in COVID-19, by a mathematical analysis based on graph theory. We performed a quantitative analysis of the literature in terms of lexical analysis and on how words are connected to each other. Despite a large difference in timeframe (throughout the last 23 years for ILD and in the last 1.5 years for COVID-19), the numbers of papers included in this study were similar. Our results show a clear discrepancy between rehabilitation proposed for COVID-19 and ILD. In ILD, the term “rehabilitation” and other related words such as “exercise” and “program” resulted in lower values of centrality and higher values of eccentricity, meaning relatively less importance of the training during the process of care in rehabilitation of patients with ILD. Conversely, “rehabilitation” was one of the most cited terms in COVID-19 literature, strongly associated with terms such as “exercise”, “physical”, and “program”, entailing a multidimensional approach of the rehabilitation for these patients. This could also be due to the widespread studies conducted on rehabilitation on COVID-19, with Chinese and Italian researchers more involved. The assessment of the terms used for the description of the rehabilitation may help to program shared rehabilitation knowledge and avoid literature misunderstandings.

A primer on machine learning techniques for genomic applications

High throughput sequencing technologies have enabled the study of complex biological aspects at single nucleotide resolution, opening the big data era. The analysis of large volumes of heterogeneous "omic" data, however, requires novel and efficient computational algorithms based on the paradigm of Artificial Intelligence. In the present review, we introduce and describe the most common machine learning methodologies, and lately deep learning, applied to a variety of genomics tasks, trying to emphasize capabilities, strengths and limitations through a simple and intuitive language. We highlight the power of the machine learning approach in handling big data by means of a real life example, and underline how described methods could be relevant in all cases in which large amounts of multimodal genomic data are available.