Cerebral autoregulation (CA) encompasses a series of physiological mechanisms that are necessary to regulate blood flow in the brain. The procedure for CA assessment via the autoregulatory index (ARI) requires the estimate of the critical closing pressure (CrCP). The study aims at investigating the impact of the strategy exploited for CrCP estimation on ARI by comparing three approaches: i) fixed CrCP at 12 mmHg (CrCP12); ii) first harmonic (H1) method applied to waveforms of arterial pressure (AP) and cerebral blood velocity (CBv); iii) 2-point technique using mean and diastolic AP and CBv values (2Pm). Analysis was carried out over AP and CBv signals recorded in 25 healthy subjects (age: 44 ± 10 yrs, 12 females, 13 males) at rest in supine position and during active standing. Computation of CrCP was complemented by the assessment of the resistance-area product (RAP). We found that the H1 and 2Pm methods led to different values of CrCP and RAP. However, the strategy selected for the CrCP computation did not affect the ARI estimation, and this result held regardless of the experimental condition. We conclude that the CrCP12 strategy can be safely utilized instead of more complex methods for the CA characterization based on ARI.

Footstep recognition is an emerging biometric that identifies or verifies users based on footstep pressure patterns obtained while walking. However, the impact of covariates on footstep recordings is not well understood, unlike more established biometric traits such as fingerprint and facial recognition. Therefore, this study used unsupervised hierarchical clustering (HCA) to examine the internal and external covariate influence on spatial and temporal footstep features of twenty individuals. Using 22 cluster validity indices, a robust HCA technique identified two distinct clusters in spatial representations (i.e., peak pressure images) and temporal representations (i.e., ground reaction force (GRF) and center of pressure (COP) time series) of the gait patterns. The clusters determined in both feature domains were distinguishable by body weight, age, race, and shoe type. Interestingly, trends related to sex and walking speed existed only in the temporal domain. These findings suggest dual implications for footstep biometric systems, which may leverage covariate information as soft biometrics to improve user recognition, or require mitigation to limit model bias and improve generalization to new users and conditions.

Standard diagnostic methods for evaluating the severity of brain injuries resulting from cardiac arrest, such as the Glasgow Coma Scale, exhibit subjective biases that lead to potentially fatal misclassifications, where life-support systems are prematurely withdrawn from patients who might otherwise recover. This study utilizes an open dataset from the International Cardiac Arrest Research Consortium to develop and evaluate a 3D convolutional neural network (CNN) model for classifying outcomes in comatose patients after cardiac arrest. The electroencephalographic (EEG) signals from the dataset are preprocessed by resampling, filtering, and standardizing signal length (10 seconds) and channel count. The model's architecture comprises 3D convolutional neural networks with subsequent layers for vectorization, compression, and further automatic feature extraction. Evaluation metrics focus on the area under the receiver operating characteristic curve, confusion matrix, accuracy, and F1 score. Results show that the 3D-CNN model outperforms existing 2D-CNN models in classifying outcomes for comatose patients, exhibiting a higher area under the receiver operating characteristic curve.

Dynamic assessment of cerebral blood flow (CBF) is crucial for the prevention and treatment of cerebrovascular diseases. Near-field coupling (NFC), due to its non-invasive and continuous advantages, holds significant promise in the dynamic assessment of CBF compared to traditional detection methods. However, challenges related to individual differences persist. This study integrates electrocardiographic signal with NFC to explore the feasibility of reflecting diverse individual CBF change patterns. Ten healthy volunteers were selected to undergo continuous monitoring and analysis of ECG and CBF signals in both supine rest and a 30° tilted position. Using heart rate variability (HRV) as a reference, the CBF signal feature patterns before and after tilt were analyzed. Finally, two distinct groups with different regulatory capabilities were identified based on CBF features. The results indicate that after positional changes, short-term HRV increased in all volunteers, followed by the manifestation of three different change patterns. CBF features revealed two change patterns: increase and decrease. Different patterns were associated with the main mechanisms of CBF regulation and the degree of cardiac pulsation changes. Furthermore, the significance levels for distinguishing the two groups based on the two CBF features were less than 0.05 and 0.01, respectively. NFC technology shows potential in discriminating dynamic changes in CBF among different individuals.

In the development of brain-computer interfaces (BCIs), precise decoding of motor outputs is crucial. This study presents an enhanced Kalman filter approach that integrates electromyography (EMG) with electrocorticography (ECoG) to improve force estimation in pinching tasks. By incorporating EMG data as a state variable in the filter, we aim to account for musculoskeletal dynamics, enhancing the accuracy of force predictions. This integration significantly improves the decoding performance, particularly during dynamic force phases. The results confirm the importance of embedding musculoskeletal dynamics into ECoG-based BCIs, which may help improve prosthetic control and motor rehabilitation for people with motor impairments.

The experimental use of CMOS high-density neural probes enables the wide field observation of the electrical activity of neural circuits at the resolution of single neurons. Optogenetic light stimulation allows to control and modulate the activity of neural cells, in a genetically selective manner. The combination of these techniques can be a powerful approach for investigating mechanisms of brain diseases and of information processing in the brain. The main obstacle to combining such techniques, however, is the photosensitivity of CMOS circuits and consequent photoelectric artefacts affecting electrophysiological recordings. This paper presents a CMOS based high-density SiNAPS neural implant which was designed to be combined with a tapered optical fiber. A photoelectric shield was created on a 512-channel probe, where the electrodes are arranged in 4 columns with 29.5 μm electrode pitch. Results show the light sensitivity of the probe with and without the photoelectric shield and the capability of recording light-evoked responses in vivo. More than 180 neurons were recorded without any light-induced distortion of the electrophysiological signals.

Characterizing geometric and topological properties of protein structures encompassing surface pockets, interior cavities, and cross channels is important for understanding their functions. Our knowledge of protein structures has been greatly advanced by AI-powered structure prediction tools, with AlphaFold2 (AF2) providing accurate 3D structure predictions for most protein sequences. Nonetheless, there is a substantial lack of function annotations and corresponding functional surface topographical information. We develop a method to predict functional pockets, along with their associated Gene Ontology (GO) terms and Enzyme Commission (EC) numbers, for a set of 65,013 AF2-predicted human non-singleton representative structures, which can be mapped to 186,095 "non-fragment" AF2-predicted human protein structures. The identification of functional pockets, along with their respective GO terms and EC numbers, is achieved by combining topological data analysis and the deep learning method of DeepFRI. All predicted functional pockets for these 65,013 AF2-predicted human representative structures are accessible at: https://cfold.bme.uic.edu/castpfold.

Atrial Fibrillation (AF) is a heart rhythm disorder characterized by rapid and irregular atrial contractions, which can increase the risk of stroke and decrease patients' quality of life. One of the main techniques to treat AF is RF catheter ablation, which involves electrically isolating the pulmonary veins from the rest of the atrium, based on point lesions surrounding the veins. There is still discussion in the community as to what is the optimal interlesion distance to improve the long-term results of AF ablation. A Python tool has been developed that, starting from the AF ablation procedure data, finds the optimal sequence of ablations surrounding the pulmonary veins and thus can calculate the distance between the entire sequence of lesions. The automated algorithm proved to be effective in most cases and in almost all cases semi-automatically. The work provides a tool for the community that can help to optimize AF ablation. In the future, the algorithm could be improved to be 100% automatic, although right now it is already useful and several clinical studies are underway using this tool.

Deep learning models are being adopted and applied across various critical medical tasks, yet they are primarily trained to provide point predictions without providing degrees of confidence. Medical practitioner's trustworthiness of deep learning models is increased when paired with uncertainty estimations. Conformal Prediction has emerged as a promising method to pair machine learning models with prediction intervals, allowing for a view of the model's uncertainty. However, popular uncertainty estimation methods for conformal prediction fail to provide highly accurate heteroskedastic intervals. In this paper, we propose a method to estimate the uncertainty of each sample by calculating the variance obtained from a Deep Regression Forest. We show that the deep regression forest variance improves the efficiency and coverage of normalized inductive conformal prediction when applied on an anti-cancer drug sensitivity prediction task.

Electrical Impedance Tomography (EIT) is a non-invasive imaging technique used to monitor mechanically ventilated patients at the bedside. This study focuses on evaluating the estimation of pulmonary perfusion and the effects of two common noise sources on the outcomes of a hypertonic saline contrast bolus procedure. We used the first-pass kinetics modeling in EIT data to estimate purely lung and hybrid pixels. Furthermore, we analyzed how signal drift and cardiac partial volume effect can yield misleading outcomes. If left uncompensated, the drift noise showed an overestimation of 33%, while the cardiac partial volume effect showed an underestimation of 13.9% in the maximum slope value. We evaluated the model performance using simulated and actual data within physiologically feasible limits.