Abstract
Patients with severe acquired brain injury and prolonged disorders of consciousness (pDoC) are characterized by high clinical complexity and high risk to develop medical complications. The ...present multi-center longitudinal study aimed at investigating the impact of medical complications on the prediction of clinical outcome by means of machine learning models. Patients with pDoC were consecutively enrolled at admission in 23 intensive neurorehabilitation units (IRU) and followed-up at 6 months from onset via the Glasgow Outcome Scale—Extended (GOSE). Demographic and clinical data at study entry and medical complications developed within 3 months from admission were collected. Machine learning models were developed, targeting neurological outcomes at 6 months from brain injury using data collected at admission. Then, after concatenating predictions of such models to the medical complications collected within 3 months, a cascade model was developed. One hundred seventy six patients with pDoC (M: 123, median age 60.2 years) were included in the analysis. At admission, the best performing solution (k-Nearest Neighbors regression, KNN) resulted in a median validation error of 0.59 points IQR 0.14 and a classification accuracy of dichotomized GOS-E of 88.6%. Coherently, at 3 months, the best model resulted in a median validation error of 0.49 points IQR 0.11 and a classification accuracy of 92.6%. Interpreting the admission KNN showed how the negative effect of older age is strengthened when patients’ communication levels are high and ameliorated when no communication is present. The model trained at 3 months showed appropriate adaptation of the admission prediction according to the severity of the developed medical complexity in the first 3 months. In this work, we developed and cross-validated an interpretable decision support tool capable of distinguishing patients which will reach sufficient independence levels at 6 months (GOS-E > 4). Furthermore, we provide an updated prediction at 3 months, keeping in consideration the rehabilitative path and the risen medical complexity.
Rehabilitation medicine is facing a new development phase thanks to a recent wave of rigorous clinical trials aimed at improving the scientific evidence of protocols. This phenomenon, combined with ...new trends in personalised medical therapies, is expected to change clinical practice dramatically. The emerging field of Rehabilomics is only possible if methodologies are based on biomedical data collection and analysis. In this framework, the objective of this work is to develop a systematic review of machine learning algorithms as solutions to predict motor functional recovery of post-stroke patients after treatment.
We conducted a comprehensive search of five electronic databases using the Patient, Intervention, Comparison and Outcome (PICO) format. We extracted health conditions, population characteristics, outcome assessed, the method for feature extraction and selection, the algorithm used, and the validation approach. The methodological quality of included studies was assessed using the prediction model risk of bias assessment tool (PROBAST). A qualitative description of the characteristics of the included studies as well as a narrative data synthesis was performed.
A total of 19 primary studies were included. The predictors most frequently used belonged to the areas of demographic characteristics and stroke assessment through clinical examination. Regarding the methods, linear and logistic regressions were the most frequently used and cross-validation was the preferred validation approach.
We identified several methodological limitations: small sample sizes, a limited number of external validation approaches, and high heterogeneity among input and output variables. Although these elements prevented a quantitative comparison across models, we defined the most frequently used models given a specific outcome, providing useful indications for the application of more complex machine learning algorithms in rehabilitation medicine.
Stroke related motor function deficits affect patients' likelihood of returning to professional activities, limit their participation in society and functionality in daily living. Hence, robot-aided ...gait rehabilitation needs to be fruitful and effective from a motor learning perspective. For this reason, optimal human-robot interaction strategies are necessary to foster neuroplastic shaping during therapy. Therefore, we performed a systematic search on the effects of different control algorithms on quantitative objective gait parameters of post-acute stroke patients.
We conducted a systematic search on four electronic databases using the Population Intervention Comparison and Outcome format. The heterogeneity of performance assessment, study designs and patients' numerosity prevented the possibility to conduct a rigorous meta-analysis, thus, the results were presented through narrative synthesis.
A total of 31 studies (out of 1036) met the inclusion criteria, without applying any temporal constraints. No controller preference with respect to gait parameters improvements was found. However, preferred solutions were encountered in the implementation of force control strategies mostly on rigid devices in therapeutic scenarios. Conversely, soft devices, which were all position-controlled, were found to be more commonly used in assistive scenarios. The effect of different controllers on gait could not be evaluated since conspicuous heterogeneity was found for both performance metrics and study designs.
Overall, due to the impossibility of performing a meta-analysis, this systematic review calls for an outcome standardisation in the evaluation of robot-aided gait rehabilitation. This could allow for the comparison of adaptive and human-dependent controllers with conventional ones, identifying the most suitable control strategies for specific pathologic gait patterns. This latter aspect could bolster individualized and personalized choices of control strategies during the therapeutic or assistive path.
Abstract
Background
Rehabilitation treatments and services are essential for the recovery of post-stroke patients’ functions; however, the increasing number of available therapies and the lack of ...consensus among outcome measures compromises the possibility to determine an appropriate level of evidence. Machine learning techniques for prognostic applications offer accurate and interpretable predictions, supporting the clinical decision for personalised treatment. The aim of this study is to develop and cross-validate predictive models for the functional prognosis of patients, highlighting the contributions of each predictor.
Methods
A dataset of 278 post-stroke patients was used for the prediction of the class transition, obtained from the modified Barthel Index. Four classification algorithms were cross-validated and compared. On the best performing model on the validation set, an analysis of predictors contribution was conducted.
Results
The Random Forest obtained the best overall results on the accuracy (76.2%), balanced accuracy (74.3%), sensitivity (0.80), and specificity (0.68). The combination of all the classification results on the test set, by weighted voting, reached 80.2% accuracy. The predictors analysis applied on the Support Vector Machine, showed that a good trunk control and communication level, and the absence of bedsores retain the major contribution in the prediction of a good functional outcome.
Conclusions
Despite a more comprehensive assessment of the patients is needed, this work paves the way for the implementation of solutions for clinical decision support in the rehabilitation of post-stroke patients. Indeed, offering good prognostic accuracies for class transition and patient-wise view of the predictors contributions, it might help in a personalised optimisation of the patients’ rehabilitation path.
The inflammatory, reparative and regenerative mechanisms activated in ischemic stroke patients immediately after the event cooperate in the response to injury, in the restoration of functions and in ...brain remodeling even weeks after the event and can be sustained by the rehabilitation treatment. Nonetheless, patients' response to treatments is difficult to predict because of the lack of specific measurable markers of recovery, which could be complementary to clinical scales in the evaluation of patients. Considering that Extracellular Vesicles (EVs) are carriers of multiple molecules involved in the response to stroke injury, in the present study, we have identified a panel of EV-associated molecules that (i) confirm the crucial involvement of EVs in the processes that follow ischemic stroke, (ii) could possibly profile ischemic stroke patients at the beginning of the rehabilitation program, (iii) could be used in predicting patients' response to treatment. By means of a multiplexing Surface Plasmon Resonance imaging biosensor, subacute ischemic stroke patients were proven to have increased expression of vascular endothelial growth factor receptor 2 (VEGFR2) and translocator protein (TSPO) on the surface of small EVs in blood. Besides, microglia EVs and endothelial EVs were shown to be significantly involved in the intercellular communications that occur more than 10 days after ischemic stroke, thus being potential tools for the profiling of patients in the subacute phase after ischemic stroke and in the prediction of their recovery.
COVID-19 cases are increasing around the globe with almost 5 million of deaths. We propose here a deep learning model capable of predicting the duration of the infection by means of information ...available at hospital admission. A total of 222 patients were enrolled in our observational study. Anagraphical and anamnestic data, COVID-19 signs and symptoms, COVID-19 therapy, hematochemical test results, and prior therapies administered to patients are used as predictors. A set of 55 features, all of which can be taken in the first hours of the patient’s hospitalization, was considered. Different solutions were compared achieving the best performance with a sequential convolutional neural network-based model merged in an ensemble with two different meta-learners linked in cascade. We obtained a median absolute error of 2.7 days (IQR = 3.0) in predicting the duration of the infection; the error was equally distributed in the infection duration range. This tool could preemptively give an outlook of the COVID-19 patients’ expected path and the associated hospitalization effort. The proposed solution could be viable in tackling the huge burden and the logistics complexity of hospitals or rehabilitation centers during the pandemic waves.
Graphical abstract
With data taken ad admission, entering a PCA-based feature selection, a
k
-fold cross-validated CNN-based model was implemented. After external texting, a median absolute error of 2.7 days IQR = 3 days
Poor dynamic balance and impaired gait adaptation to different contexts are hallmarks of people with neurological disorders (PwND), leading to difficulties in daily life and increased fall risk. ...Frequent assessment of dynamic balance and gait adaptability is therefore essential for monitoring the evolution of these impairments and/or the long-term effects of rehabilitation. The modified dynamic gait index (mDGI) is a validated clinical test specifically devoted to evaluating gait facets in clinical settings under a physiotherapist's supervision. The need of a clinical environment, consequently, limits the number of assessments. Wearable sensors are increasingly used to measure balance and locomotion in real-world contexts and may permit an increase in monitoring frequency. This study aims to provide a preliminary test of this opportunity by using nested cross-validated machine learning regressors to predict the mDGI scores of 95 PwND via inertial signals collected from short steady-state walking bouts derived from the 6-minute walk test. Four different models were compared, one for each pathology (multiple sclerosis, Parkinson's disease, and stroke) and one for the pooled multipathological cohort. Model explanations were computed on the best-performing solution; the model trained on the multipathological cohort yielded a median (interquartile range) absolute test error of 3.58 (5.38) points. In total, 76% of the predictions were within the mDGI's minimal detectable change of 5 points. These results confirm that steady-state walking measurements provide information about dynamic balance and gait adaptability and can help clinicians identify important features to improve upon during rehabilitation. Future developments will include training of the method using short steady-state walking bouts in real-world settings, analysing the feasibility of this solution to intensify performance monitoring, providing prompt detection of worsening/improvements, and complementing clinical assessments.
Patients with Disorder of Consciousness (DoC) entering Intensive Rehabilitation Units after a severe Acquired Brain Injury have a highly variable evolution of the state of consciousness which is a ...complex aspect to predict. Besides clinical factors, electroencephalography has clearly shown its potential into the identification of prognostic biomarkers of consciousness recovery. In this retrospective study, with a dataset of 271 patients with DoC, we proposed three different Elastic-Net regressors trained on different datasets to predict the Coma Recovery Scale-Revised value at discharge based on data collected at admission. One dataset was completely EEG-based, one solely clinical data-based and the last was composed by the union of the two. Each model was optimized, validated and tested with a robust nested cross-validation pipeline. The best models resulted in a median absolute test error of 4.54 IQR = 4.56, 3.39 IQR = 4.36, 3.16 IQR = 4.13 for respectively the EEG, clinical and hybrid model. Furthermore, the hybrid model for what concerns overcoming an unresponsive wakefulness state and exiting a DoC results in an AUC of 0.91 and 0.88 respectively. Small but useful improvements are added by the EEG dataset to the clinical model for what concerns overcoming an unresponsive wakefulness state. Data-driven techniques and namely, machine learning models are hereby shown to be capable of supporting the complex decision-making process the practitioners must face.
Purpose
COVID-19 pandemic has affected most components of health systems including rehabilitation. The study aims to compare demographic and clinical data of patients admitted to an intensive ...rehabilitation unit (IRU) after severe acquired brain injuries (sABIs), before and during the pandemic.
Materials and methods
In this observational retrospective study, all patients admitted to the IRU between 2017 and 2020 were included. Demographics were collected, as well as data from the clinical and functional assessment at admission and discharge from the IRU. Patients were grouped in years starting from March 2017, and the 2020/21 cohort was compared to those admitted between March 2017/18, 2018/19, and 2019/20. Lastly, the pooled cohort March 2017 to March 2020 was compared with the COVID-19 year alone.
Results
This study included 251 patients (
F
: 96 (38%): median age 68 years IQR = 19.25, median time post-onset at admission: 42 days, IQR = 23). In comparison with the pre-pandemic years, a significant increase of hemorrhagic strokes (
p
< 0.001) and a decrease of traumatic brain injuries (
p
= 0.048), a reduction of the number of patients with a prolonged disorder of consciousness admitted to the IRU (
p
< 0.001) and a lower length of stay (
p
< 0.001) were observed in 2020/21.
Conclusions
These differences in the case mix of sABI patients admitted to IRU may be considered another side-effect of the pandemic. Facing this health emergency, rehabilitation specialists need to adapt readily to the changing clinical and functional needs of patients’ addressing the IRUs.
Detecting signs of residual neural activity in patients with altered states of consciousness is a crucial issue for the customization of neurorehabilitation treatments and clinical decision-making. ...With this large observational prospective study, we propose an innovative approach to detect residual signs of consciousness via the assessment of the amount of autonomic information coded within the brain. The latter was estimated by computing the mutual information (MI) between preprocessed EEG and ECG signals, to be then compared across consciousness groups, together with the absolute power and an international qualitative labeling. One-hundred seventy-four patients (73 females, 42%) were included in the study (median age of 65 years IQR = 20, MCS +: 29, MCS -: 23, UWS: 29). Electroencephalography (EEG) information content was found to be mostly related to the coding of electrocardiography (ECG) activity, i.e., with higher MI (p < 0.05), in Unresponsive Wakefulness Syndrome and Minimally Consciousness State minus (MCS -). EEG-ECG MI, besides clearly discriminating patients in an MCS - and +, significantly differed between lesioned areas (sides) in a subgroup of unilateral hemorrhagic patients. Crucially, such an accessible and non-invasive measure of residual consciousness signs was robust across electrodes and patient groups. Consequently, exiting from a strictly neuro-centric consciousness detection approach may be the key to provide complementary insights for the objective assessment of patients' consciousness levels and for the patient-specific planning of rehabilitative interventions.