To develop and validate deep learning models leveraging CT imaging for the prediction and classification of brain stroke conditions, with the potential to enhance accuracy and support clinical decision-making. This retrospective, bi-center study included data from 250 patients, with a dataset of 8186 CT images collected from 2017 to 2022. Two AI models were developed using the Expanded ResNet101 deep learning framework as a two-step model. Model performance was evaluated using confusion matrices, supplemented by external validation with an independent dataset. External validation was conducted by an expert and two external members. Overall accuracy, confidence intervals, Cohen's Kappa value, and McNemar's test P-values were calculated. A total of 8186 CT images were incorporated, with 6386 images used for the training and 900 datasets for testing and validation in Model 01. Further, 1619 CT images were used for training and 600 datasets for testing and validation in Model 02. The average accuracy, precision, and F1 score for both models were assessed: Model 01 achieved 99.6 %, 99.4 %, and 99.6 % respectively, whereas Model 02 achieved 99.2 %, 98.8 %, and 99.1 %. The external validation accuracies were 78.6 % (95 % CI: 0.73,0.83; P < 0.001) and 60.2 % (95 % CI: 0.48,0.70; P < 0.001) for Models 01 and 02 respectively, as evaluated by the expert. Deep learning models demonstrated high accuracy, precision, and F1 scores in predicting outcomes for brain stroke patients. With larger cohort and diverse radiologic mimics, these models could support clinicians in prognosis and decision-making.

Background: Although quite a very few studies have tested structural connectivity changes following an intervention, it reflects only selected key brain regions in the motor network. Thus, the understanding of structural connectivity changes related to the motor recovery process remains unclear. Objective: This study investigated structural connectivity changes of the motor execution network following a combined intervention of low-frequency repetitive transcranial magnetic stimulation (LF-rTMS) and intensive occupational therapy (OT) after a stroke using graph theory approach. Methods: Fifty-six stroke patients underwent Fugl-Meyer Assessment (FMA), Wolf Motor Function Test-Functional Ability Scale (WMFT-FAS), diffusion tensor imaging (DTI), and T1 weighted imaging before and after the intervention. We examined graph theory measures related to twenty brain regions using structural connectomes. Results: The ipsilesional and contralesional hemisphere showed structural connectivity changes post-intervention after stroke. We found significantly increased regional centralities and nodal efficiency within the frontal pole and decreased degree centrality and nodal efficiency in the ipsilesional thalamus. Correlations were found between network measures and clinical assessments in the cuneus, postcentral gyrus, precentral gyrus, and putamen of the ipsilesional hemisphere. The contralesional areas such as the caudate, cerebellum, and frontal pole also showed significant correlations. Conclusions: This study was helpful to expand the understanding of structural connectivity changes in both hemispheric networks during the motor recovery process following LF-rTMS and intensive OT after stroke.

Purpose: To evaluate activity changes associated with the intervention of low-frequency repetitive transcranial magnetic stimulation (rTMS) and intensive occupational therapy (OT) after stroke using functional magnetic resonance (fMRI). Methods: Seventy stroke patients were scanned while performing finger tapping tasks twice, before and 12 days after the intervention. Recovery of motor functions assessed using Fugl-Meyer Assessment (FMA) and Wolf Motor Function Test-Functional Ability Scale (WMFT-FAS) for upper extremity at each time point. An fMRI analysis was performed, and a region of interest (ROI) analysis was conducted using percentage signal changes (% SC) to determine the magnitude of activation. Results: FMA and WMFT-FAS were significantly increased from pre-intervention to post-intervention. Intervention related activations were seen in the ipsilesional premotor cortex (PMC) and primary motor cortex (M1), thalamo-cortico regions with the paretic hand movements. With the unaffected hand movements, significant clusters in the contralesional primary somatosensory cortex (S1), superior parietal cortex, and bilateral cerebellum were observed. The ROI-based analysis revealed that ipsilesional M1, contralesional PMC, and supplementary motor area (SMA) showed significantly higher results with the paretic hand movements, a trend toward a significant decrease in the contralesional S1 with the unaffected hand movements from the pre-intervention to post-intervention. Conclusions: Our findings suggest that gains in motor functions produced by the intervention of rTMS and intensive OT in hemiparesis stroke patients may be associated with the ipsilesional hemisphere and contralesional hemisphere as well. Identifying rTMS and OT intervention based on cortical patterns may help to implement rTMS in motor rehabilitation after stroke.

In recent years, neurorehabilitation has been actively used to treat motor paralysis after stroke. However, the impacts of rehabilitation on neural networks in the brain remain largely unknown. Therefore, we investigated changes in structural neural networks after rehabilitation therapy in patients who received a combination of low-frequency repetitive transcranial magnetic stimulation (LF-rTMS) and intensive occupational therapy (intensive-OT) as neurorehabilitation. Fugl-Meyer assessment (FMA) for upper extremity (FMA-UE) and Action Research Arm Test (ARAT), both of which reflected upper limb motor function, were conducted before and after rehabilitation therapy. At the same time, diffusion tensor imaging (DTI) and three-dimensional T1-weighted imaging (3D T1WI) were performed. After analyzing the structural connectome based on DTI data, measures related to connectivity in neural networks were calculated using graph theory. Rehabilitation therapy prompted a significant increase in connectivity with the isthmus of the cingulate gyrus in the ipsilesional hemisphere (p < 0.05) in patients with left-sided paralysis, as well as a significant decrease in connectivity with the ipsilesional postcentral gyrus (p < 0.05). These results indicate that LF-rTMS combined with intensive-OT may facilitate motor function recovery by enhancing the functional roles of networks in motor-related areas of the ipsilesional cerebral hemisphere.