A critical decision-step in the emergency treatment of ischemic stroke is whether or not to administer thrombolysis — a treatment that can result in good recovery, or deterioration due to symptomatic intracranial haemorrhage (SICH). Certain imaging features based upon early computerized tomography (CT), in combination with clinical variables, have been found to predict SICH, albeit with modest accuracy. In this proof-of-concept study, we determine whether machine learning of CT images can predict which patients receiving tPA will develop SICH as opposed to showing clinical improvement with no haemorrhage. Clinical records and CT brains of 116 acute ischemic stroke patients treated with intravenous thrombolysis were collected retrospectively (including 16 who developed SICH). The sample was split into training (n = 106) and test sets (n = 10), repeatedly for 1760 different combinations. CT brain images acted as inputs into a support vector machine (SVM), along with clinical severity. Performance of the SVM was compared with established prognostication tools (SEDAN and HAT scores; original, or after adaptation to our cohort). Predictive performance, assessed as area under receiver-operating-characteristic curve (AUC), of the SVM (0.744) compared favourably with that of prognostic scores (original and adapted versions: 0.626–0.720; p < 0.01). The SVM also identified 9 out of 16 SICHs, as opposed to 1–5 using prognostic scores, assuming a 10% SICH frequency (p < 0.001). In summary, machine learning methods applied to acute stroke CT images offer automation, and potentially improved performance, for prediction of SICH following thrombolysis. Larger-scale cohorts, and incorporation of advanced imaging, should be tested with such methods.

SignificanceSimple voluntary movements (e.g., reaching or gripping) deteriorate with distraction, suggesting that the attention-control system—which suppresses distraction—influences motor control. Here, we tested the causal dependency of simple movements on attention control, and its neuroanatomical basis, in healthy elderly and patients with focal brain lesions. Not only did we find that attention control correlates with motor performance, correcting for lesion size, fatigue, etc., but we found a revealing pattern of dissociations: Severe motor impairment could occur with normal attention control whereas impaired attention control never occurred with disproportionately milder motor impairment—suggesting that attention control is required for normal motor performance. One implication is that a component of stroke paralysis arises from poor attentional control, which could itself be a therapeutic target.

More than half of all stroke patients can be expected to have a lesion location classifiable into 1 of the 3 principal attention networks. Our results have potential implications for therapy personalization in focal brain diseases including stroke.

BackgroundMultiplayer video games promoting exercise-based rehabilitation may facilitate motor learning, by increasing motivation through social interaction. However, a major design challenge is to enable meaningful inter-subject interaction, whilst allowing for significant skill differences between players. We present a novel motor-training paradigm that allows real-time collaboration and performance enhancement, across a wide range of inter-subject skill mismatches, including disabled vs. able-bodied partnerships.MethodsA virtual task consisting of a dynamic ball on a beam, is controlled at each end using independent digital force-sensing handgrips. Interaction is mediated through simulated physical coupling and locally-redundant control. Game performance was measured in 16 healthy-healthy and 16 patient-expert dyads, where patients were hemiparetic stroke survivors using their impaired arm. Dual-player was compared to single-player performance, in terms of score, target tracking, stability, effort and smoothness; and questionnaires probing user-experience and engagement.ResultsPerformance of less-able subjects (as ranked from single-player ability) was enhanced by dual-player mode, by an amount proportionate to the partnership’s mismatch. The more abled partners’ performances decreased by a similar amount. Such zero-sum interactions were observed for both healthy-healthy and patient-expert interactions. Dual-player was preferred by the majority of players independent of baseline ability and subject group; healthy subjects also felt more challenged, and patients more skilled.ConclusionThis is the first demonstration of implicit skill balancing in a truly collaborative virtual training task leading to heightened engagement, across both healthy subjects and stroke patients.Electronic supplementary materialThe online version of this article (doi:10.1186/s12984-017-0319-x) contains supplementary material, which is available to authorized users.

Motor-training software on tablets or smartphones (Apps) offer a low-cost, widely-available solution to supplement arm physiotherapy after stroke. We assessed the proportions of hemiplegic stroke patients who, with their plegic hand, could meaningfully engage with mobile-gaming devices using a range of standard control-methods, as well as by using a novel wireless grip-controller, adapted for neurodisability. We screened all newly-diagnosed hemiplegic stroke patients presenting to a stroke centre over 6 months. Subjects were compared on their ability to control a tablet or smartphone cursor using: finger-swipe, tap, joystick, screen-tilt, and an adapted handgrip. Cursor control was graded as: no movement (0); less than full-range movement (1); full-range movement (2); directed movement (3). In total, we screened 345 patients, of which 87 satisfied recruitment criteria and completed testing. The commonest reason for exclusion was cognitive impairment. Using conventional controls, the proportion of patients able to direct cursor movement was 38–48%; and to move it full-range was 55–67% (controller comparison: p>0.1). By comparison, handgrip enabled directed control in 75%, and full-range movement in 93% (controller comparison: p<0.001). This difference between controllers was most apparent amongst severely-disabled subjects, with 0% achieving directed or full-range control with conventional controls, compared to 58% and 83% achieving these two levels of movement, respectively, with handgrip. In conclusion, hand, or arm, training Apps played on conventional mobile devices are likely to be accessible only to mildly-disabled stroke patients. Technological adaptations such as grip-control can enable more severely affected subjects to engage with self-training software.

Background and Purpose Elevated homocysteine (tHcy) levels are known to be associated with increased risk of ischemic stroke (IS). Given that both tHcy and IS are heritable traits, we investigated a potential genetic relationship between homocysteine levels and stroke risk by assessing 18 polymorphisms previously associated with tHcy levels for their association with IS and its subtypes. Methods Previous meta-analysis results from an international stroke collaborative network, METASTROKE, were utilized to assess association of the 18 tHcy associated SNPs in 12,389 IS cases and 62,004 controls. We also investigated the associations in regions located within 50kb from the 18 tHcy related SNPs, and the association of a genetic risk score including the 18 SNPs. Results One SNP located in the RASIP1 gene and a cluster of three SNPs located at and near SLC17A3 were significantly associated with IS (P<0.0003) after correcting for multiple testing. For stroke subtypes, the sentinel SNP located upstream of MUT was significantly associated with SVD (small vessel disease) (P=0.0022), while one SNP located in MTHFR was significantly associated with LVD (large vessel disease) (P=0.00019). A genetic risk score including the 18 SNPs did not show significant association with IS or its subtypes. Conclusions This study found several potential associations with IS and its subtypes: an association of an MUT variant with SVD, an MTHFR variant with LVD, and associations of RASIP1 and SLC17A3 variants with overall IS.

Cerebral small vessel disease (SVD) is a common cause of ageing-associated physical and cognitive impairment. Identifying SVD is important for both clinical and research purposes but is usually dependent on radiologists' evaluation of brain scans. Computer tomography (CT) is the most widely used brain imaging technique but for SVD it shows a low signal-to-noise ratio, and consequently poor inter-rater reliability. We therefore propose a novel framework based on multiple instance learning (MIL) to distinguish between absent/mild SVD and moderate/severe SVD. Intensity patches are extracted from regions with high probability of containing lesions. These are then used as instances in MIL for the identification of SVD. A large baseline CT dataset, consisting of 590 CT scans, was used for evaluation. We achieved approximately 75% accuracy in classifying two different types of SVD, which is high for this challenging problem. Our results outperform those obtained by either standard machine learning methods or current clinical practice.

The strongest stroke lesion location associations with 1-week recovery were identified, and it was shown that anatomical information accounts for a sizeable proportion of early recovery variability.