Paul Bentley

Background: Thrombolysis and mechanical thrombectomy represent the most successful stroke innovations over the last 30 years. Quantifying innovation in stroke is essential for identifying productive research lines and prioritizing funding, but health care lacks validated methods for measuring innovation. Objective: This study aimed to systematically evaluate the relationship between stroke-related patents and publications, demonstrate the feasibility of using large language models (LLMs) in this process, and identify the most rapidly advancing innovations in stroke care by mapping them to a theoretical innovation life cycle. Methods: The Open Patent Services (European Patent Office) and PubMed databases were searched between 1993 and 2023 for “stroke OR cerebrovascular.” In this bibliometric patent-publication analysis, a 13 billion–parameter Llama LLM was trained to identify patents related to stroke disease, as opposed to other references to the word “stroke,” on a manually labeled subset of 5000 patents and assessed using 5-fold cross-validation. The LLM filtered irrelevant results, and the resulting patent codes were grouped into innovation clusters. For each cluster, annual patent and publication counts were normalized to adjust for global trends. Cluster-specific growth curves were plotted to analyze the rates and characteristics of growth. The innovation life cycle stage for each innovation cluster was estimated by fitting a sigmoid curve to the patent and publication data consistent with the diffusion of innovations theory by Rogers. Results: The cross-validated accuracy of the LLM was 99.2%, with a sensitivity of 96.5% and a specificity of 99.6%. An initial bibliometric search retrieved 237,035 patents and 486,664 research publications. A manual review of a random sample of patents before filtering revealed that only 11.2% (56/500) were relevant to stroke. After LLM filtering, of the 237,035 patents, 28,225 (11.9%) stroke-related patents remained. These were grouped into 7 innovation clusters: pharmacological treatment, alternative medicine, rehabilitation devices, medical imaging, diagnostic testing, surgical devices, and artificial intelligence (AI) methods. Patent and publication counts were strongly correlated across clusters (Spearman rs=0.65-0.92; P<.006) except for pharmacological treatment (rs=0.09) and alternative medicine (rs=0.55). Pharmacological treatments were the top-performing cluster over the last 30 years, accounting for 49.3% (36,005/73,094) of all patents, but patent activity in this area has plateaued since the late 2000s. AI methods, rehabilitation devices, and medical imaging exhibited exponential rates of patent growth, with annual normalized increases of 39.2%, 15.9%, and 5.8% compared to 16.9%, 5.3%, and 2.2% for publications, respectively. Conclusions: Applying an LLM to publicly available patent and publication data provides a scalable way to quantify innovation in stroke. Pharmacological treatment appears to have entered a saturation phase, whereas AI methods, rehabilitation devices, and medical imaging remain in rapid growth, highlighting areas of greatest traction for future research and investment.

Inducing muscle activity with electric current - Functional Electrical Stimulation (FES) - is a key therapy for improving extremity function after neurological trauma (e.g. post-stroke). Positive outcomes, however, are heavily dependent on modulating stimulation to maximise response while limiting rapid FES-induced muscle fatigue. Feedback control is not possible today with standard electromyography (EMG); electrical stimulation obscures EMG measurement, severely limiting range of use and benefits. Mechanomyography (MMG), though underexplored, is immune to electrical artifacts, enabling closed-loop modulation. We introduce and validate a novel MMG-FES system to optimise therapeutic stimulation post-stroke. It combines a new wearable with two myography modalities (pressure P_MMG, microphone M_MMG), an MMG- Tibialis Anterior (TA) musculotendon model and an MMG-derived fatigue index. An isometric FES fatigue protocol is executed on control ($N=15$) and post-stroke ($N=3$) participants, recording force and MMG signals. The P_MMG Mean Value (MV) signals consistently decreased with fatigue, showing strong Pearson correlations ($\bar{r}$) with force decline in control ($\bar{r}=0.740$) and stroke ($\bar{r}=0.928$) groups ($p \le 0.005$). The P_MMG MV-driven model also accurately predicted force decline, achieving mean coefficients of determination ($R^{2}$) of 0.741 (control) and 0.774 (stroke), with strong prediction correlations ($\bar{r} > 0.87, p < 0.01$). Conversely, M_MMG prediction had weaker correlation to force and fatigue. We conclude: 1) the pressure-based P_MMG sensor is a robust, non-invasive FES-induced fatigue indicator; 2) the P_MMG-driven model allows continuous estimation of force capacity; and 3) the combined system enables closed-loop FES modulation to optimize rehabilitation.

BACKGROUND Thrombolysis and mechanical thrombectomy represent the most successful stroke innovations over the last 30 years. Quantifying innovation in stroke is essential for identifying productive research lines and prioritizing funding, but health care lacks validated methods for measuring innovation. OBJECTIVE This study aimed to systematically evaluate the relationship between stroke-related patents and publications, demonstrate the feasibility of using large language models (LLMs) in this process, and identify the most rapidly advancing innovations in stroke care by mapping them to a theoretical innovation life cycle. METHODS The Open Patent Services (European Patent Office) and PubMed databases were searched between 1993 and 2023 for “stroke OR cerebrovascular.” In this bibliometric patent-publication analysis, a 13 billion–parameter Llama LLM was trained to identify patents related to stroke disease, as opposed to other references to the word “stroke,” on a manually labeled subset of 5000 patents and assessed using 5-fold cross-validation. The LLM filtered irrelevant results, and the resulting patent codes were grouped into innovation clusters. For each cluster, annual patent and publication counts were normalized to adjust for global trends. Cluster-specific growth curves were plotted to analyze the rates and characteristics of growth. The innovation life cycle stage for each innovation cluster was estimated by fitting a sigmoid curve to the patent and publication data consistent with the diffusion of innovations theory by Rogers. RESULTS The cross-validated accuracy of the LLM was 99.2%, with a sensitivity of 96.5% and a specificity of 99.6%. An initial bibliometric search retrieved 237,035 patents and 486,664 research publications. A manual review of a random sample of patents before filtering revealed that only 11.2% (56/500) were relevant to stroke. After LLM filtering, of the 237,035 patents, 28,225 (11.9%) stroke-related patents remained. These were grouped into 7 innovation clusters: pharmacological treatment, alternative medicine, rehabilitation devices, medical imaging, diagnostic testing, surgical devices, and artificial intelligence (AI) methods. Patent and publication counts were strongly correlated across clusters (Spearman rs=0.65-0.92; P<.006) except for pharmacological treatment (rs=0.09) and alternative medicine (rs=0.55). Pharmacological treatments were the top-performing cluster over the last 30 years, accounting for 49.3% (36,005/73,094) of all patents, but patent activity in this area has plateaued since the late 2000s. AI methods, rehabilitation devices, and medical imaging exhibited exponential rates of patent growth, with annual normalized increases of 39.2%, 15.9%, and 5.8% compared to 16.9%, 5.3%, and 2.2% for publications, respectively. CONCLUSIONS Applying an LLM to publicly available patent and publication data provides a scalable way to quantify innovation in stroke. Pharmacological treatment appears to have entered a saturation phase, whereas AI methods, rehabilitation devices, and medical imaging remain in rapid growth, highlighting areas of greatest traction for future research and investment.

Estimating progression of acute ischemic brain lesions – or biological lesion age - holds huge practical importance for hyperacute stroke management. The current best method for determining lesion age from non-contrast computerised tomography (NCCT), measures Relative Intensity (RI), termed Net Water Uptake (NWU). We optimised lesion age estimation from NCCT using a convolutional neural network – radiomics (CNN-R) model trained upon chronometric lesion age (Onset Time to Scan: OTS), while validating against chronometric and biological lesion age in external datasets (N = 1945). Coefficients of determination (R²) for OTS prediction, using CNN-R, and RI models were 0.58 and 0.32 respectively; while CNN-R estimated OTS showed stronger associations with ischemic core:penumbra ratio, than RI and chronometric, OTS (ρ² = 0.37, 0.19, 0.11); and with early lesion expansion (regression coefficients >2x for CNN-R versus others) (all comparisons: p < 0.05). Concluding, deep-learning analytics of NCCT lesions is approximately twice as accurate as NWU for estimating chronometric and biological lesion ages.

Background: Although previous studies investigated the main predictors of outcomes after endovascular thrombectomy (EVT) in patients aged 80 years and older, less is known about the impact of the procedural features on outcomes in elderly patients. The aim of this study was to investigate the influence of EVT technical procedures on the main 3-month outcomes in a population of patients aged 80 years and older. Methods: This observational, prospective, single-centre study included consecutive patients with acute LVO ischaemic stroke of the anterior circulation. The study outcomes were functional independence at 3 months after EVT (defined as a mRS score of 0–2), successful reperfusion (mTICI ≥ 2b), incidence of haeamorrhagic transformation, and 90-day all cause of mortality. Results: Our cohort included 497 patients with acute ischaemic stroke due to LVO treated with EVT. Among them, 105 (21.1%) patients were aged ≥ 80 years. In the elderly group, multivariable regression analysis showed that thromboaspiration technique vs stent-retriever was the single independent predictor of favourable post-procedural TICI score (OR = 7.65, 95%CI = 2.22–26.32, p = 0.001). Conclusions: Our study suggests that EVT for LVO stroke in the elderly could be safe. The use of thromboaspiration was associated with positive reperfusion outcome in this population. Further studies in larger series are warranted to confirm the present results and to evaluate the safety and efficacy of EVT in the elderly and oldest adults.

Background: Upper limb (UL) recovery after stroke is strongly dependent upon rehabilitation dose. Rehabilitation technologies present pragmatic solutions to dose enhancement, complementing therapeutic activity within conventional rehabilitation, connecting clinicians with patients remotely, and empowering patients to drive their own recovery. To date, rehabilitation technologies have been poorly adopted. Understanding the barriers to adoption may shape strategies to enhance technology use and therefore increase rehabilitation dose, thus optimizing recovery potential. Objective: We examined the usability, acceptability, and adoption of a self-directed, exercise-gaming technology within a heterogeneous stroke survivor cohort and investigated how stroke survivor characteristics, technology usability, and attitudes toward technology influenced adoption. Methods: A feasibility study of a novel exercise-gaming technology for self-directed UL rehabilitation in early subacute stroke survivors (N=30) was conducted in an inpatient, acute hospital setting. Demographic and clinical characteristics were recorded; participants’ performance in using the system (usability) was assessed using a 4-point performance rating scale (adapted from the Barthel index), and adherence with the system was electronically logged throughout the trial. The technology acceptance model was used to formulate a survey examining the acceptability of the system. Spearman rank correlations were used to examine associations between participant characteristics, user performance (usability), end-point technology acceptance, and intervention adherence (adoption). Results: The technology was usable for 87% (n=26) of participants, and the overall technology acceptance rating was 68% (95% CI 56%-79%). Participants trained with the device for a median of 26 (IQR 16-31) minutes daily over an enrollment period of 8 (IQR 5-14) days. Technology adoption positively correlated with user performance (usability) (ρ=0.55; 95% CI 0.23-0.75; P=.007) and acceptability as well as domains of perceived usefulness (ρ=0.42; 95% CI 0.09-0.68; P=.03) and perceived ease of use (ρ=0.46; 95% CI 0.10-0.74; P=.02). Technology acceptance decreased with increased global stroke severity (ρ=−0.56; 95% CI −0.79 to −0.22; P=.007). Conclusions: This technology was usable and acceptable for the majority of the cohort, who achieved an intervention dose with technology-facilitated, self-directed UL training that exceeded conventional care norms. Technology usability and acceptability were determinants of adoption and appear to be mediated by stroke severity. The results demonstrate the importance of selecting technologies for stroke survivors on the basis of individual needs and abilities, as well as optimizing the accessibility of technologies for the target user group. Facilitating changes in stroke survivors’ beliefs and attitudes toward rehabilitation technologies may enhance adoption. Further work is needed to understand how technology can be optimized to benefit those with more severe stroke.

The cornerstone of stroke care is expedient management that varies depending on the time since stroke onset. Consequently, clinical decision making is centered on accurate knowledge of timing and often requires a radiologist to interpret Computed Tomography (CT) of the brain to confirm the occurrence and age of an event. These tasks are particularly challenging due to the subtle expression of acute ischemic lesions and the dynamic nature of their appearance. Automation efforts have not yet applied deep learning to estimate lesion age and treated these two tasks independently, so, have overlooked their inherent complementary relationship. To leverage this, we propose a novel end-to-end multi-task transformer-based network optimized for concurrent segmentation and age estimation of cerebral ischemic lesions. By utilizing gated positional self-attention and CT-specific data augmentation, the proposed method can capture long-range spatial dependencies while maintaining its ability to be trained from scratch under low-data regimes commonly found in medical imaging. Furthermore, to better combine multiple predictions, we incorporate uncertainty by utilizing quantile loss to facilitate estimating a probability density function of lesion age. The effectiveness of our model is then extensively evaluated on a clinical dataset consisting of 776 CT images from two medical centers. Experimental results demonstrate that our method obtains promising performance, with an area under the curve (AUC) of 0.933 for classifying lesion ages ≤4.5 hours compared to 0.858 using a conventional approach, and outperforms task-specific state-of-the-art algorithms.