Accuracy of commercial large language model (ChatGPT) to predict the diagnosis for prehospital patients suitable for ambulance transport decisions: Diagnostic accuracy study

Prehospital Emergency Care, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 5

Published: Jan. 31, 2025

While ambulance transport decisions guided by artificial intelligence (AI) could be useful, little is known of the accuracy AI in making patient diagnoses based on pre-hospital care report (PCR). The primary objective this study was to assess ChatGPT (OpenAI, Inc., San Francisco, CA, USA) predict a patient's diagnosis using PCR comparing reference standard assigned experienced paramedics. secondary classify cases where did not agree with as paramedic correct, or equally correct. This diagnostic used zero-shot learning model and greedy decoding. A convenience sample PCRs from students analyzed an untrained ChatGPT-4 determine single most likely diagnosis. provided reviewing each giving differential three items. trained prehospital professional assessed concordant non-concordant one diagnoses. If non-concordant, two board-certified emergency physicians independently decided if more diagnosed 78/104 (75.0%) correctly (95% confidence interval: 65.3-82.7%). Among 26 disagreement, judgment that 6/26 (23.0%) There only case 104 (0.96%) would have been potentially dangerous (under-triage). In study, overall diagnose patients their medical services 75.0%. considered less than diagnosis, commonly critical diagnosis-potentially leading over-triage. under-triage rate <1%.

Language: Английский

---

Integrating AI-driven wearable devices and biometric data into stroke risk assessment: A review of opportunities and challenges

Clinical Neurology and Neurosurgery, Journal Year: 2024, Volume and Issue: 249, P. 108689 - 108689

Published: Dec. 10, 2024

Stroke is a leading cause of morbidity and mortality worldwide, early detection risk factors critical for prevention improved outcomes. Traditional stroke assessments, relying on sporadic clinical visits, fail to capture dynamic changes in such as hypertension atrial fibrillation (AF). Wearable technology (devices), combined with biometric data analysis, offers transformative approach by enabling continuous monitoring physiological parameters. This narrative review was conducted using systematic identify analyze peer-reviewed articles, reports, case studies from reputable scientific databases. The search strategy focused articles published between 2010 till date pre-determined keywords. Relevant were selected based their focus wearable devices AI-driven technologies prevention, diagnosis, rehabilitation. literature categorized thematically explore applications, opportunities, challenges, future directions. explores the current landscape assessment, focusing role detection, personalized care, integration into practice. highlights opportunities presented predictive analytics, where algorithms can provide tailored interventions. Personalized powered machine learning, enable individualized care plans. Furthermore, telemedicine facilitates remote patient rehabilitation, particularly underserved areas. Despite these advances, challenges remain. Issues accuracy, privacy concerns, wearables healthcare systems must be addressed fully realize potential. As evolves, its application could revolutionize improving outcomes reducing global burden stroke.

Language: Английский

---

Artificial intelligence in stroke risk assessment and management via retinal imaging

Frontiers in Computational Neuroscience, Journal Year: 2025, Volume and Issue: 19

Published: Feb. 17, 2025

Retinal imaging, used for assessing stroke-related retinal changes, is a non-invasive and cost-effective method that can be enhanced by machine learning deep algorithms, showing promise in early disease detection, severity grading, prognostic evaluation stroke patients. This review explores the role of artificial intelligence (AI) patient care, focusing on imaging integration into clinical workflows. has revealed several microvascular including decrease central artery diameter an increase vein diameter, both which are associated with lacunar intracranial hemorrhage. Additionally, such as arteriovenous nicking, increased vessel tortuosity, arteriolar light reflex, decreased fractals, thinning nerve fiber layer also reported to higher risk. AI models, Xception EfficientNet, have demonstrated accuracy comparable traditional risk scoring systems predicting For diagnosis, models like Inception, ResNet, VGG, alongside classifiers, shown high efficacy distinguishing patients from healthy individuals using imaging. Moreover, random forest model effectively distinguished between ischemic hemorrhagic subtypes based features, superior predictive performance compared characteristics. support vector achieved classification pial collateral status. Despite this advancements, challenges lack standardized protocols modalities, hesitance trusting AI-generated predictions, insufficient data electronic health records, need validation across diverse populations, ethical regulatory concerns persist. Future efforts must focus validating ensuring algorithm transparency, addressing issues enable broader implementation. Overcoming these barriers will essential translating technology personalized care improving outcomes.

Language: Английский

---

Systematic Review of Prehospital Prediction Models for Identifying Intracerebral Haemorrhage in Suspected Stroke Patients

Healthcare, Journal Year: 2025, Volume and Issue: 13(8), P. 876 - 876

Published: April 11, 2025

Introduction: The prompt prehospital identification of intracerebral haemorrhage (ICH) may allow very early delivery treatments to limit bleeding. Current stroke assessment tools have limited accuracy for the detection ICH as they were designed recognise all strokes, not specifically. This systematic review aims evaluate performance models in distinguishing from other causes suspected stroke. Methods: We adhered Preferred Reporting Items Systematic Reviews and Meta-Analyses guidelines. Following a predefined strategy, we searched three electronic databases via Ovid (MEDLINE, EMBASE, CENTRAL) July 2023 studies published English, without date restrictions. Subsequently, data extraction was performed, methodological quality assessed using Prediction Model Risk Bias Assessment Tool. Results: After eliminating duplicates, 6194 records screened titles abstracts. full-text 137 studies, 9 prediction included. Five these described differentiate between subtypes, distinguished ischaemic stroke, one model developed specifically identify ICH. All having high risk bias, particularly analysis domain. varied, with area under receiver operating characteristic curve ranging 0.73 0.91. commonly included following predictors ICH: impaired consciousness, headache, speech or language impairment, systolic blood pressure, nausea vomiting, weakness paralysis limbs. Conclusions: support diagnosis ICH, but existing limitations, making them unreliable informing practice. Future should aim address identified limitations include broader range strokes develop practical identifying Combining point-of-care tests might further improve

Language: Английский

---

WSO Action Plan for Stroke Prehospital Care: Top Two Priorities

CNS Neuroscience & Therapeutics, Journal Year: 2025, Volume and Issue: 31(4)

Published: April 1, 2025

ABSTRACT This editorial commentary describes the consensus reached by a group of experts from World Stroke Organization regarding two top priorities to improve stroke prehospital care on global stage. The first priority is effective action awareness, and second research development point‐of‐care diagnostic technologies.

Language: Английский

---