Data Extraction from Free-Text Reports on Mechanical Thrombectomy in Acute Ischemic Stroke Using ChatGPT: A Retrospective Analysis.

Background Procedural details of mechanical thrombectomy in patients with ischemic stroke are important predictors of clinical outcome and are collected for prospective studies or national stroke registries. To date, these data are collected manually by human readers, a labor-intensive task that is prone to errors. Purpose To evaluate the use of the large language models (LLMs) GPT-4 and GPT-3.5 to extract data from neuroradiology reports on mechanical thrombectomy in patients with ischemic stroke. Materials and Methods This retrospective study included consecutive reports from patients with ischemic stroke who underwent mechanical thrombectomy between November 2022 and September 2023 at institution 1 and between September 2016 and December 2019 at institution 2. A set of 20 reports was used to optimize the prompt, and the ability of the LLMs to extract procedural data from the reports was compared using the McNemar test. Data manually extracted by an interventional neuroradiologist served as the reference standard. Results A total of 100 internal reports from 100 patients (mean age, 74.7 years ± 13.2 [SD]; 53 female) and 30 external reports from 30 patients (mean age, 72.7 years ± 13.5; 18 male) were included. All reports were successfully processed by GPT-4 and GPT-3.5. Of 2800 data entries, 2631 (94.0% [95% CI: 93.0, 94.8]; range per category, 61%-100%) data points were correctly extracted by GPT-4 without the need for further postprocessing. With 1788 of 2800 correct data entries, GPT-3.5 produced fewer correct data entries than did GPT-4 (63.9% [95% CI: 62.0, 65.6]; range per category, 14%-99%; P < .001). For the external reports, GPT-4 extracted 760 of 840 (90.5% [95% CI: 88.3, 92.4]) correct data entries, while GPT-3.5 extracted 539 of 840 (64.2% [95% CI: 60.8, 67.4]; P < .001). Conclusion Compared with GPT-3.5, GPT-4 more frequently extracted correct procedural data from free-text reports on mechanical thrombectomy performed in patients with ischemic stroke. © RSNA, 2024 Supplemental material is available for this article.

Clinical impact of panel gene sequencing on therapy of advanced cancers of the digestive system: a retrospective, single center study.

BACKGROUND: Panel gene sequencing is an established diagnostic tool for precision oncology of solid tumors, but its utility for the treatment of cancers of the digestive system in clinical routine is less well documented. METHODS: We retrospectively identified patients with advanced or metastatic gastrointestinal, pancreaticobiliary or hepatic cancers who received panel gene sequencing at a tertiary university hospital from 2015 to 2022. For these cases, we determined the spectrum of genetic alterations, clinicopathological parameters and treatment courses. Assessment of actionability of genetic alterations was based on the OncoKB database, cancer-specific ESMO treatment guidelines, and recommendations of the local molecular tumor board. RESULTS: In total, 155 patients received panel gene sequencing using either the Oncomine Focus (62 cases), Comprehensive (91 cases) or Childhood Cancer Research Assay (2 cases). The mean age of patients was 61 years (range 24-90) and 37% were female. Most patients suffered from either colorectal cancer (53%) or cholangiocellular carcinoma (19%). 327 genetic alterations were discovered in 123 tumor samples, with an average number of 2.1 alterations per tumor. The most frequently altered genes were TP53, KRAS and PIK3CA. Actionable gene alterations were detected in 13.5-56.8% of tumors, according to ESMO guidelines or the OncoKB database, respectively. Thirteen patients were treated with targeted therapies based on identified molecular alterations, with a median progression-free survival of 8.8 months. CONCLUSIONS: Actionable genetic alterations are frequently detected by panel gene sequencing in patients with advanced cancers of the digestive tract, providing clinical benefit in selected cases. However, for the majority of identified actionable alterations, sufficient clinical evidence for targeted treatments is still lacking.

Impact of concomitant cardiovascular medications on overall survival in patients with liver cirrhosis.

OBJECTIVES OF THE ARTICLE: Liver cirrhosis is the end-stage liver disease associated with poor prognosis and cardiovascular comorbidity could significantly impact mortality of cirrhotic patients. We conducted a large, retrospective study to investigate the survival impact of cardiovascular co-medications in patients with liver cirrhosis. MATERIALS AND METHODS: A study-specific R package was processed on the local databases of partner institutions within the Observational Health Data Sciences and Informatics consortium, namely Columbia University, New York City (NYC), USA and Ajou University School of Medicine (AUSOM), South Korea. Patients with cirrhosis diagnosed between 2000 and 2020 were included. Final analysis of the anonymous survival data was performed at Medical Faculty Mannheim, Heidelberg University. RESULTS: We investigated a total of 32,366 patients with liver cirrhosis. Our data showed that administration of antiarrhythmics amiodarone or digoxin presented as a negative prognostic indicator (p = 0.000 in both cohorts). Improved survival was associated with angiotensin-converting enzyme inhibitor ramipril (p = 0.005 in NYC cohort, p = 0.075 in AUSOM cohort) and angiotensin II receptor blocker losartan (p = 0.000 in NYC cohort, p = 0.005 in AUSOM cohort). Non-selective beta blocker carvedilol was associated with a survival advantage in the NYC (p = 0.000) cohort but not in the AUSOM cohort (p = 0.142). Patients who took platelet inhibitor clopidogrel had a prolonged overall survival compared to those without (p = 0.000 in NYC cohort, p = 0.003 in AUSOM cohort). CONCLUSION: Concomitant cardiovascular medications are associated with distinct survival difference in cirrhotic patients. Multidisciplinary management is needed for a judicious choice of proper cardiovascular co-medications in cirrhotic patients.

Efficacy of Co-Medications in Patients with Alcoholic Liver Disease.

BACKGROUND: Alcoholic liver disease (ALD) is still increasing and leads to acute liver injury but also liver cirrhosis and subsequent complications such as liver failure or hepatocellular carcinoma (HCC). As most patients fail to achieve alcohol abstinence, it is essential to identify alternative treatment options in order to improve the outcome of ALD patients. METHODS: Evaluating two large cohorts of patients with ALD from the USA and Korea with a total of 12,006 patients, we investigated the effect on survival of aspirin, metformin, metoprolol, dopamine, and dobutamine drugs in patients with ALD between 2000 and 2020. Patient data were obtained through the "The Observational Health Data Sciences and Informatics consortium," an open-source, multi-stakeholder, and interdisciplinary collaborative effort. RESULTS: The use of aspirin (p = 0.000, p = 0.000), metoprolol (p = 0.002, p = 0.000), and metformin (p = 0.000, p = 0.000) confers a survival benefit for both AUSOM- and NY-treated cohorts. Need of catecholamines dobutamine (p = 0.000, p = 0.000) and dopamine (p = 0.000, p = 0.000) was strongly indicative of poor survival. ß-Blocker treatment with metoprolol (p = 0.128, p = 0.196) or carvedilol (p = 0.520, p = 0.679) was not shown to be protective in any of the female subgroups. CONCLUSION: Overall, our data fill a large gap in long-term, real-world data on patients with ALD, confirming an impact of metformin, acetylsalicylic acid, and ß-blockers on ALD patient's survival. However, gender and ethnic background lead to diverse efficacy in those patients.

Value of contrast-enhanced ultrasound for vascular diseases of the liver - current indications and open questions.

Contrast-enhanced ultrasonography (CEUS) is a widely available and well-tolerated technique that can expand the diagnosis of a variety of vascular liver diseases. This paper presents an overview of the current possibilities of the use of CEUS in vascular liver diseases. Particularly where Doppler sonography has technical limitations, CEUS provides additional opportunities to visualize vascular thrombosis and other obstructions restricting blood flow. When CT or MRI contrast agents cannot be used because of severe allergy or renal insufficiency, CEUS can be a valuable diagnostic alternative and has demonstrated comparable diagnostic performance in at least some vascular liver diseases, such as portal vein thrombosis. In addition, CEUS works without radiation and, therefore, might be particularly suitable for young patients and children. This may be useful, for example, in congenital disorders such as persistent umbilical vein or preduodenal portal vein. Vascular liver disease is rare and comprehensive data are still lacking, but the available literature provides promising insights into potential new ways to study vascular liver disease. Although most studies are based on small sample sizes or even case reports, the high diagnostic utility is undisputed.

[Budd-Chiari syndrome, review and illustration]. / Budd-Chiari Syndrom, Review und Illustration der Bildgebung.

Budd-Chiari syndrome is a rare vascular disorder characterized by obstruction of the hepatic venous outflow. Various diseases causing coagulopathy play a role in aetiology, such as myeloproliferative disorders. Acute vascular occlusion may lead to acute phlebitis with fever. The classic triad of acute liver failure may be present with ascites, hepatomegaly, and abdominal pain. However, subacute courses of disease were also observed. Because of the variable symptoms and severity extent, depending on the acuity of the course and the extent of the affected vessels, diagnosis is often difficult. Sonography, as a ubiquitously available and cost-effective diagnostic tool, plays a leading role. Doppler ultrasonography can be used to visualize hemodynamics in particular. In acute thrombotic occlusion, the affected hepatic veins usually cannot or only partially be visualized. In non-occluding thrombi, turbulent flow patterns may develop in the area of venous outflow obstruction, and flow velocity is then increased in the area of stenosis. Contrast enhanced ultrasound offers even better specificity of diagnosis. Computed tomography and magnetic resonance imaging can directly visualize thrombi and the cause of obstruction. Once the diagnosis is confirmed, anticoagulation must be initiated, but therapy of the underlying disease must also be started. If symptom-controlling measures are not sufficient, angioplasty/stenting to reopen short-segment stenoses or implantation of a TIPSS device may be considered. Liver transplantation remains ultima ratio. As studies on the precision of diagnostic methods are controversial, the characteristics of imaging for BCS are therefore summarized in this review on the basis of several illustrating case reports.

[Prevention of gastrointestinal cancer]. / Prävention gastrointestinaler Tumore.

Throughout the past decades, considerable progress has been made in the (early) diagnosis and treatment of gastrointestinal cancers. However, the prognosis for advanced stages of gastrointestinal tumors remains limited for many patients and approximately one third of all tumor patients die as a result of gastrointestinal tumors. The prevention and early detection of gastrointestinal tumors is therefore of great importance.For this reason, we summarize the current state of knowledge and recommendations for the primary, secondary and tertiary prevention of esophageal, stomach, pancreas, liver and colorectal cancer in the following.

A future role for health applications of large language models depends on regulators enforcing safety standards.

Among the rapid integration of artificial intelligence in clinical settings, large language models (LLMs), such as Generative Pre-trained Transformer-4, have emerged as multifaceted tools that have potential for health-care delivery, diagnosis, and patient care. However, deployment of LLMs raises substantial regulatory and safety concerns. Due to their high output variability, poor inherent explainability, and the risk of so-called AI hallucinations, LLM-based health-care applications that serve a medical purpose face regulatory challenges for approval as medical devices under US and EU laws, including the recently passed EU Artificial Intelligence Act. Despite unaddressed risks for patients, including misdiagnosis and unverified medical advice, such applications are available on the market. The regulatory ambiguity surrounding these tools creates an urgent need for frameworks that accommodate their unique capabilities and limitations. Alongside the development of these frameworks, existing regulations should be enforced. If regulators fear enforcing the regulations in a market dominated by supply or development by large technology companies, the consequences of layperson harm will force belated action, damaging the potentiality of LLM-based applications for layperson medical advice.

Usage Behaviour and Adoption Criteria for Mobile Health Solutions in Patients with Chronic Diseases in Gastroenterology.

Introduction: Mobile Health (mHealth) applications allow for new possibilities and opportunities in patient care. Their potential throughout the whole patient journey is undisputed. However, the eventual adoption by patients depends on their acceptance of and motivation to use mHealth applications as well as their adherence. Therefore, we investigated the motivation and drivers of acceptance for mHealth and developed an adapted model of the Unified Theory of Acceptance and Use of Technology (UTAUT2). Methods: We evaluated 215 patients with chronic gastroenterological diseases who answered a questionnaire including all model constructs with 7-point Likert scale items. Our model was adapted from the Unified Theory of Acceptance and Use in Technology 2 and includes influencing factors such as facilitating conditions, performance expectancy, hedonic motivation, social influence factors, effort expectancy, as well as personal empowerment and data protection concerns. Model evaluation was performed with structural equation modelling with PLS-SEM. Bootstrapping was performed for hypothesis testing. Results and Conclusion: Patients had a median age of 55.5 years, and the gender ratio was equally distributed. Forty percent received a degree from a university, college, technical academy, or engineering school. The majority of patients suffered from chronic liver disease, but patients with inflammatory bowel diseases, GI cancers, and pancreatic diseases were also included. Patients considered their general technology knowledge as medium to good or very good (78%). Actual usage of mHealth applications in general was rare, while the intention to use them was high. The leading acceptance factor for mHealth applications in our patient group was feasibility, both in terms of technical requirements and the intuitiveness and manageability of the application. Concerns about data privacy did not significantly impact the intention to use mobile devices. Neither the gamification aspect nor social influence factors played a significant role in the intention to use mHealth applications. Interpretation: Most of our patients were willing to spend time on a mHealth application specific to their disease on a regular basis. Acceptance and adherence are ensured by efficient utilization that requires minimum effort and compatible technologies as well as support in case of difficulties. Social influence and hedonic motivation, which were part of UTAUT2, as well as data security concerns, were not significantly influencing our patients' intention to use mHealth applications. A literature review revealed that drivers of acceptance vary considerably among different population and patient groups. Therefore, healthcare and mHealth providers should put effort into understanding their specific target groups' drivers of acceptance. We provided those for a cohort of patients from gastroenterology in this project.

Privacy-preserving large language models for structured medical information retrieval.

Most clinical information is encoded as free text, not accessible for quantitative analysis. This study presents an open-source pipeline using the local large language model (LLM) "Llama 2" to extract quantitative information from clinical text and evaluates its performance in identifying features of decompensated liver cirrhosis. The LLM identified five key clinical features in a zero- and one-shot manner from 500 patient medical histories in the MIMIC IV dataset. We compared LLMs of three sizes and various prompt engineering approaches, with predictions compared against ground truth from three blinded medical experts. Our pipeline achieved high accuracy, detecting liver cirrhosis with 100% sensitivity and 96% specificity. High sensitivities and specificities were also yielded for detecting ascites (95%, 95%), confusion (76%, 94%), abdominal pain (84%, 97%), and shortness of breath (87%, 97%) using the 70 billion parameter model, which outperformed smaller versions. Our study successfully demonstrates the capability of locally deployed LLMs to extract clinical information from free text with low hardware requirements.

LLM-AIx: An open source pipeline for Information Extraction from unstructured medical text based on privacy preserving Large Language Models.

In clinical science and practice, text data, such as clinical letters or procedure reports, is stored in an unstructured way. This type of data is not a quantifiable resource for any kind of quantitative investigations and any manual review or structured information retrieval is time-consuming and costly. The capabilities of Large Language Models (LLMs) mark a paradigm shift in natural language processing and offer new possibilities for structured Information Extraction (IE) from medical free text. This protocol describes a workflow for LLM based information extraction (LLM-AIx), enabling extraction of predefined entities from unstructured text using privacy preserving LLMs. By converting unstructured clinical text into structured data, LLM-AIx addresses a critical barrier in clinical research and practice, where the efficient extraction of information is essential for improving clinical decision-making, enhancing patient outcomes, and facilitating large-scale data analysis. The protocol consists of four main processing steps: 1) Problem definition and data preparation, 2) data preprocessing, 3) LLM-based IE and 4) output evaluation. LLM-AIx allows integration on local hospital hardware without the need of transferring any patient data to external servers. As example tasks, we applied LLM-AIx for the anonymization of fictitious clinical letters from patients with pulmonary embolism. Additionally, we extracted symptoms and laterality of the pulmonary embolism of these fictitious letters. We demonstrate troubleshooting for potential problems within the pipeline with an IE on a real-world dataset, 100 pathology reports from the Cancer Genome Atlas Program (TCGA), for TNM stage extraction. LLM-AIx can be executed without any programming knowledge via an easy-to-use interface and in no more than a few minutes or hours, depending on the LLM model selected.

Multivariate Platelet Analysis Differentiates Between Patients with Alzheimer's Disease and Healthy Controls at First Clinical Diagnosis.

BACKGROUND: Early diagnosis of Alzheimer's disease (AD) is challenging, and easily accessible biomarkers are an unmet need. Blood platelets frequently serve as peripheral model for studying AD pathogenesis and might represent a reasonable biomarker source. OBJECTIVE: In the present study, we investigated the potential to differentiate AD patients from healthy controls (HC) based on blood count, platelet morphology, and function as well as molecular markers at the time of first clinical diagnosis. METHODS: Blood samples from 40 AD patients and 29 age-matched HC were included for determination of 78 parameter by blood counting, platelet morphometry, aggregometry, flow cytometry (CD62P, CD63, activated fibrinogen receptor), protein quantification of nicotinic acetylcholine receptor α7 (nAChRα7) and caveolin-1 (CAV-1), and miRNA quantification (miR-26b, miR-199a, miR-335). Group comparison between patients and controls was performed in univariate and multivariate statistical analyses. RESULTS: AD patients showed significantly lower aggregation response to ADP and arachidonic acid and significantly decreased CD62P and CD63 surface expression induced by ADP and U46619 compared to HC. Relative nAChRα7 and CAV-1 expression was significantly higher AD platelets than in HC. Multivariate analysis of 63 parameter revealed significant differences between AD patients and healthy controls. The best performing feature model revealed a sensitivity of 96.6%, a specificity of 80.0%, and a positive predictive value of 89.3%. No grouping could be achieved by using single parameter groups. CONCLUSION: Significant differences between platelet characteristics from AD patients and HC at the time of first clinical diagnosis were observed. The best performing parameter can be used as a blood-based biomarker for AD diagnosis in a multivariate model in addition to the standardized mental tests.