Evaluating the Performance of Large Language Models in Predicting Diagnostics for Spanish Clinical Cases in Cardiology DOI Creative Commons
J. Delaunay, J. Cusidó

Applied Sciences, Год журнала: 2024, Номер 15(1), С. 61 - 61

Опубликована: Дек. 25, 2024

This study explores the potential of large language models (LLMs) in predicting medical diagnoses from Spanish-language clinical case descriptions, offering an alternative to traditional machine learning (ML) and deep (DL) techniques. Unlike ML DL models, which typically rely on extensive domain-specific training complex data preprocessing, LLMs can process unstructured text directly without need for specialized datasets. unique characteristic allows faster implementation eliminates risks associated with overfitting, are common that require tailored each new dataset. In this research, we investigate capacities several state-of-the-art based Spanish textual descriptions cases. We measured impact prompt techniques temperatures quality diagnosis. Our results indicate Gemini Pro Mixtral 8x22b generally performed well across different techniques, while Medichat Llama3 showed more variability, particularly few-shot prompting technique. Low specific such as zero-shot Retrieval-Augmented Generation (RAG), tended yield clearer accurate diagnoses. highlights a disruptive approaches, efficient, scalable, flexible solution diagnostics, non-English-speaking population.

Язык: Английский

AI versus human-generated multiple-choice questions for medical education: a cohort study in a high-stakes examination DOI Creative Commons
Alex Kwok-Keung Law,

Jerome Lok Tsun So,

Chun Tat Lui

и другие.

BMC Medical Education, Год журнала: 2025, Номер 25(1)

Опубликована: Фев. 8, 2025

Язык: Английский

Процитировано

2
Large language models for diabetes training: a prospective study DOI Creative Commons
Haoxuan Li, Zehua Jiang, Zhouyu Guan

и другие.

Science Bulletin, Год журнала: 2025, Номер unknown

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

1
Evaluating the Accuracy, Reliability, Consistency, and Readability of Different Large Language Models in Restorative Dentistry DOI Creative Commons
Zeyneb Merve Özdemir, Emre Yapici

Journal of Esthetic and Restorative Dentistry, Год журнала: 2025, Номер unknown

Опубликована: Март 2, 2025

This study aimed to evaluate the reliability, consistency, and readability of responses provided by various artificial intelligence (AI) programs questions related Restorative Dentistry. Forty-five knowledge-based information 20 (10 patient-related 10 dentistry-specific) were posed ChatGPT-3.5, ChatGPT-4, ChatGPT-4o, Chatsonic, Copilot, Gemini Advanced chatbots. The DISCERN questionnaire was used assess reliability; Flesch Reading Ease Flesch-Kincaid Grade Level scores utilized readability. Accuracy consistency determined based on chatbots' questions. Copilot demonstrated "good" while ChatGPT-3.5 showed "fair" reliability. Chatsonic exhibited highest "DISCERN total score" for questions, ChatGPT-4o performed best dentistry-specific No significant differences found in among chatbots (p > 0.05). accuracy (93.3%) had lowest (68.9%). ChatGPT-4 between repetitions. Performance AIs varied terms accuracy, when responding Dentistry promising results academic patient education applications. However, generally above recommended levels materials. utilization AI has an increasing impact aspects dentistry. Moreover, if restorative dentistry prove be reliable comprehensible, this may yield outcomes future.

Язык: Английский

Процитировано

0
Evaluating the Performance of Large Language Models in Predicting Diagnostics for Spanish Clinical Cases in Cardiology DOI Creative Commons
J. Delaunay, J. Cusidó

Applied Sciences, Год журнала: 2024, Номер 15(1), С. 61 - 61

Опубликована: Дек. 25, 2024

This study explores the potential of large language models (LLMs) in predicting medical diagnoses from Spanish-language clinical case descriptions, offering an alternative to traditional machine learning (ML) and deep (DL) techniques. Unlike ML DL models, which typically rely on extensive domain-specific training complex data preprocessing, LLMs can process unstructured text directly without need for specialized datasets. unique characteristic allows faster implementation eliminates risks associated with overfitting, are common that require tailored each new dataset. In this research, we investigate capacities several state-of-the-art based Spanish textual descriptions cases. We measured impact prompt techniques temperatures quality diagnosis. Our results indicate Gemini Pro Mixtral 8x22b generally performed well across different techniques, while Medichat Llama3 showed more variability, particularly few-shot prompting technique. Low specific such as zero-shot Retrieval-Augmented Generation (RAG), tended yield clearer accurate diagnoses. highlights a disruptive approaches, efficient, scalable, flexible solution diagnostics, non-English-speaking population.

Язык: Английский

Процитировано

1