Customized GPT-4V(ision) for radiographic diagnosis: can large language model detect supernumerary teeth? DOI Creative Commons
Enes Mustafa AŞAR, İrem İpek, Kubra Bılge

и другие.

BMC Oral Health, Год журнала: 2025, Номер 25(1)

Опубликована: Май 21, 2025

With the growing capabilities of language models like ChatGPT to process text and images, this study evaluated their accuracy in detecting supernumerary teeth on periapical radiographs. A customized GPT-4V model (CGPT-4V) was also developed assess whether domain-specific training could improve diagnostic performance compared standard GPT-4o models. One hundred eighty radiographs (90 with 90 without teeth) were using GPT-4 V, GPT-4o, a fine-tuned CGPT-4V model. Each image assessed separately standardized prompt "Are there any radiograph above?" avoid contextual bias. Three dental experts scored responses three-point Likert scale for positive cases binary negatives. Chi-square tests ROC analysis used compare performances (p < 0.05). Among three models, CGPT-4 V exhibited highest accuracy, correctly 91% cases, 77% 63% GPT-4V. The demonstrated significantly lower false rate (16%) than (42%). statistically significant difference found between 0.001), while no observed or GPT-4o. Additionally, successfully identified multiple where present. These findings highlight potential GPT radiology. Future research should focus multicenter validation, seamless clinical integration, cost-effectiveness support real-world implementation.

Язык: Английский

ANALYSIS OF APPLICATIONS OF ARTIFICIAL INTELLIGENCE IN MATERIAL STORAGE: A BIBLIOGRAPHICAL STUDY DOI Creative Commons
Samara da Costa Montenegro, Jhuly de Souza Veloso, Daniel Nascimento-e-Silva

и другие.

Revista Multidisciplinar do Nordeste Mineiro, Год журнала: 2025, Номер 2(01), С. 1 - 28

Опубликована: Янв. 30, 2025

This study aimed to analyze ten publications that portray the application of artificial intelligence in material storage. The conceptual bibliographic method was used its four stages: a) formulation primary and accessory research questions, b) data collection scientific databases, c) analysis organization collected data, d) generation interpretation answers formulated questions. results showed goals studies focused on problematic situations can be considered complex, methods consisted numerous techniques procedures, tools applied were varied large quantity, conclusions show is a technology effectively solve problems help overcome storage challenges. conclusion points out more complex problem or challenge faced, greater effectiveness solving helping it. study's main contribution science highlights need for logistics professionals know how apply practice.

Язык: Английский

Процитировано

0
Customized GPT-4V(ision) for radiographic diagnosis: can large language model detect supernumerary teeth? DOI Creative Commons
Enes Mustafa AŞAR, İrem İpek, Kubra Bılge

и другие.

BMC Oral Health, Год журнала: 2025, Номер 25(1)

Опубликована: Май 21, 2025

With the growing capabilities of language models like ChatGPT to process text and images, this study evaluated their accuracy in detecting supernumerary teeth on periapical radiographs. A customized GPT-4V model (CGPT-4V) was also developed assess whether domain-specific training could improve diagnostic performance compared standard GPT-4o models. One hundred eighty radiographs (90 with 90 without teeth) were using GPT-4 V, GPT-4o, a fine-tuned CGPT-4V model. Each image assessed separately standardized prompt "Are there any radiograph above?" avoid contextual bias. Three dental experts scored responses three-point Likert scale for positive cases binary negatives. Chi-square tests ROC analysis used compare performances (p < 0.05). Among three models, CGPT-4 V exhibited highest accuracy, correctly 91% cases, 77% 63% GPT-4V. The demonstrated significantly lower false rate (16%) than (42%). statistically significant difference found between 0.001), while no observed or GPT-4o. Additionally, successfully identified multiple where present. These findings highlight potential GPT radiology. Future research should focus multicenter validation, seamless clinical integration, cost-effectiveness support real-world implementation.

Язык: Английский

Процитировано

0