EyeDiff: text-to-image diffusion model improves rare eye disease diagnosis DOI Creative Commons
Ruoyu Chen, Weiyi Zhang, Bowen Liu

и другие.

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Ноя. 27, 2024

Abstract The rising prevalence of vision-threatening retinal diseases poses a significant burden on the global healthcare systems. Deep learning (DL) offers promising solution for automatic disease screening but demands substantial data. Collecting and labeling large volumes ophthalmic images across various modalities encounters several real-world challenges, especially rare diseases. Here, we introduce EyeDiff, text-to-image model designed to generate multimodal from natural language prompts evaluate its applicability in diagnosing common EyeDiff is trained eight large-scale datasets using advanced latent diffusion model, covering 14 image over 80 ocular diseases, adapted ten multi-country external datasets. generated accurately capture essential lesional characteristics, achieving high alignment with text as evaluated by objective metrics human experts. Furthermore, integrating significantly enhances accuracy detecting minority classes eye surpassing traditional oversampling methods addressing data imbalance. effectively tackles issue imbalance insufficiency typically encountered addresses challenges collecting annotated images, offering transformative enhance development expert-level diagnosis models field.

Язык: Английский

Generative artificial intelligence in graduate medical education DOI Creative Commons

Ravi Janumpally,

Suparna Nanua,

Andy Ngo

и другие.

Frontiers in Medicine, Год журнала: 2025, Номер 11

Опубликована: Янв. 10, 2025

Generative artificial intelligence (GenAI) is rapidly transforming various sectors, including healthcare and education. This paper explores the potential opportunities risks of GenAI in graduate medical education (GME). We review existing literature provide commentary on how could impact GME, five key areas opportunity: electronic health record (EHR) workload reduction, clinical simulation, individualized education, research analytics support, decision support. then discuss significant risks, inaccuracy overreliance AI-generated content, challenges to authenticity academic integrity, biases AI outputs, privacy concerns. As technology matures, it will likely come have an important role future but its integration should be guided by a thorough understanding both benefits limitations.

Язык: Английский

Процитировано

2
Künstliche Intelligenz in der Medizin – Chancen und Risiken aus ethischer Sicht DOI

Saskia Metan,

Florian Bruns

Deleted Journal, Год журнала: 2025, Номер unknown

Опубликована: Апрель 2, 2025

Imaging disciplines, such as ophthalmology, offer a wide range of opportunities for the beneficial use artificial intelligence (AI). The analysis images and data by trained algorithms has potential to facilitate making diagnosis patient care not just in ophthalmology. If AI brings about advances clinical practice that benefit patients, this is ethically be welcomed; however, respect self-determination patients security must guaranteed. Traceability explainability would strengthen trust automated decision-making enable ultimate medical responsibility. It should noted are only good unbiased used train them. likely lead loss skills on part doctors (deskilling), counteracted, example through improved training. Accompanying ethics research necessary identify those aspects require regulation. In principle, taken ensure serves people adapts their needs, other way round.

Язык: Английский

Процитировано

1
Blinking characteristics analyzed by a deep learning model and the relationship with tear film stability in children with long-term use of orthokeratology DOI Creative Commons
Wu Yue, Siyuan Wu, Yao Yu

и другие.

Frontiers in Cell and Developmental Biology, Год журнала: 2025, Номер 12

Опубликована: Янв. 28, 2025

Purpose Using deep learning model to observe the blinking characteristics and evaluate changes their correlation with tear film in children long-term use of orthokeratology (ortho-K). Methods 31 (58 eyes) who had used ortho-K for more than 1 year age gender-matched controls were selected follow-up our ophthalmology clinic from 2021/09 2023/10 this retrospective case-control study. Both groups underwent comprehensive ophthalmological examinations, including Ocular Surface Disease Index (OSDI) scoring, Keratograph 5M, LipiView. A system based on U-Net Swim-Transformer was proposed observation characteristics. The frequency incomplete blinks (IB), complete (CB) rate (IBR) within 20 s, as well duration closing, closed, opening phases blink wave calculated by system. Relative IPH% defined ratio mean s maximum value indicate extent blinking. Furthermore, accuracy, precision, sensitivity, specificity, F1 score overall U-Net-Swin-Transformer model, its consistency built-in algorithm evaluated well. Independent t-test Mann-Whitney test analyze patterns between wearer group control group. Spearman’s rank relationship stability. Results Our demonstrated high performance (accuracy = 98.13%, precision 96.46%, sensitivity 98.10%, specificity 0.9727) patterns. OSDI scores, conjunctival redness, lipid layer thickness (LLT), meniscus height did not change significantly two groups. Notably, exhibited shorter first (11.75 ± 7.42 vs. 14.87 7.93 p 0.030) average non-invasive break-up times (NIBUT) (13.67 7.0 16.60 7.24 0.029) compared They a higher IB (4.26 2.98 2.36 2.55, < 0.001), IBR (0.81 0.28 0.46 0.39, relative (0.3229 0.1539 0.2233 0.1960, 0.004) prolonged eye-closing phase (0.18 0.08 0.15 0.07 0.032) (0.35 0.12 0.14 0.015) controls. In addition, analysis revealed negative NIBUT (for first-NIBUT, r −0.292, 0.004; avg-NIBUT, −0.3512, 0.001) ortho-K. Conclusion U-net achieved optimal Children presented an increase eye closing phase. increased associated decreased stability, indicating importance monitoring children’s status clinical follow-up.

Язык: Английский

Процитировано

0
AI Image Generation Technology in Ophthalmology: Use, Misuse and Future Applications DOI Creative Commons

Benjamin Phipps,

Xavier Hadoux, Bin Sheng

и другие.

Progress in Retinal and Eye Research, Год журнала: 2025, Номер unknown, С. 101353 - 101353

Опубликована: Март 1, 2025

Язык: Английский

Процитировано

0
Embracing generative AI in ophthalmology DOI
Frank Larkin, Mingguang He

British Journal of Ophthalmology, Год журнала: 2024, Номер 108(10), С. 1333 - 1334

Опубликована: Сен. 20, 2024

Язык: Английский

Процитировано

0
Assessing AI chatbots efficacy in ophthalmic triage and referrals: A comparative study DOI
Piyush Jain,

Sarita Panda,

Ankita Mishra

и другие.

IP International Journal of Ocular Oncology and Oculoplasty, Год журнала: 2024, Номер 10(3), С. 135 - 139

Опубликована: Ноя. 2, 2024

To evaluate the efficacy of AI chatbots (OpenAI ChatGPT GPT-3.5, Google Bard, and WebMD) compared to human ophthalmology trainees in triage referrals for common ophthalmic conditions. A single-center study was conducted at MKCG Medical College, Berhampur, Odisha, involving six trainees. The performance assessed based on diagnostic accuracy categorization. Key indicators included top three suggested diagnoses concordance recommendations investigations referrals. Physician respondents identified correct diagnosis among suggestions 95% cases. Bard achieved 90% accuracy, 85%, WebMD 20%. High observed between physician demonstrate promising potential supporting referral decisions While expertise remains crucial, tools can augment improve efficiency, enhance patient care. Future research should focus refining algorithms, integrating clinical data, exploring real-world implementation strategies.

Язык: Английский

Процитировано

0
EyeDiff: text-to-image diffusion model improves rare eye disease diagnosis DOI Creative Commons
Ruoyu Chen, Weiyi Zhang, Bowen Liu

и другие.

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Ноя. 27, 2024

Abstract The rising prevalence of vision-threatening retinal diseases poses a significant burden on the global healthcare systems. Deep learning (DL) offers promising solution for automatic disease screening but demands substantial data. Collecting and labeling large volumes ophthalmic images across various modalities encounters several real-world challenges, especially rare diseases. Here, we introduce EyeDiff, text-to-image model designed to generate multimodal from natural language prompts evaluate its applicability in diagnosing common EyeDiff is trained eight large-scale datasets using advanced latent diffusion model, covering 14 image over 80 ocular diseases, adapted ten multi-country external datasets. generated accurately capture essential lesional characteristics, achieving high alignment with text as evaluated by objective metrics human experts. Furthermore, integrating significantly enhances accuracy detecting minority classes eye surpassing traditional oversampling methods addressing data imbalance. effectively tackles issue imbalance insufficiency typically encountered addresses challenges collecting annotated images, offering transformative enhance development expert-level diagnosis models field.

Язык: Английский

Процитировано

0