You’ve Got Two Teachers: Co-evolutionary Image and Report Distillation for Semi-supervised Anatomical Abnormality Detection in Chest X-Ray

Sun, Jinghan; Wei, Dong; Xu, Zhe; Lu, Donghuan; Liu, Hong; Wang, Liansheng; Zheng, Yefeng

doi:10.1007/978-3-031-43907-0_35

Cited by 2 publications

(3 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This disparity could stem from a noticeable dearth of datasets specifically tailored for medical visual grounding. The MS-CXR dataset, a recent published visual grounding dataset , makes some improvement of medical image visual grounding application, some publications Sun et al, 2023a;b) comes out base on it. Nevertheless, even as this dataset becomes more widely recognized, there remains a limited body of academic work exploring its potential and applications, highlighting a crucial area for future research and development.…”

Section: Visual Groundingmentioning

confidence: 99%

“…Consequently, there is a growing demand for automated radiology report generation, which is anticipated to alleviate the workload of radiologists and mitigate diagnostic inaccuracies. The rapid advancements in artificial intelligence, particularly in the domains of computer vision and natural language processing, have made automated medical report generation a feasible reality (e.g., Liu et al, 2021;Wang et al, 2023a). A prominent challenge in automated medical report generation is long text generation.…”

Section: Radiology Report Generationmentioning

confidence: 99%

“…The methods encompasses standard image captioning techniques, including Show-Tell (Vinyals et al, 2015), Att2in (Xu et al, 2015), AdaAtt (Lu et al, 2017), Transformer (Vaswani et al, 2017), and M2Transformer (Cornia et al, 2020). Additionally, the evaluation methods also have medical report generation methods, specifically R2Gen , R2GenCMN Chen et al (2021), MSAT (Wang et al, 2022b), and METransformer (Wang et al, 2023a). To provide fair comparisons, we employ the exact same prompting structure (Few-shot mixed examples prompt) to help GPT-4V generate the medical report.…”

Section: Comparison With Sotamentioning

confidence: 99%

See 2 more Smart Citations

A Comprehensive Study of GPT-4V’s Multimodal Capabilities in Medical Imaging

Li,

Liu,

Wang

et al. 2023

Preprint

View full text Add to dashboard Cite

This paper presents a comprehensive evaluation of GPT-4V’s capabilities across diverse medical imaging tasks, including Radiology Report Generation, Medical Visual Question Answering (VQA), and Visual Grounding. While prior efforts have explored GPT-4V’s performance in medical imaging, to the best of our knowledge, our study represents the first quantitative evaluation on publicly available benchmarks. Our findings highlight GPT-4V’s potential in generating descriptive reports for chest X-ray images, particularly when guided by well-structured prompts. However, its performance on the MIMIC-CXR dataset benchmark reveals areas for improvement in certain evaluation metrics, such as CIDEr. In the domain of Medical VQA, GPT-4V demonstrates proficiency in distinguishing between question types but falls short of prevailing benchmarks in terms of accuracy. Furthermore, our analysis finds the limitations of conventional evaluation metrics like the BLEU score, advocating for the development of more semantically robust assessment methods. In the field of Visual Grounding, GPT-4V exhibits preliminary promise in recognizing bounding boxes, but its precision is lacking, especially in identifying specific medical organs and signs. Our evaluation underscores the significant potential of GPT-4V in the medical imaging domain, while also emphasizing the need for targeted refinements to fully unlock its capabilities.

show abstract

Section: Visual Groundingmentioning

confidence: 99%

Section: Radiology Report Generationmentioning

confidence: 99%