Summary of Development Of a Large-scale Dataset Of Chest Computed Tomography Reports in Japanese and a High-performance Finding Classification Model, by Yosuke Yamagishi et al.
Development of a Large-scale Dataset of Chest Computed Tomography Reports in Japanese and a High-performance Finding Classification Model
by Yosuke Yamagishi, Yuta Nakamura, Tomohiro Kikuchi, Yuki Sonoda, Hiroshi Hirakawa, Shintaro Kano, Satoshi Nakamura, Shouhei Hanaoka, Takeharu Yoshikawa, Osamu Abe
First submitted to arxiv on: 20 Dec 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A recent surge in advancements in large language models underscores the pressing need for high-quality multilingual medical datasets. Specifically, Japan’s dominance in CT scanner deployment and utilization has been hindered by the lack of large-scale Japanese radiology datasets, thereby stunting the development of specialized language models for medical imaging analysis. To address this gap, researchers aimed to develop a comprehensive Japanese CT report dataset through machine translation and establish a tailored language model for structured finding classification. Additionally, they sought to create a rigorously validated evaluation dataset via expert radiologist review. A translated dataset was generated using GPT-4o mini, with the training dataset consisting of 22,778 machine-translated reports and the validation dataset comprising 150 radiologist-revised reports. The team developed CT-BERT-JPN based on the “tohoku-nlp/bert-base-japanese-v3” architecture to extract 18 structured findings from Japanese radiology reports. Results indicate strong translation performance with BLEU scores of 0.731 and 0.690, as well as ROUGE scores ranging from 0.770 to 0.876 for the Findings section and from 0.748 to 0.857 for the Impression section. CT-BERT-JPN demonstrated superior performance compared to GPT-4o in 11 out of 18 conditions, including lymphadenopathy (+14.2%), interlobular septal thickening (+10.9%), and atelectasis (+7.4%). The model maintained F1 scores exceeding 0.95 in 14 out of 18 conditions and achieved perfect scores in four conditions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study aims to create a large-scale Japanese CT report dataset and develop a specialized language model for medical imaging analysis. Researchers used machine translation to translate the CT-RATE dataset into Japanese, then developed a language model called CT-BERT-JPN to analyze the reports. The results show that the translated reports are very similar to the original ones, and the language model is good at finding certain things in the reports. |
Keywords
» Artificial intelligence » Bert » Bleu » Classification » Gpt » Language model » Nlp » Rouge » Translation
