Validation of a Deep Learning Chest X-ray Interpretation Model: Integrating Large-Scale AI and Large Language Models for Comparative Analysis with ChatGPT
Problem
Chest X-ray interpretation by AI needed validation; the study compared an assistive CXR reading model against ChatGPT for diagnostic accuracy and clinical utility.
Why it matters
Accurate chest X-ray reading directly affects diagnosis quality and patient care in medical imaging.
What the AI did
KARA-CXR and ChatGPT readings of 2000 randomly selected chest X-rays were evaluated by two radiologists on five qualitative factors including accuracy, false findings, and hallucinations.
Results
KARA-CXR achieved significantly higher diagnostic accuracy than ChatGPT on 2000 chest X-rays across five qualitative factors.
Impact
Peer-reviewed validation showed KARA-CXR outperformed ChatGPT for chest X-ray reading at one institution.
Record
| Venue | Diagnostics |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Kyu Hong Lee, Ro Woon Lee, Ye Eun Kwon |
| Citations | 29 (OpenAlex, at import time) |
| References | 28 |
| Model or method | KARA-CXR; ChatGPT |
| Organization | Kakao Brain |
| Primary source | https://doi.org/10.3390/diagnostics14010090 |
| Open access | gold |
| Keywords | chest X-ray, deep learning, ChatGPT, KARA-CXR, diagnostic accuracy, radiology |
Importance rating
real but narrower advance
Accurate chest X-ray reading affects diagnosis quality and patient care; validation at a single institution on 2000 images comparing an assistive model to ChatGPT — real but institution-limited advance.