Diagnosing Heart Failure from Chest X-Ray Images Using Deep Learning
Problem
Diagnosing heart failure from chest X-ray images is difficult and rarely automated. The study tested whether deep learning could diagnose heart failure from chest X-rays.
Why it matters
Faster, accurate heart failure detection from routine X-rays could improve timely treatment for a deadly condition.
What the AI did
A deep learning algorithm using data augmentation and transfer learning was trained on cardiologist-relabeled NIH chest X-ray images to classify heart failure, with heatmap visualization of decisions.
Results
82% accuracy diagnosing heart failure on 638 relabeled images (260 normal, 378 heart failure) from 952 NIH images.
Impact
Single-study demonstration of 82% diagnostic accuracy; no clinical deployment stated.
Record
| Venue | International Heart Journal |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Takuya Matsumoto, Satoshi Kodera, Hiroki Shinohara, Hirotaka Ieki, Toshihiro Yamaguchi, Yasutomi Higashikuni, Arihiro Kiyosue, Kaoru Ito et al. |
| Citations | 66 (OpenAlex, at import time) |
| References | 11 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1536/ihj.19-714 |
| Open access | diamond |
| Keywords | deep learning, chest X-ray, heart failure, diagnosis, transfer learning |
Importance rating
real but narrower advance
Heart failure diagnosis support from routine X-rays is clinically useful, but abstract gives no scale, prevalence, or mortality figures; single 638-image study, no deployment.