Computerized assisted evaluation system for canine cardiomegaly via key points detection with deep learning.
Mengni Zhang1, Kai Zhang1, Deying Yu2
1New Ruipeng Pet Healthcare Group Co. LTD., Beijing, 100010, China.
Preventive Veterinary Medicine
|June 12, 2021
Summary
This study presents a deep learning platform to help diagnose canine cardiomegaly using X-ray images. The system accurately identifies key points to calculate the vertebral heart score, aiding veterinarians in clinical evaluations.
Area of Science:
- Veterinary Medicine
- Medical Imaging
- Artificial Intelligence
Background:
- Cardiomegaly is a primary indicator of heart disease in dogs, often diagnosed via imaging.
- Deep learning shows significant promise for advancing diagnostic capabilities in veterinary medicine.
Purpose of the Study:
- To develop and validate a deep learning-assisted platform for diagnosing canine cardiomegaly.
- To improve the accuracy and efficiency of canine heart size evaluation using medical imaging.
Main Methods:
- Utilized HRNet (high resolution network) for detecting 16 key points on canine lateral X-ray images.
- Trained and validated the model on 2274 X-ray images, with external testing on 396 images.
- Implemented a post-processing step to refine key point detection accuracy.
Main Results:
- Achieved an initial average performance (AP) of 86.4% for key point detection.
- Enhanced AP to 90.9% after applying a post-processing correction procedure.
- Demonstrated the system's effectiveness in a clinical context for evaluating canine cardiomegaly.
Conclusions:
- The developed deep learning platform provides a clinically applicable tool for assisting in canine cardiomegaly diagnosis.
- The system's high accuracy in key point detection and VHS calculation supports its utility in veterinary practice.
- This technology has the potential to enhance diagnostic workflows for canine heart conditions.
