Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Assessing Cardiomegaly in Dogs Using a Simple CNN Model

Nikhil Deekonda

TL;DR

This work tackles automated assessment of canine cardiomegaly from radiographs using VHS-based labels. It introduces DogHeart, a lightweight CNN with four convolutional and four fully connected layers trained on 2,000 images to perform three-class classification (small, normal, large) and achieves 72% test accuracy, approaching a deeper VGG16 baseline at 74.8% while offering greater efficiency. The methodology emphasizes a compact architecture, GPU-accelerated training, and evaluation on a public dataset, highlighting practical value for veterinary practice and early intervention. The findings suggest that a simple model can yield competitive performance for veterinary radiography tasks, with room for improvements through data augmentation and extended evaluation across diverse datasets.

Abstract

This paper introduces DogHeart, a dataset comprising 1400 training, 200 validation, and 400 test images categorized as small, normal, and large based on VHS score. A custom CNN model is developed, featuring a straightforward architecture with 4 convolutional layers and 4 fully connected layers. Despite the absence of data augmentation, the model achieves a 72\% accuracy in classifying cardiomegaly severity. The study contributes to automated assessment of cardiac conditions in dogs, highlighting the potential for early detection and intervention in veterinary care.

Assessing Cardiomegaly in Dogs Using a Simple CNN Model

TL;DR

This work tackles automated assessment of canine cardiomegaly from radiographs using VHS-based labels. It introduces DogHeart, a lightweight CNN with four convolutional and four fully connected layers trained on 2,000 images to perform three-class classification (small, normal, large) and achieves 72% test accuracy, approaching a deeper VGG16 baseline at 74.8% while offering greater efficiency. The methodology emphasizes a compact architecture, GPU-accelerated training, and evaluation on a public dataset, highlighting practical value for veterinary practice and early intervention. The findings suggest that a simple model can yield competitive performance for veterinary radiography tasks, with room for improvements through data augmentation and extended evaluation across diverse datasets.

Abstract

This paper introduces DogHeart, a dataset comprising 1400 training, 200 validation, and 400 test images categorized as small, normal, and large based on VHS score. A custom CNN model is developed, featuring a straightforward architecture with 4 convolutional layers and 4 fully connected layers. Despite the absence of data augmentation, the model achieves a 72\% accuracy in classifying cardiomegaly severity. The study contributes to automated assessment of cardiac conditions in dogs, highlighting the potential for early detection and intervention in veterinary care.
Paper Structure (23 sections, 2 equations, 2 figures, 1 table)

This paper contains 23 sections, 2 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methods
  4. Motivation
  5. Environment Setup
  6. Data Preprocessing
  7. Model Architecture
  8. Training and Hyperparameters
  9. Learning Rate and Optimization Algorithm
  10. Validation and Best Model Selection
  11. Test Dataset Evaluation
  12. Results
  13. Datasets
  14. Results
  15. Challenges and Limitations
  16. ...and 8 more sections

Figures (2)

  • Figure 1: Architecture of the custom CNN model.
  • Figure 2: Example images of each class