Parallel training improves covid diagnosis model speed and accuracy
Parallel Training Using a CNN-DNN Architecture for Accelerated Development of Diagnostic Models
Computer Vision and Pattern Recognition
Summary
Training deep learning models to detect diseases in medical images can take a long time, especially with large and growing datasets like during a pandemic. The authors collected CT scans from COVID-19 and other pneumonia patients and used a combined CNN-DNN approach that breaks images into parts to speed up training by running in parallel. This method achieved similar or better accuracy while reducing training time significantly. Their approach helps create diagnostic tools faster, which can be important in urgent health situations.
What this means in practice
- •For hospital data teams: Accelerate training of diagnostic models for COVID-19 from CT scans to enable faster deployment during outbreaks.
- •For medical imaging companies: Build faster diagnostic AI tools by using parallel CNN-DNN architectures that handle large imaging datasets efficiently.$Commercial implications: Enables development of commercial diagnostic software optimized for speed and accuracy using large medical image data.
Tested on one dataset.
Authors
Janine Weber-Hamacher, Astha Jaiswal, Philipp Fervers, Dorotya Móré, Athanasios Giannakis, Ricarda Fischbach, Andreas Michael Bucher, Rahil Shahzad, Jonathan Kottlors, Thorsten Persigehl, Axel Klawonn
Abstract
Artificial intelligence has shown promise in assisting radiologists in imaging-based diagnosis across a wide range of diseases. Efficient training of large deep learning models is essential to cope with extremely large data sets or dynamically growing disease data, like in a pandemic like situation. In this retrospective study, we collected 300 CT scans from COVID-19 and non-COVID-19 pneumonia patients from three different centers in Germany. We investigated a hybrid CNN-DNN network model based on image decomposition and localization that naturally supports parallel and efficient training of deep learning models. In total, 156 models with three different architectures were trained to capture features at different levels resulting in 12 patient-level COVID-19 diagnosis models. Diagnostic performance as well as time saving were measured. The highest accuracy was obtained from DenseNet121 and 3D CNN models with a parallel CNN-DNN approach, resulting in $88.78\%$ training, $76.67\%$ validation and $76.03\%$ test accuracy for the DenseNet121 with $4\times4\times1$ subdomains and $87.72\%$ training, $76.82\%$ validation and $74.86\%$ test accuracy, respectively, for the 3D CNN with $4\times4\times1$ subdomains. The strongest reduction in parallel training time by a factor of $31$ was observed for the 3D CNN model and $4\times4\times2$ subdomains. Our parallel training approach improves efficiency as well as performance enabling rapid model development, among others crucial for pandemic preparedness.