[Paper Review] COVID-ResNet: A Deep Learning Framework for Screening of COVID19 from Radiographs

Muhammad Ali Farooq, Abdul Hafeez|arXiv (Cornell University)|Mar 31, 2020

COVID-19 diagnosis using AI13 references438 citations

TL;DR

The paper presents COVIDResNet, an open-source 3-step fine-tuning of pre-trained ResNet-50 to classify COVID-19, other pneumonia, and normal chest radiographs, achieving state-of-the-art accuracy on the COVIDx dataset using progressive resizing and discriminative learning rates.

ABSTRACT

In the last few months, the novel COVID19 pandemic has spread all over the world. Due to its easy transmission, developing techniques to accurately and easily identify the presence of COVID19 and distinguish it from other forms of flu and pneumonia is crucial. Recent research has shown that the chest Xrays of patients suffering from COVID19 depicts certain abnormalities in the radiography. However, those approaches are closed source and not made available to the research community for re-producibility and gaining deeper insight. The goal of this work is to build open source and open access datasets and present an accurate Convolutional Neural Network framework for differentiating COVID19 cases from other pneumonia cases. Our work utilizes state of the art training techniques including progressive resizing, cyclical learning rate finding and discriminative learning rates to training fast and accurate residual neural networks. Using these techniques, we showed the state of the art results on the open-access COVID-19 dataset. This work presents a 3-step technique to fine-tune a pre-trained ResNet-50 architecture to improve model performance and reduce training time. We call it COVIDResNet. This is achieved through progressively re-sizing of input images to 128x128x3, 224x224x3, and 229x229x3 pixels and fine-tuning the network at each stage. This approach along with the automatic learning rate selection enabled us to achieve the state of the art accuracy of 96.23% (on all the classes) on the COVIDx dataset with only 41 epochs. This work presented a computationally efficient and highly accurate model for multi-class classification of three different infection types from along with Normal individuals. This model can help in the early screening of COVID19 cases and help reduce the burden on healthcare systems.

Motivation & Objective

• Motivate the need for accurate, accessible screening of COVID-19 from chest radiographs.
• Provide an open-source, reproducible dataset and model for differentiating COVID-19 from other pneumonia and normal cases.
• Develop a 3-step fine-tuning strategy to improve performance and reduce training time.

Proposed method

• Use a pre-trained ResNet-50 backbone and fine-tune it with a 3-step progressive resizing of inputs to 128x128x3, 224x224x3, and 229x229x3.
• Apply progressive resizing combined with automatic learning rate selection to optimize training.
• Implement discriminative learning rates to tailor layer-wise updates during fine-tuning.
• Leverage open-access COVIDx dataset for multi-class classification of three infection types plus normal cases.
• Aim for computational efficiency and high accuracy with limited epochs.

Experimental results

Research questions

• RQ1 Can a ResNet-50 based architecture distinguish COVID-19 from other pneumonia and normal cases in chest radiographs?
• RQ2 Does progressive resizing with discriminative learning rates improve training efficiency and accuracy on the COVIDx dataset?
• RQ3 What is the achievable accuracy and training efficiency when fine-tuning a pre-trained network for multi-class COVID-19 screening from radiographs?

Key findings

• Achieves state-of-the-art accuracy of 96.23% on the COVIDx dataset across all classes.
• Employs 41 training epochs to reach the reported performance.
• Demonstrates a computationally efficient approach for multi-class classification of infection types plus normal cases.
• Provides a 3-step fine-tuning scheme that progressively resizes inputs and tunes the network at each stage.
• Highlights open-source and open-access data and model for reproducibility and community use.