Adaptive Feature Selection Guided Deep Forest for COVID-19 Classification with Chest CT

Abstract: Chest computed tomography (CT) becomes an effective tool to assist the diagnosis of coronavirus disease-19 (COVID-19). Due to the outbreak of COVID-19 worldwide, using the computed-aided diagnosis technique for COVID-19 classification based on CT images could largely alleviate the burden of clinicians. In this paper, we propose an Adaptive Feature Selection guided Deep Forest (AFS-DF) for COVID-19 classification based on chest CT images. Specifically, we first extract location-specific features from CT images. Then, in order to capture the high-level representation of these features with the relatively small-scale data, we leverage a deep forest model to learn high-level representation of the features. Moreover, we propose a feature selection method based on the trained deep forest model to reduce the redundancy of features, where the feature selection could be adaptively incorporated with the COVID-19 classification model. We evaluated our proposed AFS-DF on COVID-19 dataset with 1495 patients of COVID-19 and 1027 patients of community acquired pneumonia (CAP). The accuracy (ACC), sensitivity (SEN), specificity (SPE) and AUC achieved by our method are 91.79%, 93.05%, 89.95% and 96.35%, respectively. Experimental results on the COVID-19 dataset suggest that the proposed AFS-DF achieves superior performance in COVID-19 vs. CAP classification, compared with 4 widely used machine learning methods.

• The paper introduces an adaptive deep forest model that integrates feature selection to improve the classification of COVID-19 versus CAP.
• It extracts location-specific radiomic features from chest CT images and achieves high performance with 91.79% accuracy and a 96.35% AUC.
• The study highlights the model's potential to automate CT scan interpretations, offering clinical benefits by reducing diagnostic workload.

Adaptive Feature Selection Guided Deep Forest for COVID-19 Classification Using Chest CT Images

The paper "Adaptive Feature Selection Guided Deep Forest for COVID-19 Classification with Chest CT," authored by Liang Sun et al., introduces an innovative approach for classifying COVID-19 and community acquired pneumonia (CAP) using chest CT images. The COVID-19 pandemic has prompted a critical demand for accurate and timely diagnostic tools, and this paper contributes to that need by proposing an Adaptive Feature Selection Guided Deep Forest (AFS-DF) model.

Summary of Methodology

The AFS-DF model combines adaptive feature selection with deep forest learning to enhance the classification of COVID-19 versus CAP. Initially, the researchers extract location-specific features from chest CT images, including volume, infected lesion number, histogram distribution, surface area, and radiomics features. These features are intended to capture intricate patterns in CT images, relevant to diagnosing lung infections.

The deep forest algorithm leverages these features to learn higher-level representations while circumventing the extensive parameter tuning typically associated with deep learning. Furthermore, this approach incorporates an adaptive feature selection mechanism. The feature selection is dynamically integrated into the model training, discarding redundant features based on calculated importance, thus enhancing the model's discrimination capability.

Experimental Findings

The paper reports strong performance metrics using a dataset comprising 1495 COVID-19 patients and 1027 CAP patients, sourced from multiple hospitals. The AFS-DF model achieves an accuracy of 91.79%, a sensitivity of 93.05%, and a specificity of 89.95%. With an AUC of 96.35%, the model outperforms several conventional machine learning methods such as Logistic Regression, SVM, Random Forests, and Neural Networks. This superior performance underlines the effectiveness of feature representation and selection strategies used in AFS-DF.

Implications and Future Directions

The research presented in this paper holds both practical and theoretical implications. Practically, the deployment of such a model in clinical settings could alleviate the burden on clinicians by automating the classification of chest CT scans with high precision. Theoretically, the integration of adaptive feature selection within a deep forest framework represents a significant advance in machine learning, suggesting potential applications beyond COVID-19 diagnostics.

As machine intelligence continues to evolve, the adaptive feature selection guided methodologies could be further refined to tackle other complex classification tasks, including multi-class problems in medical imaging. Additionally, more expansive datasets and complementary modalities could be explored to further validate and amplify the efficacy of the AFS-DF model.

In conclusion, the paper contributes substantively to the ongoing effort to enhance automated diagnostic tools for respiratory diseases, demonstrating the potential of combining adaptive feature selection with deep learning frameworks. The model provides a promising approach for robust and efficient COVID-19 classification, offering insights into future applications in AI-driven healthcare solutions.