Pre-text Representation Transfer for Deep Learning with Limited and Imbalanced Data: Application to CT-Based COVID-19 Detection

Annotating medical images for disease detection is often tedious and expensive. Moreover, the available training samples for a given task are generally scarce and imbalanced. These conditions are not conducive for learning effective deep neural models. Hence, it is common to ‘transfer' neural networks trained on natural images to the medical image domain. However, this paradigm lacks in performance due to the large domain gap between the natural and medical image data. To address that, we propose a novel concept of Pre-text Representation Transfer (PRT). In contrast to the conventional transfer learning, which fine-tunes a source model after replacing its classification layers, PRT retains the original classification layers and updates the representation layers through an unsupervised pre-text task. The task is performed with (original, not synthetic) medical images, without utilizing any annotations. This enables representation transfer with a large amount of training data. This high-fidelity representation transfer allows us to use the resulting model as a more effective feature extractor. Moreover, we can also subsequently perform the traditional transfer learning with this model. We devise a collaborative representation based classification layer for the case when we leverage the model as a feature extractor. We fuse the output of this layer with the predictions of a model induced with the traditional transfer learning performed over our pre-text transferred model. The utility of our technique for limited and imbalanced data classification problem is demonstrated with an extensive five-fold evaluation for three large-scale models, tested for five different class-imbalance ratios for CT based COVID-19 detection. Our results show a consistent gain over the conventional transfer learning with the proposed method. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

D-Dimer Elevation Matters to Predict COVID-19 Severity: A Machine Learning Approach

Although previous studies using limited data have documented an association of D-dimer levels with COVID-19 severity, the role of D-dimer in the progression of COVID-19 remains unclear and requires further investigation using data from larger cohorts. We used traditional statistical modeling and machine learning methods to examine critical factors influencing the D-dimer elevation and to characterize associated risk factors of D-dimer elevation over the course of inpatient admission. We identified 20 important features to predict D-dimer levels, some of which could be used to predict and prevent the D-dimer elevation. Laboratory monitoring of D-dimer level and its risk factors at early stage can mitigate severe or death cases in COVID-19. © 2022 IEEE.

Shallow Net for COVID-19 Classification based on Biomarkers

In many cases, especially at the beginning of epidemic disaster, it is very important to be able to determine the severity of illness of a given patient. Picking up the severe status will help in directing the effort in a proper way. At the beginning, the number of classified status and the available data are limited, so, in such situation, one needs a system that can be trained based on limited data to give a trusted result. The current work focuses on the importance of the bioscience in differentiation between recovered patients and mortalities. Even with limited data, the decision trees (DT) was able to distinguish between recovered patients and mortalities with accuracy of 94%. Shallow dense network achieved accuracy of 75%. However, when a 10-fold technique was followed with the same data, the net achieved 99% of accuracy. The used data in this work was collected from King Faisal hospital in Taif city under a formal permission from the health ministry. PCA analysis confirmed that there are two parameters that have the greatest ability to differentiate between recovered patients and mortalities. ROC curve reveals that the parameters that can differentiate between recovered patients and mortalities are calcium and hemoglobin. The shallow net gives an accuracy of 92% when trained using calcium and hemoglobin only. This paper shows that with a suitable choosing of the parameters a small decision tree or shallow net can be trained quickly to decide which patient needs more attention so as to use the hospitals resources in a more reasonable way during the pandemic. All codes and data can be accessed from the following link “codes and data” © 2022. International Journal of Advanced Computer Science and Applications.All Rights Reserved.

COVID-19 prognosis using limited chest X-ray images.

The COrona VIrus Disease 2019 (COVID-19) pandemic is an ongoing global pandemic that has claimed millions of lives till date. Detecting COVID-19 and isolating affected patients at an early stage is crucial to contain its rapid spread. Although accurate, the primary viral test 'Reverse Transcription Polymerase Chain Reaction' (RT-PCR) for COVID-19 diagnosis has an elaborate test kit, and the turnaround time is high. This has motivated the research community to develop CXR based automated COVID-19 diagnostic methodologies. However, COVID-19 being a novel disease, there is no annotated large-scale CXR dataset for this particular disease. To address the issue of limited data, we propose to exploit a large-scale CXR dataset collected in the pre-COVID era and train a deep neural network in a self-supervised fashion to extract CXR specific features. Further, we compute attention maps between the global and the local features of the backbone convolutional network while finetuning using a limited COVID-19 CXR dataset. We empirically demonstrate the effectiveness of the proposed method. We provide a thorough ablation study to understand the effect of each proposed component. Finally, we provide visualizations highlighting the critical patches instrumental to the predictive decision made by our model. These saliency maps are not only a stepping stone towards explainable AI but also aids radiologists in localizing the infected area.

Robust chest CT image segmentation of COVID-19 lung infection based on limited data.

BACKGROUND: The coronavirus disease 2019 (COVID-19) affects billions of lives around the world and has a significant impact on public healthcare. For quantitative assessment and disease monitoring medical imaging like computed tomography offers great potential as alternative to RT-PCR methods. For this reason, automated image segmentation is highly desired as clinical decision support. However, publicly available COVID-19 imaging data is limited which leads to overfitting of traditional approaches. METHODS: To address this problem, we propose an innovative automated segmentation pipeline for COVID-19 infected regions, which is able to handle small datasets by utilization as variant databases. Our method focuses on on-the-fly generation of unique and random image patches for training by performing several preprocessing methods and exploiting extensive data augmentation. For further reduction of the overfitting risk, we implemented a standard 3D U-Net architecture instead of new or computational complex neural network architectures. RESULTS: Through a k-fold cross-validation on 20 CT scans as training and validation of COVID-19, we were able to develop a highly accurate as well as robust segmentation model for lungs and COVID-19 infected regions without overfitting on limited data. We performed an in-detail analysis and discussion on the robustness of our pipeline through a sensitivity analysis based on the cross-validation and impact on model generalizability of applied preprocessing techniques. Our method achieved Dice similarity coefficients for COVID-19 infection between predicted and annotated segmentation from radiologists of 0.804 on validation and 0.661 on a separate testing set consisting of 100 patients. CONCLUSIONS: We demonstrated that the proposed method outperforms related approaches, advances the state-of-the-art for COVID-19 segmentation and improves robust medical image analysis based on limited data.