Using Tree-Based Gradient Boosting to Distinguish Between Lymphoma and COVID-19

Over 600,000 new lymphoma cases and around 280,000 lymphoma-related deaths were reported in 2020. The delayed diagnosis of lymphoma has long been a problem. However, the advent of the COVID-19 pandemic, which disrupted healthcare services worldwide, may have caused more significant delays in lymphoma diagnoses. Since lymphomas can sometimes present with symptoms like COVID-19 and can affect the lungs, there is also a risk of misdiagnosis. We collected 505 lymphoma and 180 COVID-19 case reports from ScienceDirect and applied boosting methods to classify each patient as having COVID-19 or lymphoma based on the patient's age, gender and reported symptoms. LightGBM had the highest ROC AUC (0.89), meaning it best differentiates between the two diseases. Therefore, this model can be used as a screening tool to reduce the delay in lymphoma diagnosis and improve the patients' chances of survival. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

Differentiating Between COVID-19 and Tuberculosis Using Machine Learning and Natural Language Processing

Over 10 million people around the world are affected by tuberculosis (TB) every year, making it a major global health concern. With the advent of the COVID-19 pandemic, TB services in many countries have been temporarily disrupted, leading to a potential delay in the diagnosis of TB cases and many cases going under the radar. Since both diseases sometimes present similarly and generally affect the lungs, there is also a risk of misdiagnosis. This study aims to analyse the differences between COVID-19 and TB in different patients, as a first step in the creation of a TB screening tool. 180 COVID-19 and 215 TB case reports were collected from ScienceDirect. Using Natural Language Processing tools, the patient's age, gender, and symptoms were extracted from each report. Tree-based machine learning algorithms were then used to classify each case report as belonging to either disease. Overall, the cases included 252 male and 117 female patients, with 26 cases not reporting the patient's sex. The patients' ages ranged from 0 to 95 years old, with a median age of 41.5. There were 33 cases with missing age values. The most frequent symptom in the TB cases was weight loss while most COVID-19 cases listed fever as a symptom. Of all algorithms implemented, XGBoost performed best in terms of ROC AUC (86.9 %) and F1-score macro (78%). The trained model is a good starting point, which can be used by medical staff to aid in referring potential TB patients in a timely manner. This could reduce the delay in TB diagnosis as well as the TB death toll, especially in highly infected countries. © 2022 Lavoisier. All rights reserved.