Frontiers in medicine · 2025
Machine learning of immune-related laboratory indicators enables discrimination between pulmonary and extrapulmonary tuberculosis.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Gu Shuanglin, Guo Yichen, Jiang Han, Yang Chen, Li Tengyue, Ma Xiuchang, Qian Wenxian, Liang Zihao, Hu Zheng, Chen Apeng, Zeng Yi, Yi Changhua
Аннотация
INTRODUCTION: As a globally significant infectious disease, the tuberculosis bacterium can not only cause pulmonary tuberculosis through respiratory tract infection but also spread from pulmonary lesions to organs throughout the body via the blood or lymphatic system, resulting in extrapulmonary tuberculosis (EPTB). However, achieving early and accurate diagnosis of EPTB remains a major clinical challenge.
METHODS: In this study, 1,580 patients were identified as samples for analysis. Through the process of decollinearity and characterization, seven routine blood markers: Basophil (BASO), Hemoglobin (HB), Mean erythrocyte hemoglobin concentration (MCHC), mean corpuscular volume (MCV), mean platelet volume (MPV), Red Blood Cell (RBC), Lymph and red blood cell distribution width coefficient of variation (RDW-CV), one cytokine: interleukin-6 (IL-6), and three lymphocyte-related indices: CD4+ T cell, CD4+/CD8+ T cells, CD8+ T cells were included in the subsequent model construction. Then nine machine learning (ML) algorithms were used to construct predictive models.
RESULTS: The most outstanding one is the K-Nearest Neighbors (KNN) model, which has an AUC value as high as 0.846, with sensitivity and specificity of 0.769 and 0.786, respectively, shows strong prediction effectiveness. Among them, CD4+ T cells, HB, and MPV ranked the top three in terms of importance.
DISCUSSION: In conclusion, we have developed an interpretable ML model to identify EPTB based on data from routine laboratory tests, which provides some assistance in early diagnosis.
METHODS: In this study, 1,580 patients were identified as samples for analysis. Through the process of decollinearity and characterization, seven routine blood markers: Basophil (BASO), Hemoglobin (HB), Mean erythrocyte hemoglobin concentration (MCHC), mean corpuscular volume (MCV), mean platelet volume (MPV), Red Blood Cell (RBC), Lymph and red blood cell distribution width coefficient of variation (RDW-CV), one cytokine: interleukin-6 (IL-6), and three lymphocyte-related indices: CD4+ T cell, CD4+/CD8+ T cells, CD8+ T cells were included in the subsequent model construction. Then nine machine learning (ML) algorithms were used to construct predictive models.
RESULTS: The most outstanding one is the K-Nearest Neighbors (KNN) model, which has an AUC value as high as 0.846, with sensitivity and specificity of 0.769 and 0.786, respectively, shows strong prediction effectiveness. Among them, CD4+ T cells, HB, and MPV ranked the top three in terms of importance.
DISCUSSION: In conclusion, we have developed an interpretable ML model to identify EPTB based on data from routine laboratory tests, which provides some assistance in early diagnosis.