BMJ open · 2025
Development and validation of an interpretable machine learning model for predicting Philadelphia chromosome-positive acute lymphoblastic leukaemia using clinical and laboratory parameters: a single-centre retrospective study.
СтатьяВалидационное исследованиеНазвание и аннотация: Перевод пока не подготовлен
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.
Yang Wuchen, Liu Jingya, Gou Yang, Huang Xingqin, Chen Maoshan, Huang Dezhi, Wu Shengwang, Zhang Jing, Zhang Cheng, Liu Shuiqing, Peng Xiangui, Zhang Xi
Аннотация
OBJECTIVE: To develop and validate a prediction model of Philadelphia chromosome-positive acute lymphoblastic leukaemia (Ph+ALL).
DESIGN: A single-centre retrospective study.
PARTICIPANTS: This study analysed 471 newly diagnosed patients with ALL at the Second Affiliated Hospital of Army Medical University from January 2014 to December 2023.
METHODS: Clinical and laboratory parameters were collected, and the important characteristic parameters were selected using BorutaShap. Multiple machine learning (ML) models were constructed and optimised by using the active learning (AL) algorithm. Performance was evaluated using the area under the curve (AUC), comprehensive indicators and decision curve analysis. The interpretability of the model was evaluated by using SHapley Additive Interpretation (SHAP), and external validation was conducted on an independent test cohort.
RESULTS: 10 parameters were selected to construct multiple ML models. The CatBoost model integrated with an AL algorithm (CatBoost-AL) was found to be the most effective model for predicting Ph+ALL within the validation data set. This model achieved an AUC of 0.797 (95% CI 0.710 to 0.884), along with sensitivity, specificity and F1 score of 0.667, 0.864 and 0.777, respectively. The prediction performance of CatBoost-AL was further validated with an external testing set, where it maintained a strong AUC of 0.794 (95% CI 0.707 to 0.881). Using SHAP for global interpretability analysis, age, monocyte count, γ-glutamyl transferase, neutrophil count and alanine aminotransferase were identified as crucial parameters that significantly influence the diagnostic accuracy of CatBoost-AL.
CONCLUSION: An interpretable ML model and online prediction tool were developed to determine whether newly diagnosed patients with ALL are Ph+ALL. The key parameters identified by the optimal model provided a further understanding of Ph+ALL characteristics and were valuable for accurate diagnosis and treatment of Ph+ALL.
DESIGN: A single-centre retrospective study.
PARTICIPANTS: This study analysed 471 newly diagnosed patients with ALL at the Second Affiliated Hospital of Army Medical University from January 2014 to December 2023.
METHODS: Clinical and laboratory parameters were collected, and the important characteristic parameters were selected using BorutaShap. Multiple machine learning (ML) models were constructed and optimised by using the active learning (AL) algorithm. Performance was evaluated using the area under the curve (AUC), comprehensive indicators and decision curve analysis. The interpretability of the model was evaluated by using SHapley Additive Interpretation (SHAP), and external validation was conducted on an independent test cohort.
RESULTS: 10 parameters were selected to construct multiple ML models. The CatBoost model integrated with an AL algorithm (CatBoost-AL) was found to be the most effective model for predicting Ph+ALL within the validation data set. This model achieved an AUC of 0.797 (95% CI 0.710 to 0.884), along with sensitivity, specificity and F1 score of 0.667, 0.864 and 0.777, respectively. The prediction performance of CatBoost-AL was further validated with an external testing set, where it maintained a strong AUC of 0.794 (95% CI 0.707 to 0.881). Using SHAP for global interpretability analysis, age, monocyte count, γ-glutamyl transferase, neutrophil count and alanine aminotransferase were identified as crucial parameters that significantly influence the diagnostic accuracy of CatBoost-AL.
CONCLUSION: An interpretable ML model and online prediction tool were developed to determine whether newly diagnosed patients with ALL are Ph+ALL. The key parameters identified by the optimal model provided a further understanding of Ph+ALL characteristics and were valuable for accurate diagnosis and treatment of Ph+ALL.