結合機器學習與臨床指引之急性骨髓性白血病風險分層集成模型

An Ensemble Model for Acute Myeloid Leukemia Risk Stratification Recommendations by Combining Machine Learning with Clinical Guidelines

張名翔(Ming-Siang Chang)

指導教授 ： 陳倩瑜

國立臺灣大學/生物資源暨農學院/生物機電工程學系/碩士(2023年)

https://doi.org/10.6342/NTU202303243

摘要

急性骨髓性白血病（Acute myeloid leukemia, AML）是一種致命的血液疾病，由異常白細胞引起並在骨髓中發展。它會導致血小板減少，增加出血和感染的可能性。本論文開發了一個機器學習集成（ensemble）模型，使用國立台灣大學附設醫院 1213 名 AML患者的數據集，對AML風險進行分層，本研究提出的方法結合機器學習集成模型預測的結果和2017年歐洲白血病網（European LeukemiaNet 2017, ELN 2017）預測的結果，進一步合成最終的集成模型Ensemble (ML+ELN)，提出了初步的臨床風險分層建議。與ELN 2017臨床診斷建議相比，本研究的風險分層建議提供了最佳區分各種風險的能力，c-index 由0.64提升至0.66。特別在區分不利風險和中等風險上，相較於2017 ELN的p值（p-value）平均0.13，本研究的風險分層建議達到p值平均0.001的表現。

關鍵字

急性骨髓性白血病 ； 集成模型 ； 機器學習 ； 風險分層 ； 歐洲白血病網

並列摘要

Acute myeloid leukemia (AML), a fatal blood condition, is brought on by abnormal white blood cells and develops in the bone marrow. It results in a decrease in platelets, raising the possibility of bleeding and infection. This study developed an ML-based ensemble model to stratify the risk of AML using a dataset containing 1213 AML patients from the National Taiwan University Hospital. Combining the ML-based ensemble model predictions and the European LeukemiaNet (ELN) 2017 predictions, the study represents a final ensemble model (ML+ELN) for initial clinical risk stratification recommendations. Compared to the clinical diagnostic recommendations ELN 2017, the proposed risk stratification proposal provides a superior capacity to distinguish various risks and improve the c-index from 0.64 to 0.66. Especially in distinguishing unfavorable risks from moderate risks, compared with the average p-value (p-value) of 0.13 in 2017 ELN, the proposed risk stratification proposal achieves excellent performance with an average p-value of 0.001.

並列關鍵字

acute myeloid leukemia ； ensemble model ； machine learning ； risk stratification ； European LeukemiaNet

參考文獻

Albawi, S., Mohammed, T. A., & Al-Zawi, S. (2017). Understanding of a convolutional neural network. 2017 international conference on engineering and technology (ICET),

Alshdaifat, E. a., Alshdaifat, D. a., Alsarhan, A., Hussein, F., & El-Salhi, S. M. d. F. S. (2021). The Effect of Preprocessing Techniques, Applied to Numeric Features, on Classification Algorithms’ Performance. Data, 6(2), 11. https://www.mdpi.com/2306-5729/6/2/11

Bergstra, J., Yamins, D., & Cox, D. D. (2013). Hyperopt: A python library for optimizing the hyperparameters of machine learning algorithms. Proceedings of the 12th Python in science conference,

Breiman, L. (2001). Random forests. Machine learning, 45, 5-32.

Chen, T., & Guestrin, C. (2016). Xgboost: A scalable tree boosting system. Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining,