文献速递头颈部肿瘤放疗放疗物理与技术

确定头颈癌治疗相关淋巴水肿的风险与预测因素:一种使用可解释机器学习和集成特征选择的临床病理与剂量学数据挖掘方法

Determining risk and predictors of head and neck cancer treatment-related lymphedema: A clinicopathologic and dosimetric data mining approach using interpretable machine learning and ensemble feature selection.
PubMed 文献 · Clin Transl Radiat Oncol · 2024-02-28

本研究从76份头颈癌放疗计划中提取30个危及器官(含双侧颈部淋巴结分区)的剂量体积数据及临床病理资料,用集成特征选择降维后输入机器学习与竞争风险模型预测淋巴水肿。XGBoost与随机森林表现稳健,外部淋巴水肿平均F1为84±3.3%、AUC为79±11.9%,内部淋巴水肿平均F1为64±12%、AUC为78±7.9%。

要点

  • XGBoost与随机森林对外部淋巴水肿预测F1为84±3.3%、AUC为79±11.9%。
  • 两模型对内部淋巴水肿预测F1为64±12%、AUC为78±7.9%。
  • 180天时切除0-25、26-50、>50枚淋巴结的外部淋巴水肿风险分别为72.1%、95.6%、57.7%(p=0.01)。

核心结论

  • 降低头颈放疗数据维度的方法产生了有效的头颈淋巴水肿预测机器学习模型。
  • 预测模型与竞争风险模型均筛选出剂量学预测特征,且与其他研究临床病理特征的一致性支持该方法。

研究设计与样本

  • 回顾性数据挖掘研究,基于76份头颈癌放疗计划,勾画30个危及器官(含双侧颈部淋巴结分区)并提取剂量体积数据,同时收集临床病理资料。
  • 采用集成特征选择降维,以降维后特征输入机器学习模型和竞争风险模型,评估内部与外部淋巴水肿预测能力,并估计淋巴水肿事件时间与风险分层。

主要发现

  • XGBoost与随机森林表现稳健:外部淋巴水肿平均F1为84±3.3%、平均AUC为79±11.9%;内部淋巴水肿平均F1为64±12%、平均AUC为78±7.9%。
  • 预测机器学习模型与风险模型识别出共同预测因素,包括淋巴结受累体积大、多个淋巴结分区受照剂量高及手术切除淋巴结。
  • 180天时,切除0-25、26-50和>50枚淋巴结者的外部淋巴水肿风险分别为72.1%、95.6%和57.7%(p=0.01)。

关键数值与推荐

  • 外部淋巴水肿预测:XGBoost与随机森林平均F1为84±3.3%,平均AUC为79±11.9%。
  • 内部淋巴水肿预测:XGBoost与随机森林平均F1为64±12%,平均AUC为78±7.9%。
  • 180天外部淋巴水肿风险:切除0-25枚淋巴结为72.1%,26-50枚为95.6%,>50枚为57.7%(p=0.01)。

临床怎么用

  • 原文未据此给出推荐。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND AND PURPOSE: The ability to determine the risk and predictors of lymphedema is vital in improving the quality of life for head and neck (HN) cancer patients. However, selecting robust features is challenging due to the multicollinearity and high dimensionality of radiotherapy (RT) data. This study aims to overcome these challenges using an ensemble feature selection technique with machine learning (ML). MATERIALS AND METHODS: Thirty organs-at-risk, including bilateral cervical lymph node levels, were contoured, and dose-volume data were extracted from 76 HN treatment plans. Clinicopathologic data was collected. Ensemble feature selection was used to reduce the number of features. Using the reduced features as input to ML and competing risk models, internal and external lymphedema prediction capability was evaluated with the ML models, and time to lymphedema event and risk stratification were estimated using the risk models. RESULTS: Two ML models, XGBoost and random forest, exhibited robust prediction performance. They achieved average F1-scores and AUCs of 84 ± 3.3 % and 79 ± 11.9 % (external lymphedema), and 64 ± 12 % and 78 ± 7.9 % (internal lymphedema). Predictive ML and risk models identified common predictors, including bulky node involvement, high dose to various lymph node levels, and lymph nodes removed during surgery. At 180 days, removing 0-25, 26-50, and > 50 lymph nodes increased external lymphedema risk to 72.1 %, 95.6 %, and 57.7 % respectively (p = 0.01). CONCLUSION: Our approach, involving the reduction of HN RT data dimensionality, resulted in effective ML models for HN lymphedema prediction. Predictive dosimetric features emerged from both predictive and competing risk models. Consistency with clinicopathologic features from other studies supports our methodology.

来源:Clin Transl Radiat Oncol,查看原文。版权归原作者所有。

引用

  1. Teo PT, Rogacki K, Gopalakrishnan M, et al. Determining risk and predictors of head and neck cancer treatment-related lymphedema: A clinicopathologic and dosimetric data mining approach using interpretable machine learning and ensemble feature selection. Clin Transl Radiat Oncol. 2024. PMID: 38450218. doi:10.1016/j.ctro.2024.100747

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。