Najafian Keyhan, Rehany Benjamin, Nowakowski Alexander, Ghazimoghadam Saba, Pierre Kevin, Zakarian Rita, Al-Saadi Tariq, Reinhold Caroline, Babajani-Feremi Abbas, Wong Joshua K, Guiot Marie-Christine, Lacasse Marie-Constance, Lam Stephanie, Siegel Peter M, Petrecca Kevin, Dankner Matthew, Forghani Reza
Department of Computer Science, University of Saskatchewan, Saskatoon, Saskatchewan, Canada.
Augmented Intelligence and Precision Health Laboratory, McGill University. Montreal, Quebec, Canada.
Neurooncol Adv. 2024 Nov 16;6(1):vdae200. doi: 10.1093/noajnl/vdae200. eCollection 2024 Jan-Dec.
Brain metastasis invasion pattern (BMIP) is an emerging biomarker associated with recurrence-free and overall survival in patients, and differential response to therapy in preclinical models. Currently, BMIP can only be determined from the histopathological examination of surgical specimens, precluding its use as a biomarker prior to therapy initiation. The aim of this study was to investigate the potential of machine learning (ML) approaches to develop a noninvasive magnetic resonance imaging (MRI)-based biomarker for BMIP determination.
From an initial cohort of 329 patients, a subset of 132 patients met the inclusion criteria for this retrospective study. We evaluated the ability of an expert neuroradiologist to reliably predict BMIP. Thereafter, the dataset was randomly divided into training/validation (80% of cases) and test subsets (20% of cases). The ground truth for BMIP was the histopathologic evaluation of resected specimens. Following MRI sequence co-registration, advanced feature extraction techniques deriving hand-crafted radiomic features with traditional ML classifiers and convolution-based deep learning (CDL) models were trained and evaluated. Different ML approaches were used individually or using ensembling techniques to determine the model with the best performance for BMIP prediction.
Expert evaluation of brain MRI scans could not reliably predict BMIP, with an accuracy of 44%-59% depending on the semantic feature used. Among the different ML and CDL models evaluated, the best-performing model achieved an accuracy of 85% and an F1 score of 90%.
ML approaches can effectively predict BMIP, representing a noninvasive MRI-based approach to guide the management of patients with brain metastases.
脑转移瘤侵袭模式（BMIP）是一种新兴的生物标志物，与患者的无复发生存期和总生存期以及临床前模型中对治疗的不同反应相关。目前，BMIP只能通过手术标本的组织病理学检查来确定，这使得它无法在治疗开始前用作生物标志物。本研究的目的是探讨机器学习（ML）方法开发基于磁共振成像（MRI）的非侵入性生物标志物以确定BMIP的潜力。
在最初的329例患者队列中，132例患者的子集符合这项回顾性研究的纳入标准。我们评估了一位神经放射学专家可靠预测BMIP的能力。此后，将数据集随机分为训练/验证集（病例的80%）和测试子集（病例的20%）。BMIP的真实情况是对切除标本的组织病理学评估。在MRI序列配准后，使用传统ML分类器和基于卷积的深度学习（CDL）模型提取手工制作的放射组学特征的先进特征提取技术进行训练和评估。单独使用不同的ML方法或使用集成技术来确定对BMIP预测性能最佳的模型。
对脑部MRI扫描的专家评估无法可靠地预测BMIP，根据所使用的语义特征，准确率在44%至59%之间。在评估的不同ML和CDL模型中，表现最佳的模型准确率达到85%，F1分数达到90%。
ML方法可以有效预测BMIP，代表了一种基于MRI的非侵入性方法来指导脑转移瘤患者的管理。
