An investigation of machine learning methods in delta-radiomics feature analysis

Abstract

PurposeThis study aimed to investigate the effectiveness of using delta-radiomics to predict overall survival (OS) for patients with recurrent malignant gliomas treated by concurrent stereotactic radiosurgery and bevacizumab, and to investigate the effectiveness of machine learning methods for delta-radiomics feature selection and building classification models.MethodsThe pre-treatment, one-week post-treatment, and two-month post-treatment T1 and T2 fluid-attenuated inversion recovery (FLAIR) MRI were acquired. 61 radiomic features (intensity histogram-based, morphological, and texture features) were extracted from the gross tumor volume in each image. Delta-radiomics were calculated between the pre-treatment and post-treatment features. Univariate Cox regression and 3 multivariate machine learning methods (L1-regularized logistic regression [L1-LR], random forest [RF] or neural networks [NN]) were used to select a reduced number of features, and 7 machine learning methods (L1-LR, L2-LR, RF, NN, kernel support vector machine [KSVM], linear support vector machine [LSVM], or naïve bayes [NB]) was used to build classification models for predicting OS. The performances of the total 21 model combinations built based on single-time-point radiomics (pre-treatment, one-week post-treatment, and two-month post-treatment) and delta-radiomics were evaluated by the area under the receiver operating characteristic curve (AUC).ResultsFor a small cohort of 12 patients, delta-radiomics resulted in significantly higher AUC than pre-treatment radiomics (p-value<0.01). One-week/two-month delta-features resulted in significantly higher AUC (p-value<0.01) than the one-week/two-month post-treatment features, respectively. 18/21 model combinations were with higher AUC from one-week delta-features than two-month delta-features. With one-week delta-features, RF feature selector + KSVM classifier and RF feature selector + NN classifier showed the highest AUC of 0.889.ConclusionsThe results indicated that delta-features could potentially provide better treatment assessment than single-time-point features. The treatment assessment is substantially affected by the time point for computing the delta-features and the combination of machine learning methods for feature selection and classification.

Survival Prediction in Gliomas: Current State and Novel Approaches

“…Kebir et al carried out a similar analysis using 14 patients with HGG suspected of progression but used FET-PET imaging to identify 3 clusters based on 10 predominantly textural FET-PET features. Similar studies were also carried out by Chan et al and Petrova et al (64,65).…”

“…The use of imaging to assess response to treatment in glioma is of tremendous importance for both patient management and outcome assessment. With respect to computational approaches on analyzing survival outcomes, research to date has focused on two main aspects: (i) distinguishing progression from pseudo-progression (62,63); and (ii) connecting systemic therapy or radiation therapy (RT) to imaging changes (64). Artzi et al utilized data generated using conventional, dynamic contrast enhanced (DCE)-MRI and magnetic resonance spectroscopy from 20 patients to extract the enhanced lesion area using independent component analysis and choline/creatine values and compared treatment-related changes with normal-appearing white matter.…”

Intrinsic radiomic expression patterns after 20 Gy demonstrate early metabolic response of oropharyngeal cancers

“…It is generally well-known that radiomic features are susceptible to scanner variation and inter-observer variability, 53 which can often lead to inconsistent downstream machine learning models. 54,55 To the best of our knowledge, our study is the first reported head and neck cancer radiomic analysis based on prospectively acquired intra-treatment 18 F-FDG-PET images and long-term clinical follow-up. Our findings are therefore strengthened by a highfidelity dataset representing a homogenous patient cohort without inter-scanner variability across different vendors, models, or acquisition protocols.…”

“…While most published radiomic studies rely on retrospective data, our results are based on a prospective clinical trial where all images were acquired on the same scanner under the same acquisition protocol. It is generally well‐known that radiomic features are susceptible to scanner variation and inter‐observer variability, 53 which can often lead to inconsistent downstream machine learning models 54,55 . To the best of our knowledge, our study is the first reported head and neck cancer radiomic analysis based on prospectively acquired intra‐treatment 18 F‐FDG‐PET images and long‐term clinical follow‐up.…”

Artificial intelligence for clinical oncology

“…For example, radiomic features extracted from the same tumor, pre-and post-treatment, show significant discrepancies (van Dijk et al, 2019). These ''delta'' features could be used to predict patient recurrence risk and late toxicity, helping to tailor followup plans (Chang et al, 2019). Appropriately triaging patients for escalated follow-up and attention can promote decreased morbidity and more efficient healthcare resource utilization; AI leveraging EHR data has demonstrated the ability to accomplish this by selecting patients at high risk for acute-care visit while undergoing cancer therapy and assigning them to an escalated preventive care strategy (Hong et al, 2020).…”