Recursive feature elimination for biomarker discovery in resting-state functional connectivity

Abstract

Biomarker discovery involves finding correlations between features and clinical symptoms to aid clinical decision. This task is especially difficult in resting state functional magnetic resonance imaging (rs-fMRI) data due to low SNR, high-dimensionality of images, inter-subject and intra-subject variability and small numbers of subjects compared to the number of derived features. Traditional univariate analysis suffers from the problem of multiple comparisons. Here, we adopt an alternative data-driven method for identifying population differences in functional connectivity. We propose a machine-learning approach to down-select functional connectivity features associated with symptom severity in mild traumatic brain injury (mTBI). Using this approach, we identified functional regions with altered connectivity in mTBI. including the executive control, visual and precuneus networks. We compared functional connections at multiple resolutions to determine which scale would be more sensitive to changes related to patient recovery. These modular network-level features can be used as diagnostic tools for predicting disease severity and recovery profiles.

Recommendations towards standards for quantitative MRI (qMRI) and outstanding needs

“…Even with these technical considerations, qMRI biomarkers were identified and related to clinical measures of severity and progression of the disease. Resting state functional MRI results demonstrated a remarkable correlation with clinical presentation and symptomatology, perhaps creating an opportunity to replace some of the subjective symptom inventories with objective imaging‐based measures of the severity of the disease. Perfusion arterial spin labeling demonstrated decoupling of the clinical and physiological recovery of the brain .…”

Random-forest algorithm based biomarkers in predicting prognosis in the patients with hepatocellular carcinoma

“…For the prognostic signature analysis, the 369 HCC samples that contained complete clinical data were assigned into groups of good or poor prognosis according to the 5 years survival (expected survival time > 5 or < 5 years). Recursive Feature Elimination (RFE) based on Random Forest (RF) classifier, a method of supervised Machine Learning, was conducted to identify the prognostic genes in survival [18,19]. The prediction was examined by fivefolds cross-validation.…”

Decision tree based hardware power monitoring for run time dynamic power management in FPGA

“…Feature selection can circumvent the overfitting problem and reduce the resource overhead induced by activity counters used to monitor signals. We leverage recursive feature elimination algorithm to identify required features, which has been successfully applied to other domains [15]. In each iteration, the decision tree model is trained and the feature importance values are computed, according to the node impurity at different splits, which is also known as the Gini importance [14] in the decision tree.…”