On the Relationship between Feature Selection Metrics and Accuracy.

Feature selection metrics are commonly used in the machine learning pipeline to rank and select features before creating a predictive model. While many different metrics have been proposed for feature selection, final models are often evaluated by accuracy. In this paper, we consider the relationship between common feature selection metrics and accuracy. In particular, we focus on misorderings: cases where a feature selection metric may rank features differently than accuracy would. We analytically investigate the frequency of misordering for a variety of feature selection metrics as a function of parameters that represent how a feature partitions the data. Our analysis reveals that different metrics have systematic differences in how likely they are to misorder features which can happen over a wide range of partition parameters. We then perform an empirical evaluation with different feature selection metrics on several real-world datasets to measure misordering. Our empirical results generally match our analytical results, illustrating that misordering features happens in practice and can provide some insight into the performance of feature selection metrics.

Phantom Validation of DCE-MRI Magnitude and Phase-Based Vascular Input Function Measurements.

Accurate, patient-specific measurement of arterial input functions (AIF) may improve model-based analysis of vascular permeability. This study investigated factors affecting AIF measurements from magnetic resonance imaging (MRI) magnitude (AIFMAGN) and phase (AIFPHA) signals, and compared them against computed tomography (CT) (AIFCT), under controlled conditions relevant to clinical protocols using a multimodality flow phantom. The flow phantom was applied at flip angles of 20° and 30°, flow rates (3-7.5 mL/s), and peak bolus concentrations (0.5-10 mM), for in-plane and through-plane flow. Spatial 3D-FLASH signal and variable flip angle T1 profiles were measured to investigate in-flow and radiofrequency-related biases, and magnitude- and phase-derived Gd-DTPA concentrations were compared. MRI AIF performance was tested against AIFCT via Pearson correlation analysis. AIFMAGN was sensitive to imaging orientation, spatial location, flip angle, and flow rate, and it grossly underestimated AIFCT peak concentrations. Conversion to Gd-DTPA concentration using T1 taken at the same orientation and flow rate as the dynamic contrast-enhanced acquisition improved AIFMAGN accuracy; yet, AIFMAGN metrics remained variable and significantly reduced from AIFCT at concentrations above 2.5 mM. AIFPHA performed equivalently within 1 mM to AIFCT across all tested conditions. AIFPHA, but not AIFMAGN, reported equivalent measurements to AIFCT across the range of tested conditions. AIFPHA showed superior robustness.