Application of machine learning classifiers to X‐ray diffraction imaging with medically relevant phantoms

Abstract

Recent studies have demonstrated the ability to rapidly produce large field of view X-ray diffraction (XRD) images, which provide rich new data relevant to the understanding and analysis of disease. However, work has only just begun on developing algorithms that maximize the performance toward decision-making and diagnostic tasks. In this study, we present the implementation of and comparison between rules-based and machine learning (ML) classifiers on XRD images of medically relevant phantoms to explore the potential for increased classification performance. Methods: Medically relevant phantoms were utilized to provide wellcharacterized ground-truths for comparing classifier performance. Water and polylactic acid (PLA) plastic were used as surrogates for cancerous and healthy tissue, respectively, and phantoms were created with varying levels of spatial complexity and biologically relevant features for quantitative testing of classifier performance. Our previously developed X-ray scanner was used to acquire co-registered X-ray transmission and diffraction images of the phantoms. For classification algorithms, we explored and compared two rules-based classifiers (cross-correlation, or matched-filter, and linear least-squares unmixing) and two ML classifiers (support vector machines and shallow neural networks). Reference XRD spectra (measured by a commercial diffractometer) were provided to the rules-based algorithms, while 60% of the measured XRD pixels were used for training of the ML algorithms. The area under the receiver operating characteristic curve (AUC) was used as a comparative metric between the classification algorithms, along with the accuracy performance at the midpoint threshold for each classifier. Results:The AUC values for material classification were 0.994 (crosscorrelation [CC]), 0.994 (least-squares [LS]), 0.995 (support vector machine [SVM]), and 0.999 (shallow neural network [SNN]). Setting the classification threshold to the midpoint for each classifier resulted in accuracy values of CC = 96.48%, LS = 96.48%, SVM = 97.36%, and SNN = 98.94%. If only considering pixels ±3 mm from water-PLA boundaries (where partial volume effects could occur due to imaging resolution limits), the classification accuracies were CC = 89.32%, LS = 89.32%, SVM = 92.03%, and SNN = 96.79%, demonstrating an even larger improvement produced by the machine-learned algorithms in spatial regions critical for imaging tasks. Classification by transmission data alone produced an AUC of 0.773 and accuracy of 85.45%, well below the performance levels of any of the classifiers applied to XRD image data. Conclusions: We demonstrated that ML-based classifiers outperformed rulesbased approaches in terms of overall classification accuracy and improved the spatially resolved classification performance on XRD images of medical phantoms. In particular, the ML algorithms demonstrated considerably improved 532

Performance of a high-throughput full-tunnel XRD checked baggage scanning system

“…Three approaches to classification were compared using different feature sets. The data from the approximately 650 scans was used to train and test a support vector machine (SVM) classifier [7][8][9]. No certified algorithms were used in the calculation of these results.…”

Detector comparison for high-resolution 3D x-ray diffraction imaging for biospecimen analysis

“…In addition, research has identified the importance of the multicellular environment that surrounds cancer (known as the tumor micro-environment or TME) in terms of better understanding its genesis, progression, and its most effective modes of treatment 11 . Relevant TME size scales are on the order of 100's of microns, which suggests that the approximately millimeter spatial resolution demonstrated in prior work 5,12 is insufficient for TME-related research applications.…”

“…Then the range of q-values between 0.1 and 0.5 1/Å were binned by a width of 0.01 1/Å. This gives a q-value range and resolution similar to those used in our prior XRDI machine and experiments 1,5,12 .…”

“…However, numerous system architectures have been evaluated for realizing spatially resolved XRD imaging [2][3][4] . XRDI has applications that range from airport security to pathology and medicine, the latter of which is the primary motivation for this work [5][6][7][8] .…”

Towards a high-resolution volumetric x-ray diffraction imaging system for biospecimen

“…While traditionally used for material differentiation or identification, these measurements are typically made only at a single, relatively large area point on the surface of a sample. Yet, numerous applications, including material structural failure analysis, security, and medicine, require datasets that contain spatially resolved and material-specific information [5][6][7][8][9][10]. Due to recent improvements in the system architectures that generate spatially encoded XRD measurements, combined with computational advancements in signal processing and optimized reconstruction, XRD-based imaging techniques are becoming more broadly applicable, and can be tailored to task specific needs for performance metrics.…”