Cancer imaging phenomics toolkit: quantitative imaging analytics for precision diagnostics and predictive modeling of clinical outcome

Abstract

The growth of multiparametric imaging protocols has paved the way for quantitative imaging phenotypes that predict treatment response and clinical outcome, reflect underlying cancer molecular characteristics and spatiotemporal heterogeneity, and can guide personalized treatment planning. This growth has underlined the need for efficient quantitative analytics to derive high-dimensional imaging signatures of diagnostic and predictive value in this emerging era of integrated precision diagnostics. This paper presents cancer imaging phenomics toolkit (CaPTk), a new and dynamically growing software platform for analysis of radiographic images of cancer, currently focusing on brain, breast, and lung cancer. CaPTk leverages the value of quantitative imaging analytics along with machine learning to derive phenotypic imaging signatures, based on two-level functionality. First, image analysis algorithms are used to extract comprehensive panels of diverse and complementary features, such as multiparametric intensity histogram distributions, texture, shape, kinetics, connectomics, and spatial patterns. At the second level, these quantitative imaging signatures are fed into multivariate machine learning models to produce diagnostic, prognostic, and predictive biomarkers. Results from clinical studies in three areas are shown: (i) computational neuro-oncology of brain gliomas for precision diagnostics, prediction of outcome, and treatment planning; (ii) prediction of treatment response for breast and lung cancer, and (iii) risk assessment for breast cancer.

Epidermal Growth Factor Receptor Extracellular Domain Mutations in Glioblastoma Present Opportunities for Clinical Imaging and Therapeutic Development

“…The software tool used for co-registration (i.e., FLIRT (Jenkinson et al, 2002)) is available in: fsl.fmrib.ox.ac.uk. We developed the Cancer Imaging Phenomics Toolkit (CaPTk) (Davatzikos et al, 2018) to facilitate clinical translation of complex computational algorithms, without requiring computational background (e.g., identification of genetic mutation imaging signatures (Bakas et al, 2017a)). Specifically for this study, CaPTk was used for 1) initialization of seed-points required by GLISTRboost, 2) image smoothing, as well as 3) extracting the quantitative imaging features.…”

Unsupervised machine learning of radiomic features for predicting treatment response and overall survival of early stage non-small cell lung cancer patients treated with stereotactic body radiation therapy

“…For each patient, the solid component of NSCLC tumor was delineated using CaPTk software [32]. Specifically, candidate tumor regions were detected automatically using a random walk based image segmentation method based on its PET and CT images [33–35], the primary solid component was then identified by one experienced radiologist, and finally the segmentation result was further checked visually and modified manually if necessary.…”

“…For predicting each patient’s risk of mortality and nodal failure, the meta-features extracted for each patient were used to build prediction models using 3 different survival modeling techniques, including Cox proportional hazard regression (Cox regression) [30], Cox regression with LASSO (Cox_lasso) [31], and random survival forests (RSF) [32]. Particularly, the Cox regression method is a standard survival modeling technique.…”

Multivariate Analysis of Preoperative Magnetic Resonance Imaging Reveals Transcriptomic Classification of de novo Glioblastoma Patients

“…The radiophenotypic characteristics of each tumor were quantified using a comprehensive and diverse set of imaging features, extracted from all tumor sub-regions (i.e., ED, ET, NET) and all MRI sequences using the Cancer Imaging Phenomics Toolkit (CaPTk) (Davatzikos et al, 2018). The feature set extracted to build the predictive model for this study comprised of (i) volumetric measurements, (ii) morphology parameters, (iii) location information, and (iv) statistical moments of the intensity distributions.…”