1 Scientific RepoRts | 6:25295 | DOI: 10.1038/srep25295 www.nature.com/scientificreports Radiomic texture Analysis Mapping predicts Areas of true Functional MRI Activity Islam Hassan 1 , Aikaterini Kotrotsou 1 , Ali shojaee Bakhtiari 1 , Ginu A. thomas 1 , Jefrey S. Weinberg 2 , Ashok J. Kumar 1 , Raymond sawaya 2 , Markus M. Luedi 1 , pascal o. Zinn 3 & Rivka R. Colen 1,4 Individual analysis of functional Magnetic Resonance Imaging (fMRI) scans requires user-adjustment of the statistical threshold in order to maximize true functional activity and eliminate false positives. In this study, we propose a novel technique that uses radiomic texture analysis (tA) features associated with heterogeneity to predict areas of true functional activity. Scans of 15 right-handed healthy volunteers were analyzed using SPM8. The resulting functional maps were thresholded to optimize visualization of language areas, resulting in 116 regions of interests (ROIs). A board-certifed neuroradiologist classifed diferent ROIs into Expected (E) and Non-Expected (NE) based on their anatomical locations. TA was performed using the mean Echo-Planner Imaging (EPI) volume, and 20 rotation-invariant texture features were obtained for each ROI. Using forward stepwise logistic regression, we built a predictive model that discriminated between E and NE areas of functional activity, with a cross-validation AUC and success rate of 79.84% and 80.19% respectively (specifcity/ sensitivity of 78.34%/82.61%). This study found that radiomic TA of fMRI scans may allow for determination of areas of true functional activity, and thus eliminate clinician bias. Blood oxygen level dependent Magnetic Resonance Imaging (BOLD-MRI) is one of the most important tools in presurgical neuroimaging as it refects the integrated synaptic activity of neurons 1,2 . Since the development of functional MRI (fMRI) as a technique for brain mapping, it has been extensively used in multiple clinical and research applications 3 . Tis can be attributed to its non-invasive nature and high spatial and temporal resolution, which allows covering of the entire brain within a short period of time 4 . Early on, fMRI was used as a tool in neu- rocognitive research using group analysis rather than individual analysis 5 . In group analysis, data from diferent subjects are averaged in order to cancel-out random contributions and increase signal-to-noise ratio (SNR) 5 . However, individual analysis is the only option in clinical decision-making such as in presurgical brain mapping. Currently, individual analysis of fMRI data consists of steps that reduce the SNR and increase the contrast-to-noise ratio (CNR) 6 . Te ultimate goal is to maximize detection of true activity and eliminate any false positives, which is achieved through adjusting the statistical threshold of fMRI map 7 . In individual analysis, deter- mination of this threshold is arbitrary and difers from one subject to another, depending on the experience of the reporting radiologist 8 . Usually, the threshold is set to the point where maximum noise can be eliminated without afecting true activity 9 . However, it remains unclear whether the fnal fMRI map is a true representation of brain activity. Further, no method is known that eliminates non-essential or untrue activity that survives the arbitrary threshold and the limits of thresholding are not identifed. Finally, it has to be proven if a stringent threshold always results in preservation of truly active areas. Tose questions become of extreme importance in cases of clinical applications, specifcally presurgical map- ping, where accuracy is pivotal for clinical decision making 10,11 . Despite multiple validation studies of fMRI, individual fMRI results per-se cannot be considered 100% accurate due to several factors 12 . First, fMRI is an indirect measurement of brain activity, thus it is an overstretch to assume that BOLD-signal represents activity of a specifc brain region associated with the evaluated function 4,8 . Second, no statistical method can provide 1 Department of Diagnostic Radiology, the University of texas MD Anderson cancer center, Houston, texas, USA. 2 Department of neurosurgery, the University of texas MD Anderson cancer center, Houston, texas, USA. 3 Department of neurosurgery, Baylor college of Medicine, Houston, texas, USA. 4 Department of cancer Systems imaging, the University of texas MD Anderson cancer center, Houston, texas, USA. correspondence and requests for materials should be addressed to R.R.c. (email: rcolen@mdanderson.org) Received: 25 January 2016 Accepted: 14 April 2016 Published: 06 May 2016 OPEN