Radiomic features from the necrotic region on post-treatment Gadolinium T1w MRI appear to differentiate pseudo-progression from true tumor progression in primary brain tumors
Prateek Prasanna1, Raymond Huang2, Andrew Rose1, Gagandeep Singh1, Anant Madabhushi1, and Pallavi Tiwari1

1Case Western Reserve University, Cleveland, OH, United States, 2Brigham and Women's Hospital, Boston, MA, United States

Synopsis

Pseudoprogression is an early-delayed inflammatory response to chemoradiotherapy typically appearing up to 3 months post-treatment in brain tumors. On routine MRI, pseudoprogression closely mimics the appearance of true progression, thereby making their visual identification challenging. Early diagnosis of pseudoprogression has implications in management of treatment effects and subsequently survival. We present initial results of using a newly developed radiomic descriptor, CoLlAGE, in distinguishing the two pathologies. We report that CoLlAGe measurements when captured from the necrotic region as opposed to just the enhancing region on MRI can reliably distinguish psuedo-progression from true progression with 100% accuracy (n=17)

Purpose

Following resection and concomitant radiation treatment in brain tumors, approximately 50% of all patients present with suspicious findings on follow-up MRI indicative of either pseudoprogression (PsP), a benign response to radiation therapy, or “true” tumor progression. On routine MRI scans, PsP closely mimics the appearance of tumor progression, thereby making their visual identification challenging. Guidelines set by RANO/Macdonald’s criteria1,2 used in clinical diagnosis are unreliable as they are based on 2-dimensional measurements of the enhancing tumor alone. Histopathologically, PsP has been found to be related to necrotizing effects with complete absence of tumor cells and characterized by vascular dilation, and endothelial damage of normal cerebral vasculature, while true progression is characterized by the presence of tumor cells, increased cellularity, and vascular proliferation. These changes in heterogeneity across the two conditions are appreciable within and around the necrotic regions of the lesion at a pathologic scale. However radiologists are typically unable to distinguish PsP from cancer recurrence on MRI. In this work, we explore the feasibility of a new computer extracted texture (radiomic) feature called CoLlAGe3 to distinguish PsP from tumor progression in primary brain tumor patients. CoLlAGe aims to capture subtle differences in heterogeneity on a per-voxel basis by measuring entropy (mathematical construct to capture degree of disorder) of gradient orientations. We hypothesize that the CoLlAGe measurements when captured from the necrotic region as opposed to just the enhancing region on MRI, will be substantially different across “true” tumor progression and PsP, potentially capturing the underlying differences in heterogeneity as reflected on a pathologic scale.

Methods

A total of 17 MRI studies (Gd-T1w, T2w, FLAIR) were obtained in an IRB approved study, for PsP (n = 5) or tumor progression (n = 12) as established on follow-up clinical and imaging examinations (Table 1). Our workflow first involved co-registration of Gd-T1w, T2w and FLAIR sequences in order to align the anatomical structures across different imaging sequences4, followed by harmonizing protocol-specific image intensities to template distributions to account for intensity non-standardness5. Expert delineation of enhancing, and necrotic regions was performed on T1w MRI and of edema region on T2w, and FLAIR. CoLlAGe features were then extracted on a per-pixel basis for every region (edema, tumor necrosis, enhancing tumor), for each of the MRI protocols. The values across all the studies were collated for every region as a histogram, and skewness, a measure of asymmetry of probability distributions was used as a feature to distinguish PsP from tumor progression.

Results and Discussion

Skewness of the CoLlAGe distribution for every patient is shown in a linear scale in Figure 1. The normalized mean and standard deviation values of CoLlAGe skewness were 0.07 +/- 0.05 and 0.42 +/- 0.3 for PsP and cancer progression respectively. As observed in Figure 1, at a threshold of 0.08, all of the 12 tumor progression cases as well as all 5 PsP studies were correctly identified, with 100% detection accuracy. Figures 2(a)-(c) show the box-plots of intensity values for enhancing tumor, edema, and tumor necrosis respectively, while Figures 2(d)-(e) show the corresponding box-plots for CoLlAGe values extracted from each of the 3 regions on Gd-T1w scans for all 17 patient studies. Interestingly, the intensity and CoLlAGe values for PsP and tumor progression were fairly overlapping for the enhancing and edema regions, while the CoLlAGe values showed near-perfect separation in the necrotic region for Gd T1w MRI (p-value =0.0003). CoLlAGe values for tumor progression cases were observed to be more negatively skewed as compared to those from PsP cases. The differences in CoLlAGe values in the necrotic region may be on account of differences in necrotizing effects observed at a pathologic scale in patients with PsP (high hypoxia, absence of tumor cells), as opposed to those with tumor progression.

Conclusion

We presented the initial results of employing CoLlAGe, a new radiomic feature, to distinguish PsP from tumor progression with a 100% detection accuracy in n=17 studies. We identified that skewness of CoLlAGe values from within the necrotic region was markedly different between PsP and tumor progression studies (tumor progression more negatively skewed than PsP), as opposed to features obtained from enhancing regions, which showed no statistical differences in values across PsP and tumor progression studies.

Clinical Implication

Reliable distinction of PsP from true progression would allow for early identification of patients with “true” progression who are currently subject to a “wait-and-watch” as their tumor continues to grow. It would also help evaluate the efficacy of treatment procedures thereby facilitating targeted treatment in patients with early treatment failure.

Acknowledgements

Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under award numbers R21CA167811-01, R21CA179327-01, R21CA195152-01, U24CA199374-01the National Institute of Diabetes and Digestive and Kidney Diseases under award number R01DK098503-02, the DOD Prostate Cancer Synergistic Idea Development Award (PC120857); the DOD Lung Cancer Idea Development New Investigator Award (LC130463),the DOD Prostate Cancer Idea Development Award; the Ohio Third Frontier Technology development Grant, the CTSC Coulter Annual Pilot Grant, the Case Comprehensive Cancer Center Pilot Grantthe VelaSano Grant from the Cleveland Clinicthe Wallace H. Coulter Foundation Program in the Department of Biomedical Engineering at Case Western Reserve University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

References

[1] Wen, P. Y., Macdonald, D. R., Reardon, D. A., Cloughesy, T. F., Sorensen, A. G., Galanis, E., & Chang, S. M. (2010). Updated response assessment criteria for high-grade gliomas: response assessment in neuro-oncology working group. Journal of Clinical Oncology, 28(11), 1963-1972.

[2] Chinot, O. L., Macdonald, D. R., Abrey, L. E., Zahlmann, G., Kerloëguen, Y., & Cloughesy, T. F. (2013). Response assessment criteria for glioblastoma: practical adaptation and implementation in clinical trials of antiangiogenic therapy. Current neurology and neuroscience reports, 13(5), 1-11.

[3] Prasanna, P., Tiwari, P., & Madabhushi, A. (2014). Co-occurrence of Local Anisotropic Gradient Orientations (CoLlAGe): Distinguishing Tumor Confounders and Molecular Subtypes on MRI. In Medical Image Computing and Computer-Assisted Intervention–MICCAI 2014 (pp. 73-80). Springer International Publishing.

[4] Fedorov A., Beichel R., Kalpathy-Cramer J., Finet J., Fillion-Robin J-C., Pujol S., Bauer C., Jennings D., Fennessy F., Sonka M., Buatti J., Aylward S.R., Miller J.V., Pieper S., Kikinis R. 3D Slicer as an Image Computing Platform for the Quantitative Imaging Network. Magn Reson Imaging. 2012 Nov;30(9):1323-41

[5] Madabhushi, A., & Udupa, J. K. (2006). New methods of MR image intensity standardization via generalized scale. Medical Physics, 33(9), 3426-3434.

Figures

Figure 1: Per-patient CoLlAGe skewness values from necrosis region shown on a linear scale, with red denoting patients with tumor progression, while blue denotes patients with PsP.

Figure 2: (a)-(c) show the box and whisker plots of intensity values from the three subcompartments in Gd-T1w protocol for both PsP and true progression. Corresponding CoLlAGe skewness values are shown in (d)-(f). Note that only CoLlAGe values extracted from necrotic regions show significant difference.

Table 1: Acquisition parameters for Gd-T1w, T2 and FLAIR scans.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0294