Association of Apparent Diffusion Coefficient with PFS for recurrent Astrocytoma and Generation of Maps to aid in Defining Tumor Pathology
Evan Neill1, Manisha Dayal1, Joanna Phillips2, Llewellyn Jalbert1, Soonme Cha1, Annette Molinaro2, Susan Chang2, and Sarah Nelson1

1Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States, 2Neurological Surgery, University of California, San Francisco, San Francisco, CA, United States


Patients with recurrent low-grade Astrocytomas may experience a large variety of clinical disease courses and outcomes. Diffusion MR imaging and its associations with patient and tumor level characteristics provide an opportunity to use images as maps of tumor aggression and therefore prognosis. This study examines the relationships between the Apparent Diffusion Coefficient (ADC), histology and progress free survival in order to develop a colormap that highlights regions of the tumor with more aggressive characteristics. The colormaps provide an easier way for neuro-oncologists and neurosurgeons to interpret ADC images that can aid them in making decisions about how to manage their patients.


Low-Grade astrocytomas are a class of central nervous system tumor that arise from malignant astrocytes. When these tumors recur histological analysis of tissue samples that are obtained during surgical resection often results in a the conclusion that the lesion has transformed to a higher WHO grade1. Patients experience a broad spectrum of therapeutic approaches and clinical outcomes. Having a non-invasive imaging methodology that is able to identify portions of the tumor with more malignant characteristics would be helpful for monitoring disease progression, planning image guided surgery, defining where tissue samples should be obtained in order to provide the most accurate diagnosis, and making decisions about changes in treatment. Diffusion weighted MR imaging provides estimates of the Apparent Diffusion Coefficient (ADC), which has previously been linked to survival2,3,4 and histopathological parameters5 that are important for tumor characterization.


The purpose of this study was to relate ADC and histological parameters from image guided tissue samples and to determine if there was an association between progression free survival (PFS) and pre-surgical ADC parameters for patients with recurrent low grade astrocytomas. The information obtained was used to create a tumor-level color map that highlighted more malignant regions of the tumor.


55 patients with an original diagnosis of WHO Grade II astrocytoma were recruited immediately prior to surgical resection for suspected recurrence. Ten patients recurred as Grade II, 28 as Grade III, and 17 as Grade IV. A total of 91 image guided tissues samples were acquired from these patients. Preoperative MR scans that included 6 directional axial Diffusion Weighted Imaging (DWI) with b=1000s/mm2 were conducted at either 1.5 or 3 Tesla. Target sample locations were selected using BrainLab navigation software based on surgically accessible areas within the lesion that were expected to correspond to tumor. Tissue samples were immediately sectioned and fixed in 4% formalin, dehydrated by graded ethanols, and embedded in wax using standardized techniques for histopathology analysis to evaluate tumor score (rated as 0 to 3), tumor cellularity and MIB-1- labeling index. Histograms and summary statistics of ADC values were generated for each subject from regions of interest corresponding to normal appearing white matter (NAWM), the T2 hyperintense lesion (T2ALL) and contrast enhancing lesion (CEL). The association of these summary statistics with PFS was assessed using a Cox proportional hazards analysis. To compare imaging with histopathology, 5mm spherical ROIs were generated at sample locations and the median ADC values determined. Correlation between imaging and histopathological parameters were determined by iteratively choosing one random biopsy per patient and performing a Kendall Tau correlation test, as well as a single uncontrolled Kendall Tau test. In order to generate color maps, the patients were stratified into 2 sub-groups (nADC- and nADC+) based upon whether they had nADC greater or less than the population median value and histograms of normalized ADC (nADC) values in the T2ALL lesion were averaged for each sub-group. An RGB colormap for each subject was created from nADC values based upon using intensities in the probability density distribution divided by their respective maxima from nADC-, nADC+ and NAWM histograms as the intensities in different color channels.


When adjusting for age and extent of resection, there was a significant association (P<0.01) between median nADC values in the T2ALL and PFS. For subpopulations that recurred as Grade II, III, and IV the median PFS values were 990, 911, and 329 days respectively. Pre-surgery nADC values at the sample locations were correlated with MIB-1 labeling index (p<.001 without considering the multiplicity of samples per patient, and with a median p-value of 0.04 when considered over 50 random selections of one vial per patient). The color maps (see Figure 1) were able to highlight regions with low and high ADC in a manner that was easier to visually interpret than grey scale images of ADC values.


The associations of ADC metrics with PFS and the relationship of MIB-1 index with pre-surgery ADC values in patients with recurrent grade 2 astrocytomas suggest that radiologists, oncologists, and surgeons should consider the spatial variation in ADC values in order to assess whether the lesion has transformed to a more malignant phenotype. Using RGB colormaps to emphasize regions with low and high ADC values is of interest for predicting malignant transformation, for directing surgical sampling, and for serially monitoring response to treatment.


This work was supported by the NIH Brain Tumor SPORE Grant P50CA097257.


[1] Grier, et al. (2006) Oncologist 11:681-693

[2] Wen, et al. (2014) J Neurooncol 121:331-339

[3] Poussaint, et al. (2015) Neuro Oncol. pii: nov256

[4] Wu, et al. (2012) J Chinese Med Assoc. 75(11):581-8.

[5] Zikou, et al. (2012) Clin Neurol Neurosurg. 114(6):607-12

Khayal, et al. Neuro-Oncology 13.11: 1192–1201.

Kono, et al. American Journal of Neuroradiology 22: 1081-1088


Colormaps for two patients of the same WHO Grade demonstrate their ability to highlight tumors with ADC values indicative of worse outcome.

The average tumor nADC histograms for both nADC+ and nADC- patients, as well as NAWM for all. There are 60 bins that span from a value of 0 to 3.5 (corresponding approximately to ADC values of 0 to 2800).

This table gives key information on the patient groups that went into the tumor-level nADC histogram averages. An nADC value of 1.67 was the population wide median of median T2ALL nADC values.

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