Standardization in Quantitative Imaging: A Multicenter Comparison of Radiomic Features from Different Software Packages on Digital Reference Objects and Patient Data Sets

Radiomic features are being increasingly studied for clinical applications. We aimed to assess the agreement among radiomic features when computed by several groups by using different software packages under very tightly controlled conditions, which included standardized feature definitions and common image data sets. Ten sites (9 from the NCI's Quantitative Imaging Network] positron emission tomography–computed tomography working group plus one site from outside that group) participated in this project. Nine common quantitative imaging features were selected for comparison including features that describe morphology, intensity, shape, and texture. The common image data sets were: three 3D digital reference objects (DROs) and 10 patient image scans from the Lung Image Database Consortium data set using a specific lesion in each scan. Each object (DRO or lesion) was accompanied by an already-defined volume of interest, from which the features were calculated. Feature values for each object (DRO or lesion) were reported. The coefficient of variation (CV), expressed as a percentage, was calculated across software packages for each feature on each object. Thirteen sets of results were obtained for the DROs and patient data sets. Five of the 9 features showed excellent agreement with CV < 1%; 1 feature had moderate agreement (CV < 10%), and 3 features had larger variations (CV ≥ 10%) even after attempts at harmonization of feature calculations. This work highlights the value of feature definition standardization as well as the need to further clarify definitions for some features.

Abstract P5-12-03: Withdrawn

This abstract was withdrawn by the authors.

Abstract PD1-05: Aspirin use is associated with lower mammographic density in a large screening cohort

Background: Current breast cancer prevention agents have substantial side effects and do not prevent estrogen receptor negative (ER-) breast cancer. Aspirin is a promising breast cancer prevention therapy; it is cheap, safe, well tolerated, with strong biologic and epidemiologic evidence for a prevention effect on both ER- and ER+ breast cancers. However, clinical trials to date have failed to corroborate a prevention effect; these results are potentially related to study design (dose, duration of therapy and followup, population treated). We sought to evaluate the effect of aspirin on mammographic density, as breast density is a well-accepted, modifiable risk factor for both estrogen receptor positive (ER+) and ER- breast cancer. Methods: Electronic medical records from the University of Pennsylvania were retrospectively evaluated for women from a core set of 36 primary care/ObGyn practices. Individuals were selected if they had both undergone routine screening mammography during 2012-2013 and had an ambulatory visit within the year prior with a confirmed list of medication use. We selected the medication record closest to the screening exam. Logistic regression was performed to test for associations between clinically-recorded BIRADS breast density and aspirin use, after adjusting for the additional risk factors of age, body mass index (BMI) and ethnicity. Results: We identified 26000 women who fit the above criteria, of whom 19.7% reported current aspirin use and 41% were African American. Mean age was 57.3 (standard deviation [sd], 10.9) and mean BMI was 28.9 (sd, 7.3) kg/m2 for the entire cohort. Aspirin users were significantly older and had higher BMI (see Table). There was an independent, inverse association between aspirin use and mammographic density (Ptrend&lt;0.001). Compared to women with extremely dense breasts, women with fatty (OR=1.73, CI: 1.33-2.25) or scattered fibroglandular (OR=1.50; CI: 1.17-1.92) breasts were more likely to be aspirin users. A dose-response pattern was observed, as there was a lower likelihood of having extremely or heterogeneously dense breasts with increasing aspirin dose (OR=0.62, CI: 0.50-0.76 for &gt;300 mg; OR=0.84, CI=0.77-0.91 for &lt;=300 mg; compared to non-users as reference group). The association between aspirin use and density was more pronounced for women &lt;60 and for African American women (p=0.01). Conclusion: We demonstrate an independent association between aspirin use and lower mammographic density in a large, diverse screening cohort. Our results suggest that this association is stronger for younger and African American women: two groups at greater risk for ER- breast cancer. Future evaluation of this cohort will examine duration of aspirin use, and evaluate an automated measure of breast density. These results and others highlight the potential value and need for a randomized, controlled trial of aspirin as a preventive agent for breast cancer.

CharacteristicAspirin Non-UsersAspirin UsersPAge, mean (SD)55.3 (10.2)65.3 (9.8)&lt;0.0001BMI, mean (SD)28.5 (7.2)30.4 (7.6)&lt;0.0001Breast density, no. (%)  OR (95% CI)BIRADS 12006 (9.6)861 (16.9)1.73 (1.33 - 2.25)BIRADS 29346 (44.7)2859 (55.9)1.50 (1.17 - 1.92)BIRADS 38480 (40.6)1312 (25.7)1.22 (0.95 - 1.56)BIRADS 41057 (5.1)79 (1.6)1.00 (Reference)

Citation Format: Wood ME, Sprague BL, Oustimov A, Synnstvedt MB, Cuke M, Emily CF, Kontos D. Aspirin use is associated with lower mammographic density in a large screening cohort. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr PD1-05.

Abstract P2-02-04: Kinetic heterogeneity features on breast DCE-MRI as prognostic markers of breast cancer recurrence

PURPOSE

Breast tumors have been shown to be heterogeneous lesions, and tumor heterogeneity is one of the major difficulties in the way of an effective cancer diagnosis and successful treatment. We evaluate the predictive capacity of DCE-MRI kinetic heterogeneity features for assessing the risk of breast cancer recurrence as determined by a validated tumor gene expression assay.

METHOD AND MATERIALS

Bilateral Breast DCE-MRI scans were retrospectively analyzed from 57 women with estrogen receptor positive/node negative invasive breast cancer. The ages of the women at the time of the imaging ranged from 37 to 74 years with a mean age of 55.5 years. The women were imaged prone in a 1.5T scanner (GE LX echo, GE Healthcare, or Siemens Sonata, Siemens); matrix size: 512 × 512; slice thickness: 2.4-4.4 mm; flip angle: 25° or 30°. The images were collected before and after the administration of gadodiamide (Omniscan) or gadobenate dimeglumine (MultiHance) contrast agents. Dynamic contrast enhanced images were acquired at 90 second intervals for 3 post contrast time points. The women had previously undergone Oncotype Dx (Genomic Health Inc.) profiling of their tumor. The Oncotype DX assay provides the likelihood of 10-year breast cancer recurrence, using a score stratified into 3 risk categories (risk: low ≤17, medium = 18-30, high ≥ 31). Pixel-wise relative enhancement curves were computed using three post-contrast time points. Fuzzy C-means clustering was applied to partition the tumor pixels according to the variance of their relative enhancement. To capture kinetic heterogeneity, wavelet features were extracted within each tumor partition as a measure of spatial variation. Mean and variance of these features were further estimated within each region. Using these features, multivariable logistic regression was performed with leave-one-out cross-validation and feature selection to classify the tumors as high or low/medium risk. We compared our kinetic heterogeneity features against standard kinetics and texture features. Area under the curve (AUC) of the receiver operating characteristic (ROC) was used to evaluate classification performance.

RESULTS

Feature selection indicated an optimal set of 7 kinetic heterogeneity features (out of 54). The classifier based on these features had an AUC = 0.82 in classifying high versus low/medium risk tumors. Classifiers based on standard kinetics and texture features performed with AUCs of 0.69 and 0.64 respectively.

CONCLUSION

Wavelet kinetic features from breast DCE-MRI could be used to capture the spatial pattern of kinetic heterogeneity of and potentially serve as prognostic markers for the risk of recurrence. In addition, breast DCE-MRI kinetic heterogeneity features could be used to assess likelihood of recurrence and ultimately help guide therapeutic decisions. Larger studies are needed to validate these findings.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P2-02-04.

Abstract P4-01-08: Effect of bilateral salpingo-oophrectomy on breast MRI fibroglandular volume and background parenchymal enhancement for BRCA 1/2 mutation carriers

PURPOSE: Bilateral salpingo-oophrectomy (BSO) reduces breast cancer risk in women with BRCA1/2 mutations by approximately 50%. Annual screening for these high-risk women includes breast MRI and mammography. It is unknown whether BSO affects fibroglandular volume (FG) and background enhancement (BG) on breast MRI in this population. The purpose of this study is to assess the difference in FG and BG in BRCA1/2 mutation carriers before and after BSO on breast MRI.

METHOD AND MATERIALS: We compared FG and BG on contrast-enhanced breast MRI before and after BSO. Two readers blinded to both clinical and imaging history scored FG according to the BI-RADS criteria from 1–4: fatty, scattered, heterogeneously dense, or extremely dense. BG was graded based on the MRI BI-RADS scale from 1–4: minimal, mild, moderate, and marked. Average BIRADS scores of each reader were used to calculate mean FG and BG (±SD) before and after BSO, and compared using a paired t-test for significance.

RESULTS: We examined 60 women with BRCA1/2 mutations who underwent breast MRI before and after BSO from 2001–2011 at our institution. Five patients were excluded from the final analysis as they underwent mastectomy or bilateral radiation therapy before post-BSO MRI. Mean time to post-BSO MRI was 8.3 months ± 7 months. Mean pre- and post-operative FG were 2.64±0.78 and 2.58±0.75, respectively (p = 0.622). Mean pre- and post-operative BG were 2.46±0.93 and 1.87±0.81, respectively (p = 0.0001). Breast cancer was detected in 8 women at a median time of 3.4 years following BSO. In these 8 women, mean pre- and post-operative FG were 3.10±0.57 and 2.55±0.69, respectively. Mean pre- and post-operative BG were 3.20±0.67 and 2.35±0.71, respectively.

CONCLUSION: In this population of 55 BRCA1/2 mutation carriers who underwent breast MRI before and after BSO, there was a significant reduction in BG following BSO. There was no difference in FG on MRI following BSO. For the 8 patients in this cohort who were diagnosed with breast cancer after BSO, there was a trend toward higher FG and BG both before and after BSO, in comparison to the patients who were not diagnosed with breast cancer, suggesting that rates of higher FG and BG may be associated with increased risk for developing breast cancer.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-01-08.

Automatic Segmentation of Breast Carcinomas from DCE-MRI using a Statistical Learning Algorithm

Segmenting regions of high angiogenic activity corresponding to malignant tumors from DCE-MRI is a time-consuming task requiring processing of data in 4 dimensions. Quantitative analyses developed thus far are highly sensitive to external factors and are valid only under certain operating assumptions, which need not be valid for breast carcinomas. In this paper, we have developed a novel Statistical Learning Algorithm for Tumor Segmentation (SLATS) for automatically segmenting cancer from a region selected by the user on DCE-MRI. In this preliminary study, SLATS appears to demonstrate high accuracy (78%) and sensitivity (100%) in segmenting cancers from DCE-MRI when compared to segmentations performed by an expert radiologist. This may be a useful tool for delineating tumors for image-guided interventions.