Abstract P4-03-04: Computer extracted image measurements of nuclear shape and texture from H&E images appear to stratify low and high risk ER+ breast cancers assessed via oncotype DX

Background: In this study we investigate the ability of computer extracted image features (nuclear morphology and texture) from digitized H&E tissue slides to stratify women with lymph node negative (LN-), estrogen receptor positive (ER+) breast cancer (BCa) as low or high risk as determined by Oncotype DX (ODX), a 21 gene-expression assay. Each year, over 120,000 women in the United States (1 million worldwide) are diagnosed with ER+ BCa. Treatment guidelines recommend hormone therapy (HT) plus chemotherapy (CT); however, up to 85% of ER+ BCa patients will not benefit from CT, yet will still suffer its side effects. ODX yields a numeric risk score (RS) ranging from 1-100; RS<18 suggests patients will respond to HT alone while RS>30 indicates need for adjuvant CT. Unfortunately, this test is expensive (>$4000), time-consuming, and involves destructive tissue testing. The goal of this study is to show that quantitative features calculated from H&E images can accurately predict risk stratification as determined by ODX in women with LN-, ER+ BCa, suggesting a histologic image based classifier could serve as a low-cost alternative.

Methods: Digitized H&E-stained ER+ BCa tissue sampled from 111 patients (34 high and 77 low-risk as determined by ODX) were obtained from the University of Pennsylvania, the University of Medicine and Dentistry of NJ, and Case Western Reserve University. Regions of cancer were annotated manually by an expert pathologist, and representative fields of view (FOV) were chosen at 20x magnification (2000 by 2000 pixels) for each patient. A selection of nuclear boundaries was annotated manually in each FOV. For each nucleus, a set of 2343 features was extracted, including 21 morphological (size, shape, and boundary) and 2322 texture (Gabor, Local Binary Pattern, Greylevel, and Laws filter features). Using Minimum Redundancy Maximum Relevance (mRMR) feature selection, the 3 features best able to separate low and high ODX risk categories were identified and used to build a supervised Bayesian classifier. Classifier training employed a randomized 3-fold cross-validation scheme; in each trial, two-thirds of the dataset were randomly selected for training, and the remaining one-third employed for independent testing. Classifier performance was evaluated using area under the receiver operating characteristic curve (AUC), positive predictive value (PPV), and negative predictive value (NPV) with respect to low and high ODX risk categorization. Performance metrics were averaged over 100 trials of 3-fold cross-validation (see table).

Results: The mRMR method selected one morphological feature (nuclear area) and two Laws-based texture features as being highly discriminating between risk categories. The Bayesian classifier trained with these 3 features yielded high AUC, PPV, and NPV measures with low variance in distinguishing ODX risk categories. The supervised classification results indicate that quantitative image features from H&E-stained histopathology are able to accurately discriminate between low and high risk patients as determined by ODX.

Classification PerformancePerformance MetricAverage (100 Trials)Standard DeviationAUC0.870.018PPV0.810.039NPV0.880.017

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P4-03-04.

Abstract P2-02-12: Computer derived image features on DCE-MRI appear to distinguish estrogen receptor-positive breast cancers with low and high oncotype DX recurrence scores

The Oncotype DX (ODX) is a 21 panel gene-expression based assay for identifying which Estrogen Receptor-positive (ER+) breast cancer (BCa) patients are candidates for adjuvant chemotherapy. The objective of this research was to identify whether computerized texture features on a staging DCE-MRI can distinguish ER+ BCa with low and high ODX recurrence scores (RS) (i.e. to distinguish which ER+ BCa patients are more likely to benefit from adjuvant hormonal therapy from those who require chemotherapy). This would provide a non-invasive, imaging based, pre-therapeutic assessment tool for predicting the appropriate treatment regimen. This work, to the best of our knowledge, is the first attempt to quantitatively correlate low versus high risk stratification via computer derived MRI measurements to corresponding risk stratification via the ODX assay.

52 ER+ BCa patient studies with high (>30, N = 28) and low (<18, N = 24) ODX RS were available for this study from two sites; 16 breast MRIs from the Boston Medical Center using a Phillips 1.5T magnet with a 7-channel breast coil, and 36 MRIs from the Case Medical Center using a Siemens 1.5T magnet with a 8-channel breast coil. All datasets included T1w images obtained prior to, during, and after administration of 0.1 mmol/kg of Gd-DTPA and corresponding ODX RS. For each study a radiologist picked a representative slice showing the tumor and then manually segmented the region of interest (ROI) containing the lesion. Computerized image analysis tools developed in-house via the MATLAB© programming platform were applied to the manually segmented lesion ROI for each of the 52 MRI studies to quantitatively characterize the lesion via a set of (a) 6 shape, (b) 3 pharmacokinetic (Ktrans, ve, kep) based on Tofts model (PK), (c) 12 enhancement kinetic (EK), (d) 12 intensity kinetic (IK), (e) 312 textural kinetic (TK), (f) 6 dynamic local binary pattern (DLBP), and (g) 5 dynamic histogram of oriented gradient (DHoG) features. The computer extracted features were evaluated via a linear discriminant analysis (LDA) classifier in terms of their ability to distinguish ER+ BCa as having a low or high ODX RS via a 2-fold randomized cross validation scheme. At each iteration, half of the studies were randomly selected from the 52 cases and used for training the LDA classifier and the remaining 26 studies were used for independent testing. This process was repeated 200 times. Classification performance was evaluated by area under the ROC curve (AUC). Higher AUC values suggest a stronger relationship between risk stratification via MRI attributes and ODX.

Table 1Feature classAccuracy (μ±Δ)AUC (μ±Δ)DHoG87.07%±5.66%0.89±0.04DLBP85.86%±7.82%0.83±0.07EK82.36%±8.46%0.80±0.06PK81.14%±7.55%0.78±0.07TK75.93%±6.65%0.76±0.08IK76.43%±7.23%0.75±0.12Shape71.04%±6.81%0.70±0.06

Table 1 illustrates the mean and standard deviation in accuracy and AUC values over 200 runs of randomized cross validation. DHoG, DBLP and EK features yielded the highest classification accuracy and AUC. Although lesion shape has been shown to be important for discriminating benign and malignant lesions on MRI, shape appears to be less useful in distinguishing between ER+ BCa lesions with low and high ODX RS.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P2-02-12.

Abstract P2-03-01: Computer extracted image texture features on T2-weighted MRI appear to correlate with nuclear morphologic descriptors from H&E-stained histopathology in estrogen receptor positive breast cancers

Oncotype DX (ODX) is a 21 panel gene-expression based assay for predicting whether patients with estrogen receptor-positive (ER+) breast cancer (BCa) are candidates for adjuvant chemotherapy. However, the time and expense associated with genomic assays suggests the need for a non-invasive, imaging-based, pre-therapeutic tool for assessment of disease risk and selection of an appropriate treatment regimen. The objective of this research was to determine whether (a) computer extracted image features on T2-weighted (T2w) MRI and H&E stained histopathology are independently able to distinguish ER+ BCa with low and high ODX recurrence scores (RS) and (b) to determine whether there is a correlation between MRI and histologic features identified as being predictive of low and high ODX risk categories.

A total of 11 ER+ BCa patients were considered in this study, based on availability of in vivo 1.5 Tesla T2w MRI. For each study, the corresponding formalin-fixed paraffin-embedded H&E stained tissue specimens were digitized at 20x (0.5 μm/pixel) using a whole-slide scanner. Of the 11 patients, 8 were identified in the low ODX (RS < 18) and 3 in the high ODX (RS > 30) risk categories. Each dataset was accompanied by expert annotations of (a) the lesion ROI on MRI and (b) boundaries of epithelial nuclei from a representative field-of-view on the digitized histology slide.

For each MRI study, a multi-scale, multi-orientation Gabor filter bank was convolved with the annotated lesion area providing a set of 192 texture features (FMRI). For each corresponding histology image, 471 features (FHIST) were extracted describing both nuclear morphology (NM) and Laws texture (LT) within the nuclear regions. Independent 2-sample t-tests were used to identify salient features in FMRI and FHIST that are able to distinguish low and high ODX risk categories. We found that, for the MRI dataset, Gabor texture features at several scales and orientations yielded salient features (p < 0.05) while on histopathology, nuclear texture and convexity (shape) features were identified as the top discriminative features (p < 0.01). Relationships between significant features were evaluated via Spearman's rank correlation test (see table), where high correlations were observed between lesion texture on T2w MRI and nuclear texture and shape on histology.

Correlation of histologic and MRI features able to distinguish low and high ODX RSHistologic feature correlated with ODXMRI feature correlated with ODXCorrelation coefficient (ρ)p-valueLT: 70 Mean HSVGF: Scale 2: Orientation 3: min/max-0.85450.0008NM: ConvexityGF: Scale 5: Orientation 6: mean-0.85450.0008LT: 70 Mean HSVGF: Scale 2: Orientation 3: min/max-0.83640.0013LT: 70 Mean HSVGF: Scale 3: Orientation 8: mean-0.83640.0013LT: 70 Mean HSVGF: Scale 3: Orientation 2: mean-0.81820.0021

Our results suggest that quantitative features extracted on both T2w MRI and histopathology can independently distinguish between low and high risk ODX classes. Moreover, some of these MRI and histologic features appear to be significantly correlated, suggesting that information regarding tumor biology is reflected in both MRI and histologic image features.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P2-03-01.

Abstract P1-08-16: Deep sequencing of breast tumor biopsies reveals an association between pathologic complete response and reduction of TP53 clonal abundance upon brief exposure to therapy

Background: Next generation deep sequencing has revealed the existence of intra-tumor heterogeneity within subsets of breast tumors. The clinical implications of intra-tumor heterogeneity are not fully understood, however subclonal heterogeneity likely plays a role in treatment resistance. We quantify the clonal abundance of somatic mutations in breast tumor biopsies using deep targeted amplicon sequencing and assess their changes over the course of preoperative therapy (PT). We also evaluate the association of changes in clonal abundance upon brief exposure (BE) to therapy with clinical outcome.

Methods: DNA from 69 breast tumor samples obtained from BE preoperative clinical trials BrUOG 211A/211B were sequenced. Patients received a run-in dose of bevacizumab(B), nab-paclitaxel(N) or trastuzumab (T), followed by combination biologic/chemotherapy (HER2- with B/carboplatin/N; HER2+ with T/carboplatin/N). We sequenced biopsy pairs obtained pre/post 10 day exposure to run-in targeted therapy and germline and surgical tumor DNA for a subset of patients upon completion of PT. A TruSeqCustom Amplicon (Illumina) for targeted enrichment sequencing that included 1183 amplicons covering either hotspot regions or whole exonic regions from 35 commoly mutated genes in breast cancer (TCGA, Stephens, 2012; Shah); a total of 101,484 bp of the genome was represented. Sequencing was performed using IlluminaMiSeq platform and analyzed for variant calls using IlluminaBasespace. High confidence somatic mutations were identified in samples with matched germline data using VarScan2. In the absence of matched normal DNA, germline variants were eliminated using dbSNP and the 1000 Genomes Project. Minor allele frequencies (MAF) of somatic aberrations were estimated as the percentage of reads matching the variant.

Results: Approximately 5 mutations on average were found baseline and post-exposure, with a maximum mutational burden of 15 mutations in one basal breast cancer. Recurrent somatic aberrations were observed in TP53 (42%), PIK3CA (16%) and FAT4 (13%), whereas sporadic aberrations were also seen in COL1A1, PTEN, CDH1. More than 85% of samples harboring TP53 mutations exhibited MAF≥40%. Similar high clonal abundance (MAF >50%) was observed for FAT4 mutations whereas PIK3CA mutations exhibited only subclonal frequencies (MAF≤30%). We evaluated changes in clonal architecture upon BE to therapy by scoring for a change in MAF of at least 10% from baseline to post-exposure sample. We scored a total of 16 cases for clonal abundance changes in TP53 mutations upon exposure to therapy. We found 6 cases that exhibited ≥10% reduction in MAF, of which 4 achieved pCR (p = 0.03) and the remaining 2 achieved RCB I. This association was independent of therapy arm and BE regimen.

Conclusions: We found that a reduction in TP53 clonal abundance upon BE to PT is associated with clinical outcome. We are currently integrating whole genome copy-number profiles with the deep sequencing data to more accurately assess clonal architecture and changes upon exposure to therapy. Clonal changes upon BE to therapy may provide early readouts of therapy benefit and provide biological insights into mechanisms of action.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P1-08-16.

Abstract P4-01-02: Association of DCE-MRI texture features with molecular phenotypes and neoadjuvant therapy response in breast cancer

Purpose: MRI imaging phenotype features such as volume and morphology are used to characterize tumor heterogeneity and tumor response. Texture-based imaging features are important in lesion characterization but their relationship to molecular phenotypes and response is unclear. Molecular stratification of breast cancer into luminal, basal, ERBB2, and normal-like can be made based on gene expression profiles. We investigate how texture-based imaging features relate to tumor biology, genetic subtypes and neoadjuvant therapy response using MRI, histopathological and RNA-sequencing data.

Materials and Methods: Data from 74 Stage IIA to IIIB breast cancer patients enrolled in neo-adjuvant clinical trials NCT00617942 and NCT00723125 were retrospectively reviewed. We evaluated 37 gray-level co-occurrence matrix features (GLCM) on post-contrast T1 fat-suppressed images of 38 HER2− tumors and 35 HER2+ tumors. The texture features included angular second moment, contrast, correlation, first diagonal moment, entropy, regularity, roughness, line likeness and other statistical summaries. We also performed RNA-sequencing on 23 tumors and compared RNAseq-based PAM50 clustering with texture-based clustering. Patients with pCR and RCB class=I were determined to be responders and the rest were labeled non-responders. Wilcoxon signed rank test was used to compare luminal vs. basal, ER+ vs. ER− and PR+ vs. PR- tumors and determine the discriminative power of the texture features. We then performed hierarchical clustering on our patient data set based on the significant texture features and evaluated their association with subtypes, hormone receptor status and response. Statistical significance of clusters was determined by hypergeometric test.

Results: We found five MRI texture features to be significantly associated with tumor subtypes: first diagonal moment, contrast range, correlation range and entropy range (p < 0.05). These five features together differentiated basal and luminal PAM50 subtypes with p = 0.001. Our analysis also showed an association between texture features and tumor hormone status. ER− tumors clustered strongly (13 of 20 ER− cases clustered, p = 0.009) with the 23 significant ER-associated texture features. Similarly, the PR- tumors formed tight grouping (15 of 24 PR- cases clustered, p = 0.006) with 26 significant PR-associated texture features in HER2− patients. Interestingly, only two texture features, entropy range and regularity, distinguished between responders and non-responders (p = 0.04). These features will be further evaluated with DCE-MRI data capturing post brief exposure dynamics.

Conclusion: Our results indicate that certain texture features from DCE-MRI images do capture biological heterogeneity in tumors and can potentially complement standard clinical evaluations. Texture features have previously been assessed for diagnostic settings but to our knowledge this is the first study that shows association of texture features with breast cancer subtyping and neoadjuvant therapy response. We speculate that this could potentially impact clinical management decisions and therapy selection.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-01-02.