Vascularity change and tumor response to neoadjuvant chemotherapy for advanced breast cancer

Machine Learning Models and Multiparametric Magnetic Resonance Imaging for the Prediction of Pathologic Response to Neoadjuvant Chemotherapy in Breast Cancer

Simple Summary

Achieving pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in breast cancer (BC) is crucial, as pCR is a surrogate marker for survival. However, only 10–30% of patients achieve it. It is therefore essential to develop tools that enable to accurately predict response. Recently, different studies have demonstrated the feasibility of applying machine learning approaches to non-invasively predict pCR before NAC administration from magnetic resonance imaging (MRI) data. Some of those models are based on radiomics, an emerging field that allows the automated extraction of clinically relevant information from radiologic images. However, the research is still at an early stage and further data are needed. Here, we determine whether the combination of imaging data (radiomic features and perfusion/diffusion imaging biomarkers) extracted from multiparametric MRIs and clinical variables can improve pCR prediction to NAC compared to models only including imaging or clinical data, potentially avoiding unnecessary treatment and delays to surgery.

Abstract

Background: Most breast cancer (BC) patients fail to achieve pathological complete response (pCR) after neoadjuvant chemotherapy (NAC). The aim of this study was to evaluate whether imaging features (perfusion/diffusion imaging biomarkers + radiomic features) extracted from pre-treatment multiparametric (mp)MRIs were able to predict, alone or in combination with clinical data, pCR to NAC. Methods: Patients with stage II-III BC receiving NAC and undergoing breast mpMRI were retrospectively evaluated. Imaging features were extracted from mpMRIs performed before NAC. Three different machine learning models based on imaging features, clinical data or imaging features + clinical data were trained to predict pCR. Confusion matrices and performance metrics were obtained to assess model performance. Statistical analyses were conducted to evaluate differences between responders and non-responders. Results: Fifty-eight patients (median [range] age, 52 [45–58] years) were included, of whom 12 showed pCR. The combined model improved pCR prediction compared to clinical and imaging models, yielding 91.5% of accuracy with no false positive cases and only 17% false negative results. Changes in different parameters between responders and non-responders suggested a possible increase in vascularity and reduced tumour heterogeneity in patients with pCR, with the percentile 25th of time-to-peak (TTP), a classical perfusion parameter, being able to discriminate both groups in a 75% of the cases. Conclusions: A combination of mpMRI-derived imaging features and clinical variables was able to successfully predict pCR to NAC. Specific patient profiles according to tumour vascularity and heterogeneity might explain pCR differences, where TTP could emerge as a putative surrogate marker for pCR.

Higher underestimation of tumour size post-neoadjuvant chemotherapy with breast magnetic resonance imaging (MRI)-A concordance comparison cohort analysis

Objectives

The aim of this study was to evaluate the diagnostic accuracy of breast MRI for detecting residual tumor and the tumor size whether it would be affected after neoadjuvant chemotherapy.

Methods

Total 109 patients with NAC and 682 patients without NAC were included in this retrospective study. Measurement of the largest diameter of tumors at pathology was chosen as gold standard and compared with preoperative breast MRI. A concordance threshold of ±25% of maximal tumor size was used. The accuracy of MRI was graded as concordant, underestimation, or overestimation rate. Further subgroup analysis with tumor stages, histologic subgroups and intrinsic subtypes was performed.

Results

The post-NAC MRI was associated with 92.5% sensitivity, 55.2% specificity, 85.1% positive predictive value, 72.7% negative predictive value, and overall 82.6% accuracy for detecting residual tumor. In determining tumor size, the overall concordance rates of the non-NAC group and the NAC group were 43.5% and 41.3%, respectively (p = 0.678). But the overestimation rate and underestimation rate were 26.6% and 32.1% for NAC group, and 52.9% and 3.5% for the non-NAC group (p<0.001). While in the subgroups analysis, the concordance rate of the NAC group (26.7%) was lower than that of the non-NAC group (82.1%) at T3 stage (p<0.001). There were no statistically significant differences between different tumor histologic subgroups and intrinsic subtypes.

Conclusions

The overall accuracy of MRI in predicting tumor size was not affected by NAC; however, it tends to underestimate tumor size after NAC, especially in patients with T3 lesions and above.

Dynamic Diffuse Optical Tomography for Monitoring Neoadjuvant Chemotherapy in Patients with Breast Cancer

Purpose To identify dynamic optical imaging features that associate with the degree of pathologic response in patients with breast cancer during neoadjuvant chemotherapy (NAC). Materials and Methods Of 40 patients with breast cancer who participated in a longitudinal study between June 2011 and March 2016, 34 completed the study. There were 13 patients who obtained a pathologic complete response (pCR) and 21 patients who did not obtain a pCR. Imaging data from six subjects were excluded from the study because either the patients dropped out of the study before it was finished or there was an instrumentation malfunction. Two weeks into the treatment regimen, three-dimensional images of both breasts during a breath hold were acquired by using dynamic diffuse optical tomography. Features from the breath-hold traces were used to distinguish between response groups. Receiver operating characteristic (ROC) curves and sensitivity analysis were used to determine the degree of association with 5-month treatment outcome. Results An ROC curve analysis showed that this method could identify patients with a pCR with a positive predictive value of 70.6% (12 of 17), a negative predictive value of 94.1% (16 of 17), a sensitivity of 92.3% (12 of 13), a specificity of 76.2% (16 of 21), and an area under the ROC curve of 0.85. Conclusion Several dynamic optical imaging features obtained within 2 weeks of NAC initiation were identified that showed statistically significant differences between patients with pCR and patients without pCR as determined 5 months after treatment initiation. If confirmed in a larger cohort prospective study, these dynamic imaging features may be used to predict treatment outcome as early as 2 weeks after treatment initiation. ^© RSNA, 2018 Online supplemental material is available for this article.

Predicting Treatment Response of Breast Cancer to Neoadjuvant Chemotherapy Using Ultrasound-Guided Diffuse Optical Tomography

PURPOSE

To prospectively investigate ultrasound-guided diffuse optical tomography (US-guided DOT) in predicting breast cancer response to neoadjuvant chemotherapy (NAC).

MATERIALS AND METHODS

Eighty-eight breast cancer patients, with a total of 93 lesions, were included in our study. Pre- and post-last chemotherapy, size and total hemoglobin concentration (THC) of each lesion were measured by conventional US and US-guided DOT 1 day before biopsy (time point t0, THC THC0, SIZE S0) and 1 to 2 days before surgery (time point tL, THCL, SL). The relative changes in THC and SIZE of lesions after the first and last NAC cycles were considered as the variables ΔTHC and ΔSIZE. Receiver operating characteristic curve was performed to calculate ΔTHC and ΔSIZE cutoff values to evaluate pathologic response of 93 breast cancers to NAC, which were then prospectively used to predicate response of 61 breast cancers to NAC.

RESULTS

The cutoff values of ΔTHC and ΔSIZE for evaluation of breast cancers NAC treatment response were 23.9% and 42.6%. At ΔTHC 23.9%, the predicted treatment response in 61 breast lesions for the time points t1 to t3 was calculated by area under the curve (AUC), which were AUC₁ 0.534 (P=.6668), AUC₂ 0.604 (P=.1893), and AUC₃ 0.674(P =. 0.027), respectively; for ΔSIZE 42.6%, at time points t1 to t3, AUC₁ 0.505 (P=.9121), AUC₂ 0.645 (P=.0115), and AUC₃ 0.719 (P=.0018).

CONCLUSION

US-guided DOT ΔTHC 23.9% and US ΔSIZE 42.6% can be used for the response evaluation and earlier prediction of the pathological response after three rounds of chemotherapy.

A computer-aided diagnosis (CAD) scheme for pretreatment prediction of pathological response to neoadjuvant therapy using dynamic contrast-enhanced MRI texture features

OBJECTIVE

To assess whether a computer-aided, diagnosis (CAD) system can predict pathological Complete Response (pCR) to neoadjuvant chemotherapy (NAC) prior to treatment using texture features.

METHODS

Response to treatment of 44 patients was defined according to the histopatology of resected tumour and extracted axillary nodes in two ways: (a) pCR+ (Smith's Grade = 5) vs pCR- (Smith's Grade < 5); (b) pCRN+ (pCR+ and absence of residual lymph node metastases) vs pCRN - . A CAD system was developed to: (i) segment the breasts; (ii) register the DCE-MRI sequence; (iii) detect the lesion and (iv) extract 27 3D texture features. The role of individual texture features, multiparametric models and Bayesian classifiers in predicting patients' response to NAC were evaluated.

RESULTS

A cross-validated Bayesian classifier fed with 6 features was able to predict pCR with a specificity of 72% and a sensitivity of 67%. Conversely, 2 features were used by the Bayesian classifier to predict pCRN, obtaining a sensitivity of 69% and a specificity of 61%.

CONCLUSION

A CAD scheme, that extracts texture features from an automatically segmented 3D mask of the tumour, could predict pathological response to NAC. Additional research should be performed to validate these promising results on a larger cohort of patients and using different classification strategies. Advances in knowledge: This is the first study assessing the role of an automatic CAD system in predicting the pathological response to NAC before treatment. Fully automatic methods represent the backbone of standardized analysis and may help in timely managing patients candidate to NAC.

Monitoring Neoadjuvant Chemotherapy for Breast Cancer by Using Three-dimensional Subharmonic Aided Pressure Estimation and Imaging with US Contrast Agents: Preliminary Experience

Purpose To determine whether three-dimensional subharmonic aided pressure estimation (SHAPE) and subharmonic imaging can help predict the response of breast cancer to neoadjuvant chemotherapy. Materials and Methods In this HIPAA-compliant prospective study, 17 women (age range, 45-70 years) scheduled to undergo neoadjuvant therapy for breast cancer underwent ultrasonography (US) immediately before therapy and at completion of 10%, 60%, and 100% of chemotherapy. All patients provided written informed consent. At each examination, radiofrequency data were collected from SHAPE and subharmonic imaging during infusion of a US contrast agent. Maximum-frequency magnitude and mean intensity were calculated for SHAPE and subharmonic imaging. The signal differences in the tumor relative to the surrounding area were compared with the final treatment response by using the Student t test. Results Four patients left the study, and data from two patients were discarded because of technical problems. Eight patients completed the entire imaging protocol, and an additional three patients dropped out after the imaging session at completion of 10% of chemotherapy as a result of disease progression (these patients were counted as nonresponders). Patients' imaging outcomes consisted of six responders (tumor volume reduction >90%) and five partial responders or nonresponders. The results at completion of 10% of therapy showed that the subharmonic signal increased more in the tumor than in the surrounding area for responders than in partial responders or nonresponders (mean ± standard deviation, 3.23 dB ± 1.41 vs -0.88 dB ± 1.46 [P = .001], respectively, for SHAPE and 1.32 dB ± 0.73 vs -0.82 dB ± 0.88 [P = .002], respectively, for subharmonic imaging). Moreover, three patients whose tumor measurements initially increased were correctly predicted to be responders with SHAPE and subharmonic imaging after completion of 10% of therapy. Conclusion SHAPE and subharmonic imaging have the potential to help predict response to neoadjuvant chemotherapy for breast cancer as early as completion of 10% of therapy, albeit on the basis of a small sample size. ^© RSNA, 2017 Online supplemental material is available for this article.

Diffusion weighted imaging evaluated the early therapy effect of tamoxifen in an MNU-induced mammary cancer rat model

Purpose

To assess the optimal time point of diffusion-weighted imaging (DWI) for early prognosis of breast cancer following tamoxifen therapy using a methylnitrosourea (MNU)-induced ER-positive breast-cancer model.

Methods

Two groups of Sprague-Dawley rats (n = 15 for group 1; n = 10 for group 2) were used. All animals (50 days old) were intravenously injected with MNU (50 mg/kg body weight) to induce ER-positive mammary tumors. When tumors were approximately 2 cm in diameter, DWI was performed on days 0, 3, and 7, and intratumoral apparent diffusion coefficient (ADC) values were measured. Therapy started on day 0 with tamoxifen (10 mg/kg diet) and continued for 4 weeks for group 1, but only 1 week for group 2, while tumor volume was measured by caliper twice weekly. All animals of group 2 were euthanized on day 7 after imaging, and Ki-67, TUNEL, ERα, and ERβ staining were performed on tumor tissue.

Results

DW images of MNU-induced mammary tumors were successfully obtained with minimal motion artifact. For group 1, ADC change for 3 days after therapy initiation (ADC_3D) was significantly correlated with tumor-volume change until day 11, but the significant correlation between ADC change for 7 days (ADC_7D) and the tumor-volume change was observed until day 18. Similarly, for group 2, either ADC_7D or ADC_3D was significantly correlated with the tumor-volume change, but the higher significance was observed for ADC_7D. Furthermore, ADC_7D was significantly correlated with apoptotic (TUNEL stained), proliferative (Ki-67 stained), and ERβ-positive cell densities, but ADC_3D was not significantly correlated with any of those.

Conclusions

ADC_7D might be a more reliable surrogate imaging biomarker than ADC_3D to assess effectiveness of tamoxifen therapy for ER-positive breast cancer, which may enable personalized treatment. The significant correlation between ADC_7D and ERβ-positive cell density suggests that ERβ may play an important role as a therapeutic indicator of tamoxifen.

Breast cancer: assessing response to neoadjuvant chemotherapy by using US-guided near-infrared tomography

PURPOSE

To assess initial breast tumor hemoglobin (Hb) content before the initiation of neoadjuvant chemotherapy, monitor the Hb changes at the end of each treatment cycle, and correlate these findings with tumor pathologic response.

MATERIALS AND METHODS

The HIPAA-compliant study protocol was approved by the institutional review boards of both institutions. Written informed consent was obtained from all patients. Patients who were eligible for neoadjuvant chemotherapy were recruited between December 2007 and May 2011, and their tumor Hb content was assessed by using a near-infrared imager coupled with an ultrasonography (US) system. Thirty-two women (mean age, 48 years; range, 32-82 years) were imaged before treatment, at the end of every treatment cycle, and before definitive surgery. The patients were graded in terms of their final pathologic response on the basis of the Miller-Payne system as nonresponders and partial responders (grades 1-3) and near-complete and complete responders (grades 4 and 5). Tumor vascularity was assessed from total Hb (tHb), oxygenated Hb (oxyHb), and deoxygenated Hb (deoxyHb) concentrations. Tumor vascularity changes during treatment were assessed from percentage tHb normalized to the pretreatment level. A two-sample two-sided t test was used to calculate the P value and to evaluate statistical significance between groups. Bonferroni-Holm correction was applied to obtain the corrected P value for multiple comparisons.

RESULTS

There were 20 Miller-Payne grade 1-3 tumors and 14 grade 4 or 5 tumors. Mean maximum pretreatment tHb, oxyHb, and deoxyHb levels were significantly higher in grade 4 and 5 tumors than in grade 1-3 tumors (P = .005, P = .008, and P = .017, respectively). The mean percentage tHb changes were significantly higher in grade 4 or 5 tumors than in grade 1-3 tumors at the end of treatment cycles 1-3 (P = .009 and corrected P = .009, P = .002 and corrected P = .004, and P < .001 and corrected P < .001, respectively).

DISCUSSION

These findings indicate that initial tumor Hb content is a strong predictor of final pathologic response. Additionally, the tHb changes during early treatment cycles can further predict final pathologic response.