Monitoring the Response to Neoadjuvant Chemotherapy in Patients with Breast Cancer Using Ultrasound Scattering Coefficient: A Preliminary Report

Neoadjuvant chemotherapy (NAC) was initially used in locally advanced breast cancer (LABC) and the inflammatory form of cancer. Currently, it is recommended for patients with Stage 2 or 3, HER-2 positive or triple-negative (TNBC) breast cancer [1,2]. NAC reduces the tumor mass by inducing intracellular damage that causes cell death and degeneration. Response to treatment increases breast-conserving surgery, reduces recurrence and the risk of metastases and micro metastases [3]. However, the response to NAC treatment is variable, and assessment is necessary to distinguish between responders and non-responders and, if necessary, modify treatment. The response is defined and classified on the basis of changes in cancer cellularity and is divided into two categories, pathological partial response (pPR) subdivided into G1 (<9% reduction), G2 (10-29%), G3 (30-90%), G4 (>90%) and pathological complete response (pCR).The rate of responding and non-responding patients varies: pCR is seen (depending on the results presented in the literature) in 10%-31% of patients; pPR is seen in 69% to 100% of patients [4-6].Currently, in the monitoring of patients treated with NAC, a clinical breast examination(CBE), mammography (MMG), traditional B-mode ultrasound imaging (US), magnetic resonance imaging with contrast agent (CA-MRI), or diffusion-weighted (DW-MRI) can be used [7].


Introduction
Neoadjuvant chemotherapy (NAC) was initially used in locally advanced breast cancer (LABC) and the inflammatory form of cancer. Currently, it is recommended for patients with Stage 2 or 3, HER-2 positive or triple-negative (TNBC) breast cancer [1,2]. NAC reduces the tumor mass by inducing intracellular damage that causes cell death and degeneration. Response to treatment increases breast-conserving surgery, reduces recurrence and the risk of metastases and micro metastases [3]. However, the response to NAC treatment is variable, and assessment is necessary to distinguish between responders and non-responders and, if necessary, modify treatment. The response is defined and classified on the basis of changes in cancer cellularity and is divided into two categories, pathological partial response (pPR) subdivided into G1 (<9% reduction), G2 (10-29%), G3 (30-90%), G4 (>90%) and pathological complete response (pCR).The rate of responding and non-responding patients varies: pCR is seen (depending on the results presented in the literature) in 10%-31% of patients; pPR is seen in 69% to 100% of patients [4][5][6].Currently, in the monitoring of patients treated with NAC, a clinical breast examination(CBE), mammography (MMG), traditional B-mode ultrasound imaging (US), magnetic resonance imaging with contrast agent (CA-MRI), or diffusion-weighted (DW-MRI) can be used [7].
CBE has been shown to be a subjective technique with limited efficacy, especially in tumors smaller than 2cm: about 60% of residual breast cancers are undetectable by CBE [8]. Ultrasound is considered a more accurate method in assessing tumor size and monitoring residual breast tumors compared to CBE or MMG [9].
The literature analyzing the usefulness of an ultrasound examination together with MMG describes an increase in the probability of pCR prediction by up to 80% [10]. A promising method for observing regression of the disease is contrasted ultrasound, but this is not a commonly used test, and is not recommended by EFSUMB [11]. Ultrasound sonoelastography is another ultrasound technique used to monitor effectiveness of a treatment. The first published results suggest that a reduction in tumor stiffness observed by share wave elastography (SWE) allows prediction of disease response to NAC [12,13]. Evans et al. [12] have demonstrated using SWE, that the decrease in breast cancer stiffness, evaluated after the 3rd course of NAC, was a better predictor of pCR (sensitivity 59%, specificity 85%), compared to the assessment of lesion reduction using magnetic resonance (50% sensitivity, specificity 79%). Among the available radiological methods, the monitoring of tumor response during NAC using magnetic resonance imaging (MRI) is more accurate in comparison to CBE, US, or MMG; however, availability of MRI is limited, and underestimation of residual disease may affect up to 20% of patients [14,15]. The difficulties and limitations of radiological methods may be due to the fact that there is a delay in changes to tumor size and architecture, in contrast to cell death which starts several hours or days after the start of the treatment [16].
of the stroma with increased vascularization and infiltration of inflammatory cells [18]. For non-responding patients (pPR G1), tumor cells remain almost unchanged. However, in patients with pPR G2, G3, G4, there is a change in the percentage of enlarged, multinucleated, neoplastic cells [18,19]. New diagnostic tools are sought that will differentiate patients who respond to treatment from non-responders to NAC with high accuracy and at an early stage of treatment. In the literature, there are publications on the usefulness of quantitative ultrasound (QUS) techniques for monitoring reactions to NAC [19][20][21][22]. In the case of a classic B-mode examination, the image is created on the basis of an envelope of radio-frequency (RF) signals, which is subjected to intensive downstream processing. Filtering, log-compression and interpolation algorithms are used to reduce noise and reveal the edge of the tissues displayed on the ultrasound scanner screen. These procedures reduced the amount of information about the tissues examined, which can be obtained from the analysis of the original RF signal.
QUS techniques use raw RF data resulting from the interference of backscattered waves on elements included in the examined tissues, such as neoplastic cell clusters or elements of fibrous stromal tissue in the mammary gland. The theory of acoustic scattering in relation to tissue biology and the assessment of the usefulness of various ultrasound techniques for the study of cellular density was analyzed, among others, by Oelze et al. [23]. Since then, many QUS methods have been developed based on the analysis of the scattered echo, aimed at characterizing the tissue microstructure and elastic properties of the tissues. Ultrasonic parameters determined from the scattered echo that can characterize the microstructure of tissues include echo envelope statistical parameters [24,25], texture parameters [26], and scatter parameters. The latter can effectively predict the response of tumor tissue to the treatment used [21,26]. The aim of our study was to evaluate patient responses to NAC using different ultrasound techniques, namely the assessment of tumor size in B-mode imaging, stiffness assessment in elastographic examination and using quantitative ultrasound parameter. As a quantitative measure, we applied the IBSC (integrated backscatter coefficient), whose value depends on the quantity, shape, organization, and size of the scattering elements.

Patients
The study protocol was approved by the institutional review board of the Maria Skłodowska-Curie Memorial Cancer Centre and Institute of Oncology, Warsaw, Poland. All patients gave their written consent to participate in the study. From April 2016 to November 2017, 10 patients aged 32 to 75 (mean age 52.9) with a total of 13 tumors (one bifocal lesion, one trifocal lesion) were qualified for NAC at Oncology Clinic. AT (Doxorubicin, Docetaxel) and AC (Doxorubicin, Cyclophosphamide) and Taxol were used in the treatment according to international guidelines. All patients underwent a simple mastectomy with lymphadenectomy.

Histology
All patients underwent core needle biopsies (CNB) after administration of 2% lidocaine, using a biopsy gun needle (14G diameter -Pro-Mag). Three to five cores were taken from each lesion. After surgery (simple mastectomy),surgical specimens were immediately fixed in 10%buffered formalin. Representative sections from these samples were processed and routinely stained for H&E for histopathological (microscopic) examinations (HE). All tumor sampleswere evaluated by the same pathologist. Based on the pathological assessment of breast tissue, grade of malignancy, cancer subtype, information on tumor response to treatment as cellularity, degree of cell damages, residual tumor size was obtained (Tables 1,2 & 3). In order to categorize the tumor pathological response to NAC, changes in cellularity of tumors were quantified using the samples obtained from the CNB before treatment and the material obtained after the treatment and surgery, using the Miller-Payne scale [27]. In histopathological examination after NAC, tumors were classified into two categories: pathological partial response (pPR) and pathological complete response (pCR). pPR was sub-divided into G1, G2, G3, or G4, according to the extent of the changes observed. The results of the ultrasonic analysis were referred to the pathological verification. CR (complete response), PR (partial response), SD (stable disease) .
* Assessment of elastography for the 2nd course of NAC ** lesion invisible from the week after the end of the 3rd course of NAC W-width, D-depth, L-length % reduction in the largest dimension of the radial plane lesion. NST-Nonspecific type, ER-estrogen receptor, PGR-progesterone receptor, HER-human epithelial growth factor AT-Doxorubicin, Docetaxel; AC-Doxorubicin, Cyclophosphamide, IHCH-immunohistochemically examination.

Registration of Ultrasonic Data
A total of 67 B-mode ultrasound examinations with breast sonoelastography and lymph node assessment were performed in the Department of Ultrasound, Institute of Fundamental. Technological Research Polish Academy of Science in Warsaw. B-mode images and the corresponding raw RF echoes were recorded using an ultrasound scanner (Ultrasonix Sonix Touch-Research, Ultrasonix Medical Corporation, Richmond, BC,Canada) and a linear array transducer L14-5/38, with the transmitted frequency set at 10MHz, which, as measured by a hydrophone, corresponded to pulses center frequency of approximately 6.5MHz. The tumor area region of interest (ROI) was determined on the B-mode image by an experienced radiologist. Each patient underwent at least five ultrasound examinations: baseline recordings were made before the start of treatment, with subsequent scans a week after each round of chemotherapy. During each examination, the data from the focal lesion were recorded from four cross-sections (radial, radial+45°, anti-radial, anti-radial+45°). The period of participating patient monitoring was 5-6 months. The assessment of focal lesions in the breast was based on the guidelines of American College of Radiology (BI-RADS lexicon) and the standards of the Polish Ultrasound Society [28,29]. The RECIST 1.1 classification was adapted to monitor the volume of tumors [30]. Tsukuba scale was used in the sonoelastographic assessment of tumors [31]. It is a 5-point scale of classification, from Tsukuba 1, when strain is presented in whole lesion, to Tsukuba 5, when no strain is measured in the lesion and surrounding tissue (see Figure 1b, 2b, 4b and 5b). In order to collect data for quantitative analysis for each B-mode scan, 510 RF signal lines were recorded, sampled at a frequency of 40MHz. The transducers focus was always sited in the middle of the lesion. The analysis of the collected data for IBSC determination was performed offline, using proprietary programs implemented in the Matlab environment (Mathwork Natick, MA, USA).

Quantitative Ultrasound Parameters
Integrated Backscatter Coefficient (IBSC): Analysis of RF signals to determine IBSC was carried out using the method proposed by Yao et al. [32]. The ROI (tumor area) was analyzed using a sliding window method (the window was moved horizontally with one step corresponding to one RF line and vertically with the step of one sample of the analyzed signal corresponding to 0.025 microseconds). Parametric maps showing changes occurring due to NAC were built on the basis of IBSC values found in subsequent windows in the tumor area. Higher values of IBSC are represented as red and lower values as blue (Figures 8 & 9). In order to determine a single IBSC value characterizing the entire lesion, IBSC values obtained for all windows in a given section of tumor were averaged; the mean of four sections was then used.

B-Mode Ultrasound Imaging
In this study the 10 patients underwent a simple mastectomy with lymphadenectomy after chemotherapy. The mean tumor size by the B-mode examination before the treatment was 26mm (range 3-41mm), and after the treatment 8mm (range 2-34mm). The largest dimension of lesions in the radial plane (according to RECIST 1.1) was 41mm (4-41mm) before treatment, and 16mm (3-16mm) after treatment. The dimensions of neoplastic tumors decreased, meeting the RECIST criteria, Partial Response (PR) was observed in 10/13 cases, in the histopathological examination (he): pCR, G1, G3, G4), while in 2/13 of the patients Stable Disease (SD) was demonstrated (in the he pCR and G1was observed). In one of the patients in the ultrasound examination, a complete response (CR) was found (in the he pCR was observed). The results of ultrasound examinations (B-mode tumor size and elastography, before and after NAC) and tumor size after pathological examinations are presented in (Table 1). Elastography revealed 5/13 tumors that were designated as not deformable. These consisted of 3 tumors that did not respond to treatment (pPR G1) characterized by high stiffness (T4 and T5), one tumor that was pCR, and 1 tumor that was pPR G4. In 4/13 cases a decrease in stiffness was observed (in HE pPR lesions: G3 and G4). In 2 tumors, an increase in stiffness was observed; these were a pCR tumor and a pPR G4 tumor. In 7/13 tumors the stiffness did not change (in he pPR lesions: G1, G3, G4 and pCR was observed).

Results of Histopathological Verification
Histopathological examination after NAC and the surgery, revealed 4 pCR tumors, the remaining ones were pPR (where in 3 tumor G1 was observed). Lesions were verified as Invasive Carcinoma Non-Specified Type (IC NST) G2 and G3 (Table 2). (Table 3) presents the reduction percentage of cancer cells and histopathological verification performed after NAC. The images from microscopic verification and the corresponding images from the B-mode examination, before and after treatment, for two patients with extremely different response to the treatment (1 and 9) are shown in (Figure  1-6) (a, b) below. Patient 1 responded well to NAC (pCR) treatment. Patient 9 did not respond to the treatment (G1).

Quantitative Backscatter Ultrasound Results
Ten tumors which were classified by histopathological examination after NAC as pathological complete response (pCR) and pathological partial response (G2, G3, and G4) were characterized by an increase in the IBSC value of 153% (range 48% to 287%). Three tumors that were designated pPR G1 (<9% reduction) in cancer cellularity) displayed a decrease in IBSC (mean 31%: range -20% to -60%). (Figure 7) shows the percentage change of IBSC value after NAC. The IBSC value for each lesion prior to the start of treatment was used as the reference value. Damage in the tumor microstructure after each round of chemotherapy was visualized using the IBSC parametric maps. Examples of the maps developed for two patients (1 and 9), whose B-mode and microscopy images are presented in (Figure 1-6), reacting in different ways to the treatment (pCR and G1) are shown in (Figure 8,9).

Discussion
Analysis of B-mode ultrasound images of 13 breast tumors using an adapted RECIST1.1 methodology indicated a partial response in 10, a complete response in 1 tumor, and stable disease in 2 tumor. This assessment, however, did not correlate with the outcome or the dimensions of tumors and cellularity obtained in the final histopathological verification. In tumors with Partial Response and Stable Disease, pathological response as G1, but also with pCR, G3, G4 were observed. It is worth noting that the dimensions of tumors obtained in the final histopathological verification were much larger than the dimensions determined on the basis of B-mode imaging. In our study, there was no correlation of changes in the stiffness of tumors (sonoelastography) to other methods used to demonstrate a response to treatment (QUS or histopathology). The studies published so far have shown that both sonoelastography techniques (relative strains and SWE) are useful in predicting responses to NAC. Ma Y et al. showed in a group of 71 patients that the decrease in E mean and SR (Strain Ratio) after courses of chemotherapy predicted with high accuracy (AUC = 0.93 and 0.90, respectively) the response to treatment [33].
They also showed that tumors with lower initial stiffness correlated with a favorable response after the NAC (pCR). Similar results were obtained by Evans et al. [13]. Detailed analysis of pCR tumors using sonoelastography in our study showed no changes or decreased stiffness. In the histopathological evaluation of these pCR lesions, all patients showed a significant decrease in tumor cellularity compared to CNB (core biopsy). The samples appeared to contain fibrosis and stromal elastosis, which could increase the rigidity of the tumors, but this has not been demonstrated in our samples. Different results were obtained for the assessment of tumors using QUS, including tumors with pCR. There were large differences in the IBSC values between the group of tumors which responded to treatment (pCR, pPR G2, G3, and G4) and the group that did not responded to the treatment (pPR G1). In the first group, there was an increase in the IBSC value: in the second group, there was a slight decrease. Similar results were published by Sannachi et al.: 30 patients with LABC showed no changes in IBSC values in a group of non-responders. The authors, however, defined non-response as no significant differences in the microscopic assessment of cellularity of lesions and less than 50% reduction in tumor size [21].
Bearing in mind our results and reports from the literature, we question what constitutes the main source of scattering in breast tumors being examined. Czarnota et al. hypothesized that it is possible to observe the defragmentation of cell nuclei during apoptosis using high ultrasonic frequencies (>20MHz) [34]. The influence of cell properties on ultrasound scattering has been confirmed, among others by Czarnota and Kolios, who demonstrated an in vitro correlation of apoptotic cell nuclei size with the scattering intensity indicated by IBSC [35,36]. Nevertheless, the hypothesis of an observed increase in ultrasound scattering as a result of defragmentation of cell nuclei cannot be directly extrapolated to our in vivo studies of breast cancer. The length of ultrasonic waves at 6.5MHz used in our work is relatively large, and thus the wave does not interact directly with individual cells, but with larger tissue structures, such as cell clusters or stromal fibroid tissue. The relationship between the change of IBSC and the changes occurring in tissue structure should therefore be considered at the level of reconstruction of whole cell clusters or remodeling of the stroma during chemotherapy.
The interpretation of the alteration in ultrasound backscattered parameters is problematic. Spatial distribution of tissue types and their effect on ultrasonic scattering are not wholly understood. In addition, the results of histopathological examination only provide information about the condition of a part of the tumor tissue before NAC (taken by biopsy) and after the entire treatment cycle and may not be reflective of the changes taking place after each of course of NAC.

Conclusions
Early prediction of response in patients with breast cancer is crucial for planning surgery and assessing the effectiveness of the therapy. This is the first report comparing ultrasound examinations in B-mode for tumor size assessment, sonoelastography, and quantitative ultrasonography with histopathological analysis in patients undergoing neoadjuvant chemotherapy. Preliminary results on a group of 10 female patients with the presence of 13 breast cancer tumors confirmed the relationship between the result of post-operative histopathological verification and changes in the IBSC value. In the study group of patients, there were unequivocal differences in IBSC values between patients with pCR, pPR tumors (G2, G3, G4) and patients without changes in cellularity after treatment (pPR G1). The results of the response were not as unambiguous in the assessment by tumor size in the B-mode study and deformability changes using sonoelastography. The results obtained for the IBSC parameter were presented using parametric maps. Such maps enable visualization of changes taking place in tumors after chemotherapy and may be a valuable supplement to other imaging techniques used in diagnostic imaging. The QUS technique, and in particular the IBSC parameter may supplement the methods of assessing the effectiveness of treatment with NAC. In order to determine whether IBSC extends clinical benefit in the NAC assessment, further observations are warranted using a larger cohort of patients.