Original Contribution
Axial-Shear Strain Elastography for Breast Lesion Classification: Further Results From In Vivo Data

https://doi.org/10.1016/j.ultrasmedbio.2010.11.001Get rights and content

Abstract

The purpose of this work was to investigate the potential of the normalized axial-shear strain area (NASSA) feature, derived from axial-shear strain elastograms (ASSE), for breast lesion classification of fibroadenoma and cancer. This study consisted of previously acquired in vivo digital radiofrequency data of breast lesions. A total of 33 biopsy-proven malignant tumors and 30 fibroadenoma cases were included in the study, which involved three observers blinded to the original BIRADS-ultrasound scores. The observers outlined the lesions on the sonograms. The ASSEs were segmented and color-overlaid on the sonograms, and the NASSA feature from the ASSE was computed semi-automatically. Receiver operating characteristic (ROC) curves were then generated and the area under the curve (AUC) was calculated for each observer performance. A logistic regression classifier was built to compare the improvement in the AUC when using BIRADS scores plus NASSA values as opposed to BIRADS scores alone. BIRADS score ROC had an AUC of 0.89 (95% CI = 0.81 to 0.97). In comparison, the average of the AUC for all the three observers using ASSE feature alone was 0.84. However, the AUC increased to 0.94 (average of 3 observers) when BIRADS score and ASSE feature were combined. The results demonstrate that the NASSA feature derived from ASSE has the potential to improve BIRADS breast lesion classification of fibroadenoma and malignant tumors. (E-mail: [email protected])

Introduction

Ultrasound (US) elastography was introduced by Ophir et al. (1991) as a technique to image the stiffness variation in soft tissues. The technique involves acquiring US (radiofrequency [RF]/envelope) signals from an imaging plane before and after a small quasi-static compression. Typically, the pre- and postcompression frames are processed to generate images of local strain, commonly known as elastograms. When the elastogram depicts axial strain values, it is referred to as an axial strain elastogram (Ophir et al. 1999).

Based on the axial strain elastograms alone, elastography has been shown to be useful in a wide variety of clinical applications including breast lesion classification (Céspedes et al., 1993, Hiltawsky et al., 2001). The use of elastography for the reduction of the rate of unnecessary breast biopsies has been demonstrated by several groups (Garra et al., 1997, Regner et al., 2006, Barr, 2006, Itoh et al., 2006, Cho et al., 2008). Most of these reports have used the size discrepancy between sonographic and elastographic lesion appearance, along with/without strain contrast measures (Garra et al., 1997, Regner et al., 2006, Barr, 2006, Burnside et al., 2007). Some of them have used a scoring system based on strain distribution patterns (Itoh et al., 2006, Cho et al., 2008).

However, these measures from axial strain elastograms exploit only the size discrepancy between the sonographic and elastographic images of the tumor, which exists only in malignant cases. Nevertheless, it is also well documented in the literature that benign fibroadenomas are typically mobile (implying loosely-bonded to the host tissue) compared with malignant tumors, which tend to be well adhered (firmly bonded) to the host tissue (Fry, 1954, Ueno et al., 1988, Bamber et al., 1988). Therefore, it is reasonable to hypothesize that additional information regarding the bonding conditions near lesion boundaries has the potential to improve the performance of the current standard of practice in breast lesion classification by US.

Axial strain is one of the nine strain tensors that describe deformation in 3-D. We have shown that in addition to axial strain, it may be feasible to image another strain tensor in the form of axial-shear strain (ThitaiKumar et al. 2007). Note that the total shear strain, as defined in eqn (1), is the sum of the axial-shear (first-term) and lateral-shear strain component.ɛx,y=(vx+uy),where (u, v) are the lateral and axial displacement components along the x- (lateral direction) and y- (axial direction) axes, respectively.

The ultrasonic estimation of the lateral-shear strain is noisy compared with the estimation of the axial-shear component (ThitaiKumar et al. 2005). This fundamental limitation as a result of suboptimal sampling between US array beams also affects traditional elastography and sonography. However, we have shown that imaging the axial-shear component alone results in quality images of an independent constitutive tissue parameter that relates directly to shear strain. The image depicting the axial-shear strain was referred to as axial-shear strain elastogram (ASSE). We demonstrated recently that the axial-shear strain distribution pattern around an inclusion is directly influenced by the bonding at the inclusion-background boundary using simulations, gelatin-phantom experiments and breast lesions in vivo (ThitaiKumar et al. 2007). The normalized axial-shear strain area (NASSA) near the inclusion-background boundary is a feature that could be used to identify the boundary-bonding conditions (ThitaiKumar et al., 2007, Thitaikumar et al., 2008). Results from the initial feasibility study that evaluated the potential of this feature to differentiate between benign and malignant tumors in the breast are encouraging (ThitaiKumar et al. 2008). However, the initial feasibility study was restricted to a small number of in vivo cases (n = 21) that precluded a detailed statistical analysis.

In this paper, we report on a larger follow-up study done to investigate the potential of the NASSA feature to classify breast lesions into fibroadenoma and cancer, and therefore reduce unnecessary benign breast biopsies. We include the results of a statistical analysis performed to assess the improvement in the BIRADS-based breast lesion classification as a result of the addition of NASSA.

Section snippets

In vivo data

We used in vivo digital RF data of breast lesions that were acquired at the University of Vermont by Dr. Garra’s group (1997). The patient study was HIPAA-compliant and had appropriate institutional review board approval. Informed consent was obtained from all participating patients, who were informed that the RF data collected would be used at a later time for the creation of elastograms. The sonograms used in this study were reconstructed from the selected RF frames. Patients with a BIRADS

Results

Table 1 summarizes the response from each observer who participated in the study. It can be seen that the number of sonograms on which the observers was able to outline the lesion (blinded to the ASSE) was on average only 72% of the 63 cases (second column, i.e., average of 44/63, 54/63 and 38/63). This percentage improved by an average 6.3% after looking at the ASSE (third column), demonstrating the use of ASSE in confirming the lesion boundary and thereby improving sonographic lesion

Discussion

The results from the study demonstrate that the recently introduced ASSE may have a role in aiding the US classification of benign vs. malignant breast lesions. More importantly, the results based on adding the NASSA feature to the established BIRADS scores show an improvement in classification performance, as is evident from an increase in the AUC from 0.88 to 0.94. In recent years, several reports have shown that using regular axial elastograms alongside BIRADS improves the breast lesion

Conclusions

We have shown that features from recently developed ASSE, which provides information regarding lesion/host bonding characteristics, may add to existing sonographic and elastographic feature sets that are used in routine clinical practice for tumor differentiation with US. ASSEs are complementary to and can be used alongside with other coregistered images like axial strain elastograms and sonograms, to better visualize and interpret tissue information in the region of interest. ASSEs can be

Acknowledgments

The data used in this study were acquired previously for projects supported by NIH Program Projects grants P01-CA64597 and P01-EB02105. The current work was supported by NIH grant R21-CA135580. The authors would also like to thank the observers Elaine Khalil, M.D., Karen Ophir, B.S., R.D.M.S., Charlene Waldron, R.T., R.D.M.S. and Rhoda Reading, B.S., R.D.M.S., R.V.T., R.T., who volunteered their time to participate in the study.

References (22)

  • A. Itoh et al.

    Breast disease: Clinical application of US elastography for diagnosis

    Radiology

    (2006)
  • Cited by (26)

    • A Novel Elastographic Frame Quality Indicator and its use in Automatic Representative-Frame Selection from a Cine Loop

      2017, Ultrasound in Medicine and Biology
      Citation Excerpt :

      The magnitude and spatial variation of these shear strains depend mostly on the degree of bonding at the inclusion–background interface, the orientation of the inclusion–background interface with respect to the direction of compression, the stiffness contrast between the inclusion and the background and the level of axial compression applied (Knight et al. 2002; Twiss and Moores 1992). It was reported in a series of publications that even by imaging only the axial-shear strain component of the total shear strain, information related to the degree of bonding between the inclusion and the background can be reliably obtained (Thitaikumar et al. 2007a, 2007b; Thittai et al. 2011, 2012). The image of the axial-shear strain distribution, defined as the axial-shear strain elastogram (ASSE), which is computed as a lateral derivative of axial displacement (Thitaikumar et al. 2007b), has been reported to be useful for in vivo breast lesion classification into benign and malignant (Thitaikumar et al. 2008; Thittai et al. 2011).

    • An analysis of the segmentation threshold used in axial-shear strain elastography

      2015, Ultrasonics
      Citation Excerpt :

      Xu et al. [29] reported using a segmentation threshold of 20% of the applied axial strain in simulations, but adjusted this empirical value based on an estimate of the lesion-background strain contrast from ASE to account for the differences in stiffness contrast among the in vivo breast lesion cases. Recently [23], an observer study on a database containing larger number of in vivo cases was reported that used a segmentation threshold value of 50% of the peak axial–shear strain value. The study also reported the finding that displaying ASSE as a color-overlay was preferred by the observers consisting of radiologist and sonographers.

    • Elastic modulus of the prostate: A new non-invasive feature to diagnose bladder outlet obstruction in patients with benign prostatic hyperplasia

      2014, Ultrasound in Medicine and Biology
      Citation Excerpt :

      Stiffness or elasticity is quantified by the elastic modulus. Shear wave sonoelastography (SWE), a promising, newly developed sonoelastography technique, has been successfully used to determine elastic moduli of tissues in different organs to facilitate differential diagnosis (Barr et al. 2012; Berg et al. 2012; Ferraioli et al. 2012; Thittai et al. 2010, 2011). Thus, SWE may provide a new strategy for the non-invasive diagnosis of BOO.

    View all citing articles on Scopus
    View full text