Breast vibro-acoustography: initial experience in benign lesions

There are several different medical imaging modalities for the screening and diagnosis of breast cancer [1]-[3]. However, benign breast lesions are much more common than malignant lesions, and accurate diagnosis of these lesions is important for optimal care of the patient [4]. Benign breast disease (BBD) is a well-known, significant risk factor for breast cancer [5]. BBD is diagnosed when a woman has a breast biopsy for a palpable or imaging abnormality in her breast that results in benign findings.

Any imaging evaluation of the breast with high sensitivity may also be associated with increased false positive results that may leads to unnecessary (i.e. benign) biopsies, resulting in high cost as well as significant trauma and anxiety for the patients due to this invasive procedure and in some instances, they may choose unnecessary extensive varieties of surgeries such as mastectomy [6],[7]. Moreover, accurate identification of benign non-proliferative conditions such as cysts and fibroadenomas can reduce unwanted benign breast biopsies. It is, therefore, important to develop imaging tools with higher specificity to reduce false positive test results. Mammographic breast density (MBD), another known factor, reduces the sensitivity of mammograms [8]. Hence, there is an immense need for a noninvasive tool that can assess breast tissue characteristics while not influenced by MBD.

Conventional breast ultrasonography (US) is routinely used as an adjunct imaging tool to x-ray mammography for diagnosis of breast pathology; it improves sensitivity and has a considerable role in differentiation of cysts and solid nodules with higher specificity. US can also help to characterize solid breast masses [3],[9]-[14]. US features such as lesion shape, orientation to the skin line, lesion boundary, margin characteristics, echo-pattern, posterior acoustic appearance, and effects on the surrounding breast tissue are employed to reach a Breast Imaging and Reporting Data System (BI-RADS) assessment [15]. In BI-RADS, the lesions are categorized into 7 categories numbered 0 to 6. Categories 1 to 3 indicate being negative, benign findings, or probably benign, respectively. Categories 4 and 5 are suspicious for malignancy, and 6 refers to having known biopsy proven malignancy [15]. Abnormalities on screening mammography that require further evaluation are assessed as category 0; similarly, calcifications found on breast US are also in category 0 assessment [16]. US is usually used for the subsequent evaluation of BI-RADS category 0 mammograms [17].

Benign breast lesions have a characteristic sonographic appearance; cysts appear as well-circumscribed, round or oval anechoic or hypoechoic masses with unnoticeable walls and posterior acoustic enhancement [18]. Fibroadenomas typically appear as oval, well-circumscribed masses with an abrupt interface and homogeneous iso- or hypoechoic echo texture. Benign papilloma masses appear as solid masses within a dilated duct with a vascular feeding pedicle seen on color Doppler imaging [2],[9],[14]. There is, however, an overlap between benign and malignant lesion US characteristics leading to a significant number of false-positive cases, which are recommended for US-guided percutaneous biopsy [14]. An additional sonographic tool to help decrease the number of unnecessary biopsies can play an important role to reduce such a large number of unnecessary biopsies. Imaging modalities, particularly those that provide palpation-like information, can help to better diagnose and identify breast lesions.

Elasticity imaging is an emerging field of medical imaging that provides such information. Elasticity imaging consists of magnetic resonance elastography (MRE) [23],[24] which is expensive and not widely available [19],[20], conventional quasi-static ultrasound elastography [21]-[23], Acoustic Radiation Force Impulse (ARFI) imaging [24]-[26] and Shear Wave Elastography (SWE, also called SuperSonic Imaging (SSI)) [27]-[31]. ARFI and SWE use ultrasound radiation force to generate shear waves and quantify tissue elasticity from measured propagation speed of shear waves [32],[33]. The results of studies using ARFI and SWE for breast have been very promising.

Vibro-acoustography (VA) is an imaging modality that also provides palpation-like information [34]. VA is introduced as a complementary technique to improve sensitivity and specificity in clinical breast imaging [35]. Principles of VA have been described extensively [34],[36]-[39]. In VA, ultrasound is employed to produce a localized low-frequency force to vibrate the tissue. In technical terms, such a force is called “acoustic radiation force (ARF)”. The low-frequency ARF is generated by two intersecting continuous wave (CW) focused ultrasound beams at slightly different frequencies. This force, which acts as a point force, vibrates the object at a frequency equal to the difference between two US frequencies (typically in kHz range). The resulting vibrations produce an acoustic emission field that is detected by a sensitive microphone (or hydrophone). Harder tissues normally produce a significantly different acoustic emission compared to normal soft tissues. Conceptually, VA resembles palpation; i.e., detects tissue response to an exerted force on tissue. [34],[36]. However, VA benefits from a significant advantage of using a highly localized ARF, which leads to the possibility of assessing tissue properties on a small scale. As a result, VA can provide detailed information on tissue mechanics at high resolution [38],[39].

Compared to conventional US imaging, VA images are speckle free, thus, the images have high-contrast that allows detection of small structures [4],[40],[41]. Thus VA can be a complementary tool to the existing breast imaging tools. Compared to elastography techniques, VA uses dynamic acoustic radiation force in the range of 10s of kHz, which is much higher than the frequency used in quasi-static elastography, ARFI, and shear wave imaging (normally in 10s to 100s Hz range). VA mages represent the acoustic response of tissue, which is a complex function of several parameters, including the elasticity and viscosity. However, elasticity cannot be directly and quantitatively measured from this acoustic response.

Our preliminary studies demonstrated the abilities of VA imaging in various tissues [40],[42],[43] including in vivo human breast [35],[44]. In our previous studies [35],[44], we used a confocal VA system combined with mammography to image various breast abnormalities, [35]. Since VA is a new modality, characteristic of different types of lesions (benign or malignant) in VA are not necessarily known. A goal of this paper is to determine and present VA characteristics of benign breast lesions. Another goal of this paper is to evaluate the performance of such characteristics in identifying benign lesions in VA and compare the results and compare the results to those of mammography. To define the VA characteristics of benign breast masses, we adopted the features that are normally attributed to such masses in mammography. Such features, which are often defined in contrast to those of malignant masses, include morphological attributes of the mass and information related to calcifications. We will test the validity and performance of these characteristics in a reader-based study.

Women volunteers with abnormality in their clinical breast examination and/or mammography BI-RADS category 4 or less were included in this study. All patients underwent clinical mammography and breast US before participation in the study. VA imaging was done on all subjects. In total, 36 patients with benign lesions were evaluated; five were used for training and excluded from the evaluation. Also, two participants were excluded from the study because of accidental hardware failure that occurred during the study. A total of 29 patients, averaging 44 years old, with benign lesions were evaluated. The final diagnosis for 25 patients was based on histological results. In the remaining 4 patients, based on clinical imaging, the lesions were determined to be simple cysts. These cysts were aspirated with needle aspiration and disappeared completely under direct ultrasound visualization. Therefore, the final diagnosis includes six benign cysts, 15 fibroadenomas, three papillomas, and three of post-surgical scar tissue, and two focal atypical ductal hyperplasia with microcalcifications. The flow chart shown in Figure 2 demonstrates the results of VA image evaluation by three independent readers’ reviews in 29 patients with benign lesions and their final diagnosis. The lesion detection rate for each radiologist was 100% with (95% confidence interval (CI): 88% to 100%). While all 29 lesions were confirmed to be benign (25 biopsy proven and 4 cysts with aspiration), the primary clinical imaging review of reference image (mammography) indicated that four lesions were suspicious for malignancy, therefore, the correct classification as benign was 86% (95% CI: 69% to 95%). This uncertainty was also observed using the new imaging modality. Correct classification as benign by the reviewers was either 83% (95% CI: 65% to 92%) or 86% (95% CI: 69% to 95%) as shown in Figure 2. Readers 1 and 2 misclassified three of the lesions as suspicious and two as malignant. Reader 3 misclassified three of the lesions as suspicious and one as malignant. These misclassified cases are discussed in the following section.

The goal of this study was to primarily investigate the diagnostic accuracy of VA in detecting benign breast lesions. Various types of benign breast lesions including lipid cyst, benign cyst, benign fibroadenoma, and benign papilloma were studied with a laboratory-built VA system integrated with a mammography machine. The results of the study clearly show the efficacy of VA in detecting benign breast lesions. Also, the results show that the mammographic attributes of benign lesions can be adopted to identify such lesions in VA imaging.

We explored the VA characteristics of a wide range of benign lesions. Regular shape, well-defined lesion boundary, distinct or soft margin, and soft lobulation are the most benign characteristics we found on VA images. Cysts appeared on VA images as well-defined masses with a soft border as seen in Figures 7 and 8 and all reviewers could identify all cyst cases. Fibroadenomas, the most common solid breast masses that undergo breast biopsy, appeared on VA images as well-circumscribed ovoid or round masses with gentle lobulation, and in some cases were calcified as seen in US and mammography, which happens in degenerating fibroadenomas [49],[50].

Three cases of fibroadenomas and one case of papilloma that were unidentifiable by mammography were marked by our radiologist and detected by VA. Similarly seen in breast US [14], VA can help identify lesions that are obscured in mammography due to dense breast. VA also identified benign papilloma lesions as elongated nodules in a dilated duct; one papilloma was not identifiable on mammography but seen using VA. Diagnosis of papilloma is important and surgical excision should be considered for core needle biopsy proven benign papilloma - especially for lesions larger than 1.5 cm, regardless of imaging findings [41]. As expected, VA worked better than mammography in dense breast.

The cases that were misclassified by VA were also misidentified by mammography. Three of the misidentified cases were post-surgical scar tissue, as shown in Figure 9, presenting as a radial pattern distortion imitating spiculation, which is most seen in malignant tissue. It should be noted that mammographically only some of these changes can be differentiated based on morphological characteristics. Even though microcalcifications are often associated with breast carcinoma, not all spiculated lesions including microcalcifications are considered malignant. Mammography alone is often not a reliable imaging tool for making the definitive diagnosis in these cases. Additional mammographic views, breast US, clinical breast examination, and needle or surgical biopsy are often required [18].

Two other misclassifications occurred in focal atypical ductal hyperplasia (ADH) with associated calcifications as shown in Figure 10. The presence of a cluster of indeterminate microcalcifications is always suspicious of ADH or ductal carcinoma in situ (DCIS). ADH is a lesion with significant malignant potential. The diagnosis of ADH at needle core breast biopsy is normally considered as an indication for surgical excision [19],[21],[46]. In these two cases, VA and mammography were in agreement on suspicious lesions.

The VA images represented in the results section are simply reconstructed from the amplitude of the acoustic emission; additional signal and image processing algorithms were not required. The resolution of the VA images is determined by the spatial distribution of the ARF and is in the sub-millimeter range. Such resolution is generally sufficient for detecting the breast lesions. Since VA and US share common hardware (US scanner and transducer), we envision that in the future VA can be combined with conventional US imaging and become a hybrid imaging modality that can provide physicians with further clinically useful information.

We used mammography as the reference modality because the confocal VA system is integrated with a mammography machine that allowed acquiring matching mammography and VA images from the same view angle and through the same imaging window, which facilitated the comparison. We also note that traditional breast US and elastography images are obtained in the B-Plane, which is perpendicular to the C-plane images of VA. This difference in imaging plane makes it difficult to compare VA against US and elastography images to match the location of breast lesions. In contrast, we can obtain VA and mammography from an identical imaging window, thus one-to-one comparison becomes possible. We also emphasize that VA is introduced as a complementary technique to improve sensitivity and specificity in clinical breast imaging [35]. Compared to conventional US imaging, VA images are speckle free, thus, the images have high-contrast that allows detection of small structures such as microcalcifications. Thus VA can be a complementary tool to the existing breast imaging tools. Compared to elastography techniques such as ARFI [32] and SWI [33], VA uses dynamic acoustic radiation force in the range of 10s of kHz [34], which is much higher than the frequency used in quasi-static elastography, ARFI, and shear wave imaging (normally in 10s to 100s Hz range). VA images represent the acoustic response of tissue, which is a complex function of several parameters, including the elasticity and viscosity. However, elasticity cannot be directly and quantitatively measured from this acoustic response.

VA imaging of the breast with a confocal VA system in the prone position has certain limitations. One of the drawbacks of this system is limited access to parts of the breast near the chest wall. This was due to mechanical limitations caused by the patient position (prone position) which put a part of the breast close to the chest wall outside the imaging window and thus not accessible. Also, the need for two-dimensional raster scanning of the transducer resulted in slow image acquisition. To overcome such limitations of confocal VA, we recently developed a new VA system with a handheld array transducer that is implemented on a clinical ultrasound scanner. This system is now being tested in an ongoing study on patients in supine position [51].

We understand that malignant lesions are clinically more important. We decided to focus this study on benign breast lesions, because benign lesions are also clinically important as they are much more common than malignant breast lesions. Benign breast lesions cover a wide range of abnormalities with different VA characteristic, thus warranting a separate a study.

Furthermore, the case selection did not represent a typical screening process. Nonetheless, this study represented an important first step in studying the utility of VA for detection and characterization of breast masses. Despite the mentioned limitations, the results of this study show the ability of the technique to identify the benign lesions with high clarity and present VA characteristics of a wide range of benign lesions.

Although the sample size in the present study was rather small, we believe that our results can be reproduced and validated in future clinical studies. We anticipate that by implementing VA on a clinical US scanner and by using a 2D US transducer [52], VA will become a more suitable tool for clinical usage. Additional studies are needed to compare the capabilities of VA and traditional ultrasound imaging.