TL;DR
This paper introduces Quantitative Transmission Ultrasound (QT Ultrasound) as a novel 3D imaging technique for breast anatomy. It demonstrates QT Ultrasound's ability to accurately visualize breast structures and differentiate tissue types using speed of sound measurements, outperforming traditional mammography in image quality for many anatomical features.
What Problem Does This Paper Address?
Traditional X-ray mammography (XRM) is the standard for breast cancer screening, but it struggles to visualize fine anatomical details, particularly in women with dense breast tissue. This lack of detail can lead to challenges in lesion detection and characterization, potentially causing anxiety and unnecessary biopsies. While handheld ultrasound (HHUS) offers better resolution, it's limited in depth and typically provides only 2D images. The need for a high-resolution, 3D breast imaging modality that can accurately delineate anatomy and differentiate tissue types, especially in dense breasts, remains a significant challenge.
Methods
The study employs a multi-faceted approach:
- Quantitative Transmission Ultrasound (QT Ultrasound): This novel technology uses a multifrequency nonlinear 3D inverse-scattering algorithm. It performs both reflection and transmission data acquisition. The transmission data is used to create speed of sound maps, which are then used to correct reflection images for refraction and to generate 3D breast images. This technique offers high resolution (0.75 mm x 0.75 mm x 1.5 mm FWHM) and quantitative speed of sound measurements without ionizing radiation, compression, or contrast agents.
- Correlative Imaging: Cadaver breast specimens were imaged using QT Ultrasound, then sectioned and photographed. Biopsies were performed to confirm microanatomy, and these histological findings were correlated with the QT Ultrasound images.
- Speed of Sound Validation: Speed of sound measurements were taken for six histologically confirmed normal breast tissue types (fat, pectoralis muscle, ductal tissue, glandular tissue, skin, and connective tissue) in both cadaver and live volunteer breasts.
- Tissue Classification: Linear Discriminant Analysis (LDA) was used to classify different normal breast tissue types based on QT Ultrasound's three modalities: speed of sound, reflection, and attenuation. A leave-one-out cross-validation approach was used to assess classification performance.
- Visual Grading Analysis (VGA): Radiologists (readers) compared the image quality of ten normal breast structures visualized on QT Ultrasound versus XRM in live subjects. A standardized rating scale was used to assess image quality.
Datasets
- Cadaver Breasts: 3 human cadaver breast specimens were used for correlative imaging and speed of sound validation.
- Live Volunteer Breasts:
- 150 speed of sound measurements were taken from 4 normal volunteers for validation.
- 22 normal breasts (44 image sets) from living subjects were used for the visual grading analysis. These were part of a case collection study approved by IRBs at three institutions.
Key Insights and Ideas
- Speed of Sound as a Biomarker: The study strongly validates the speed of sound as a reliable, tissue-specific identifier for various breast tissues. Different tissue types exhibit distinct speed of sound values, allowing for quantitative differentiation.
- Anatomical Detail in Dense Breasts: QT Ultrasound shows promise in visualizing anatomical detail in dense breast tissue, a significant challenge for conventional mammography. The ability to differentiate glandular elements from ducts is highlighted as a key advancement.
- 3D Reconstruction Advantage: The inherent 3D nature of QT Ultrasound allows for a more comprehensive visualization of breast anatomy compared to 2D methods, facilitating better understanding of spatial relationships between structures.
- Radiologist Preference: Radiologists generally found the image quality of QT Ultrasound to be superior to XRM for visualizing most breast features, indicating its potential clinical utility.
Key Results
- Anatomical Identification: QT Ultrasound successfully identified all major anatomical features of the breast.
- Tissue Specificity: Speed of sound calculations demonstrated specific values for different breast tissues.
- LDA Classification Accuracy: The linear discriminant analysis achieved an overall accuracy of 88.5% in classifying normal breast tissue types. This accuracy improved to 91.4% when skin tissue was excluded from the classifier.
- Visual Grading Analysis:
- Readers scored QT Ultrasound image quality as equivalent or better than XRM in 69%-90% of breasts for specific tissues.
- For most features (except pectoralis muscle), readers rated QT Ultrasound image quality as equivalent or better than XRM in over 90% of breasts.
- QT Ultrasound was rated as better than XRM for skin, epidermis, and hypodermis in 69%-78% of breasts.
- The image quality of pectoralis muscle was rated superior on XRM in almost 60% of breasts.
Limitations and Future Work
- Pectoralis Muscle Visualization: The study notes that readers scored the image quality of the pectoralis muscle on XRM as better than QT Ultrasound in almost 60% of breasts. This suggests a potential area for improvement in QT Ultrasound's visualization of this specific tissue.
- Future Clinical Studies: The authors are currently conducting controlled clinical studies to evaluate the sensitivity, specificity, callback rates, and biopsy rates of QT Ultrasound in women with dense breasts to further determine its clinical value.
- Further Correlation: Future publications are planned to explore more detailed correlations with large-format and 3D histology, as well as micro-CT and high-field MRI.
Summary updated