Back to results list
Show full item record
Please use this identifier to cite or link to this item:
|Title:||Development of a 3D breast ultrasound system for providing automated annotation and coronal images||Authors:||Jiang, Weiwei||Degree:||Ph.D.||Issue Date:||2015||Abstract:||Breast cancer is a malignant tumor that originates in breast tissue. It is the most common cancer in women worldwide and is second to lung cancer as a cause of cancer-related deaths in women. Up to now, there has not been an effective method to prevent breast cancer and early detection has remained the cornerstone for breast cancer control. As one of the widely used assessment methods, breast ultrasound imaging plays a very important role in the decline of breast cancer death rates because of its advantages of radiation-free, real-time and suitable for dense breast. It is also valuable in differentiating malignant masses from benign lesions, which can greatly help to reduce the number of biopsies.However, as the routine detection method, the clinical two-dimensional (2D) breast ultrasound imaging still has some limitations. One limitation is the manual annotation method. In 2D breast ultrasound scanning, the annotation is a critical part of the ultrasound image and is used to register the image location on the breast. However, clinical ultrasound annotation procedures highly depend on the operator's experience and training. This manual procedure is a highly time-consuming and repetitive task for operators. Moreover, manual annotation involves approximate positional registration according to the operator's estimation. This procedure may lead to inaccurate annotation or even errors. The other limitation for clinical breast ultrasound imaging is that it cannot provide breast coronal images for the 2D scanning method. Images along coronal plane contain important diagnosis information such as retraction phenomenon, which is an important feature of malignant tumors. To overcome above limitations of the clinical breast ultrasound scanning method, in this study, a three-dimensional (3D) breast ultrasound system for providing automated 3D annotation and coronal images was developed. This system comprised of an ultrasound scanner with a linear probe, a compact electromagnetic spatial sensing device and a computer with customized software and video capture card. The spatial sensing device contained three parts: the sensor, transmitter and control box. The sensor was mounted on the ultrasound probe to get real-time spatial data of ultrasound images. During examination, images were generated by the scanner and digitized into the computer using the video capture card. Meanwhile, the corresponding spatial information was also sent to the computer. Based on images together with their spatial data, a program was developed by Visual C++ and Visualization Toolkits to acquire data, calculate and display image annotation, and render coronal images.
In this study, an automated 3D annotation method was proposed. In this method, three parameters were employed to indicate the image location, including the distance to the nipple, the radial angle and the probe tilt information. The 3D model, 2D pictogram and textual sequence were employed to display these three parameters. To assess system accuracy, two phantoms were designed. The distance/angle was measured by our system and also by the micrometer/vernier protractor. Very good linear correlations were obtained for both the distance measurement (y = 0.991x, R² = 0.9998) and angle measurement (y = 1.0098x, R² = 0.9995). The comparison of the results obtained by this automated annotation method and the manual annotation method also demonstrated linear regression in distance and angle. Breast phantom experiments verified that this method had good repeatability, with intra-class correlation coefficients (ICC) of 0.907 and 0.937 for the intra-rater and inter-rater reliability tests, respectively. For the clinical test of this new system, 21 breast patients were recruited and comparison between the manual annotation and our automated annotation was made. The results revealed that this new annotation method could greatly help to save examination time compared with the conventional method. Good correlation between the manual and automated annotation methods was found (R² = 0.8571 for distance, R² = 0.9983 for radial angle). The repeatability test also showed good performance of this new system in clinical setting (ICC = 0.989 for intra-rater tests, ICC = 0.927 for inter-rater tests).To obtain coronal images, a narrow-band rendering method was used based on B-mode images and their corresponding spatial data. In validation phantom experiments, 20 inclusions with different sizes (5mm - 20mm) were measured on coronal images obtained by this new system. The results obtained well correlated with those measured by a micrometer (y = 1.0147x, R² = 0.9927). Phantom tests also showed that this system had excellent intra- and inter- rater repeatability (ICC>0.995). Threes subjects with breast lesions were scanned in vivo using this new system and a commercially available three-dimensional (3D) probe. The results revealed that this new method required shorter scanning time (74s for commercial 3-D probe, 64s for our system). The tumor sizes measured on the coronal plane provided by the new method were smaller by 5.6% to 11.9% in comparison with results of the 3D probe.In conclusion, based on clinical free-hand ultrasound scanner and a spatial sensor, a 3D breast ultrasound system for providing automated annotation and coronal images was successfully developed in this study. Phantom experiments and clinical trials revealed that this system could provide great help to breast ultrasound examination.
|Subjects:||Breast -- Ultrasonic imaging.
Three-dimensional imaging in medicine.
Breast -- Diseases -- Diagnosis.
Breast -- Cancer -- Ultrasonic imaging.
Hong Kong Polytechnic University -- Dissertations
|Pages:||xix, 147 pages : illustrations ; 30 cm|
|Appears in Collections:||Thesis|
View full-text via https://theses.lib.polyu.edu.hk/handle/200/8082
Citations as of Jun 4, 2023
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.