Yujun Guo

Image Registration Accuracy of a 3-Dimensional Transrectal Ultrasound-Guided Prostate Biopsy System

Objective. For a follow‐up prostate biopsy procedure, it is useful to know the previous biopsy locations in anatomic relation to the current transrectal ultrasound (TRUS) scan. The goal of this study was to validate the performance of a 3‐dimensional TRUS‐guided prostate biopsy system that can accurately relocate previous biopsy sites. Methods. To correlate biopsy locations from a sequence of visits by a patient, the prostate surface data obtained from a previous visit needs to be registered to the follow‐up visits. Two interpolation methods, thin‐plate spline (TPS) and elastic warping (EW), were tested for registration of the TRUS prostate image to follow‐up scans. We validated our biopsy system using a custom‐built phantom. Beads were embedded inside the phantom and were located in each TRUS scan. We recorded the locations of the beads before and after pressures were applied to the phantom and then compared them with computer‐estimated positions to measure performance. Results. In our experiments, before system processing, the mean target registration error (TRE) ± SD was 6.4 ± 4.5 mm (range, 3–13 mm). After registration and TPS interpolation, the TRE was 5.0 ± 1.03 mm (range, 2–8 mm). After registration and EW interpolation, the TRE was 2.7 ± 0.99 mm (range, 1–4 mm). Elastic warping was significantly better than the TPS in most cases (P < .0011). For clinical applications, EW can be implemented on a graphics processing unit with an execution time of less than 2.5 seconds. Conclusions. Elastic warping interpolation yields more accurate results than the TPS for registration of TRUS prostate images. Experimental results indicate potential for clinical application of this method.

Multi-modality image registration using gradient vector flow intensity

Similarity measure plays a critical role in image registration. Mutual information (MI) has been proved to be a promising measure used widely in multi-modality image registration. However, applying mutual information to original intensities only takes statistical information into consideration, while spatial information is not even considered. In this paper, a novel approach is proposed to incorporate spatial information into MI through gradient vector flow (GVF). Mutual information now is calculated from the GVF-intensity (GVFI) map of the original images instead of their intensity values. Multi-modality brain image registration was performed to test the accuracy and robustness of the proposed method. Experimental results showed that the success rate of our method is higher than that of traditional MI-based registration.

Effect of 3D automated prostate segmentation for ultrasound image guided repeat biopsy application

Prostate repeat biopsy has become one of the key requirements in todays prostate cancer detection. Urologists are interested in knowing previous 3-D biopsy locations during the current visit of the patient. Eigen has developed a system for performing 3-D Ultrasound image guided prostate biopsy. The repeat biopsy tool consists of three stages: (1) segmentation of the prostate capsules from previous and current ultrasound volumes; (2) registration of segmented surfaces using adaptive focus deformable model; (3) mapping of old biopsy sites onto new volume via thin-plate splines (TPS). The system critically depends on accurate 3-D segmentation of capsule volumes. In this paper, we study the effect of automated segmentation technique on the accuracy of 3-D ultrasound guided repeat biopsy. Our database consists of 38 prostate volumes of different patients which are acquired using Philips sidefire transrectal ultrasound (TRUS) probe. The prostate volumes were segmented in three ways: expert segmentation, semi-automated segmentation, and fully automated segmentation. New biopsy sites were identified in the new volumes from different segmentation methods, and we compared the mean squared distance between biopsy sites. It is demonstrated that the performance of our fully automated segmentation tool is comparable to that of semi-automated segmentation method.

Breast-Imaging Registration: A Review

Breast cancer is the most common type of cancer in women worldwide. About ten percent of women are confronted with breast cancer in their lives. In the year 2005, it was estimated that there would be approximately 212, 118 new cases of invasive breast cancer and about 41,250 deaths in the United States. Breast cancer is most effectively treated when detected at an early stage, and the survival probability of the patient is dependent on the tumor size at detection time. The larger the tumor size, the larger the probability for the presence of metastases in vital organs. Early detection of the tumor is critical for a good prognosis. A number of different imaging methods for diagnosis and biopsy of suspicious lesions are available. X-ray mammography is the main tool used for the detection and diagnosis of breast malignancies, and it is currently the only medical imaging modality used in screening. Figure 13.1 shows two of the most common projections of mammogram: Figs. 13.1(a) and (b) are examples of mediolateral oblique (MLO) image, whereas Figs. 13.1(c) and (d) are examples of craniocaudal (CC) image. With about 70% sensitivity and 30% positive predictive value, mammography screening has been shown in clinical trials to reduce breast cancer mortality by 25% to 30% for women in the 50 to 70 age group. However, x-ray mammography has limited specificity and sensitivity. About 10% of all cancers are overlooked using x-ray mammography, especially those in dense breasts. It is estimated that approximately two-thirds of these missed cancers are detected retrospectively by radiologists. In addition, about two-thirds of lesions sent to biopsy turn out to be benign. This has led to the investigation of alternative imaging modalities such as ultrasound, magnetic resonance imaging (MRI), computed tomography (CT), positron emission tomography (PET), etc. for the detection and diagnosis of breast cancer. Ultrasound has become a valuable tool to use with mammograms because it is widely available and less expensive than other options. Breast ultrasound is used to target a specific area of concern found by the mammogram. It is a widely accepted adjunct to mammography in patients with palpable masses or symptomatic breast disease. It is well established that breast ultrasound can distinguish solid from cystic masses with an accuracy approaching 100%, and can detect lesions that are not mammographically visible.

P‐186: A Study of CRT (5‐Mpixel) vs. LCD (9‐Mpixel) Displays for Breast Lesion Detection in Full‐Field Digital Mammography and Ultrasound (FFDMUS) Data Sets via Image‐Enhancement Algorithms

The latest technological changes are fast replacing the conventional cathode ray tube (CRT) displays with liquid crystal displays (LCDs). It is thus important to understand and evaluate them. The novelty of our evaluation strategy comes from the usage of computer aided diagnostics-based on pixel intensities. This evaluation system combines both lesion segmentation and quantification. Hence it is an integrated approach. The FFDMUS ultrasound data was acquired and then displayed on LCD and CRT displays. The FFDMUS ultrasound images were segmented using the signal-to-noise ratio (SNR) algorithm. We use Hausdoff distance measure (HDM) and polyline distance metric (PDM) for performance evaluation. Our results using the HDM method on FFDMUS ultrasound images show that lesions quantified from LCD images show a 29% improvement compared to lesions quantified from CRT images. A similar behavior was observed using PDM method. Hence we conclude that use of LCD displays for mammography applications with image enhancement techniques will have a greater diagnostic accuracy compared to the CRT displays.

Effect of PDE-based noise removal on GVF-based deformation model on lesion detection in breast phantom x-ray images from Fischer’s fused FFDM and ultrasound (FFDMUS) imaging system

It has been recently established that fusion of multi-modalities has led to better diagnostic capability and increased sensitivity and specificity. Fischer has been developing fused full-field digital mammography and ultrasound (FFDMUS) system. In FFDMUS, two sets of acquisitions are performed: 2-D X-ray and 3-D ultrasound. The segmentation of acquired lesions in phantom images is important: (1) to assess the image quality of X-ray and ultrasound images; (2) to register multi-modality images, and (3) to establish an automatic lesion detection methodology to assist the radiologist. In this paper, we studied the effect of PDE-based smoother on the gradient vector flow (GVF)-based active contour model for breast lesion detection. CIRS X-ray phantom images were acquired using FFDMUS, and region of interest (ROI) samples were extracted. PDE-based smoother was implemented to generate noise free images. The GVF-based strategy was then implemented on these noise free samples. Initial contours were set as default, and GVF snake then converged to extract lesion topology. The performance index was calculated by computing the difference between estimated lesion area and ideal lesion area. Our performance index with GVF (without PDE smoothing) yielded an average percentage error of 10.32%, while GVF with PDE yielded an average error of 9.61%, an improvement of 7%. We also optimized our PDE smoother for least GVF error estimation, and to our observation, we found the optimal number of iteration was 140. We also tested our program written in C++ on synthetic datasets.