Date of Award
Doctor of Philosophy (PhD)
Worldwide, breast cancer is the most common cancer among women. In the United States alone, the American cancer society has estimated there will be 271,270 new breast cancer cases in 2019, and 42,260 lives will be lost to the disease. Ultrasound (US), mammography, and magnetic resonance imaging (MRI) are regularly used for breast cancer diagnosis and therapy monitoring. However, they sometimes fail to diagnose breast cancer effectively. These shortcomings have motivated researchers to explore new modalities. One of these modalities, diffuse optical tomography (DOT), utilizes near-infrared (NIR) light to reveal the optical properties of tissue. NIR-based DOT images the contrast between a suspected lesion’s location and the background tissue, caused by the higher NIR absorption of the hemoglobin which characterizes tumors. The limitation of high light scattering inside tissue is minimized by using ultrasound image to find the tumor location.
This thesis focuses on developing a compact, low-cost ultrasound guided diffuse optical tomography imaging system and on improving optical image reconstruction by extracting the tumor’s location and size from co-registered ultrasound images. Several electronic components have been redesigned and optimized to save space and cost and to improve the user experience. In terms of software and algorithm development, manual extraction of tumor information from ultrasound images has been replaced by using a semi-automated ultrasound image segmentation algorithm that reduces the optical image reconstruction time and operator dependency. This system and algorithm have been validated with phantom and clinical data and have demonstrated their efficacy. An ongoing clinical trial will continue to gather more patient data to improve the robustness of the imaging algorithm.
Another part of this research focuses on ovarian cancer diagnosis. Ovarian cancer is the most deadly of all gynecological cancers, with a less than 50% five-year survival rate. This cancer can evolve without any noticeable symptom, which makes it difficult to diagnose in an early stage. Although ultrasound-guided photoacoustic tomography (PAT) has demonstrated potential for early detection of ovarian cancer, clinical studies have been very limited due to the lack of robust PAT systems.
In this research, we have customized a commercial ultrasound system to obtain real-time co-registered PAT and US images. This system was validated with several phantom studies before use in a clinical trial. PAT and US raw data from 30 ovarian cancer patients was used to extract spectral and statistical features for training and testing classifiers for automatic diagnosis. For some challenging cases, the region of interest selection was improved by reconstructing co-registered Doppler images. This study will be continued in order to obtain quantitative tissue properties using US-guided PAT.
Mark Anastasio, Adam Bauer, Hong Chen, Yuan-chuan Tai,