My lab focuses on the diagnosis and treatment of cancer using advanced imaging techniques. There are 3 main projects: tomosynthesis imaging, computer aided diagnosis, and breast modeling.
First, I lead a team from the Ravin Advanced Imaging Laboratories (RAI Labs) that collaborates closely with Siemens Healthcare to develop breast tomosynthesis imaging, a form of limited-angle tomography also known as "tomo" or "3D mammography." Tomosynthesis can acquire a 3D image quickly, easily, and at comparable dose as conventional mammography. By improving both sensitivity and specificity of breast cancer diagnosis, tomo has become the most exciting recent development in breast cancer screening, and the only technology with the potential to replace mammography in the near future. We recently concluded both an NIH-sponsored clinical trial as well as a multi-center FDA trial for Siemens.
Fig. 1: Sample images from our studies. This patient presented with a very subtle, indistinct mass as shown in the standard mammogram (left). Even with the magnification view, the mass is still very subtle (middle). The tomo scan easily reveals a spiculated mass which was biopsied to reveal invasive ductal carcinoma.
Second, computer aided diagnosis (CAD) is an interdisciplinary field combining both imaging and informatics. We developed computer vision algorithms to localize suspicious mammographic lesions. We also created predictive models that use machine learning and statistical analysis in order to classify mammograms as benign versus malignant. During these studies, we compiled one of the largest multi-institution breast cancer databases with approximately 5000 cases. In ongoing studies funded by NIH and DOD, we are exploring the relationship of imaging findings with genomic markers to predict which cases of DCIS are likely to be more aggressive and progress toward invasive cancer. This addresses the clinically significant problem of overdiagnosis, and provides women with more personalized risk assessment to inform their treatment decisions.
Finally, I have been working with colleagues from RAI Labs to create new virtual breast models that are based on actual patient data. These models go far beyond conventional phantoms in portraying realistic breast anatomy. Furthermore, we can transform these virtual models into physical form using the latest 3D printing technology, as shown in Fig. 2 below. Such physical phantoms can be scanned on actual mammography and tomosynthesis systems, allowing us to measure image quality in new ways that are not only quantitative but also clinically relevant. We continue to refine the realism of these physical phantoms, and seek to develop new procedures for quality control and system evaluation.
Fig. 2: (Left) Physical breast phantom opened up in butterfly position to demonstrate its realistic interior structures. (Right) Actual tomo images of the phantom acquired by two different tomo systems. The dramatic differences in image appearance support the need for further research to evaluate and optimize these systems.
I am one of the six core faculty members of the Carl E. Ravin Advanced Imaging Labs (RAI Labs), which includes over 30 faculty, staff and students devoted to research in advanced digital imaging techniques and applications for improved health care. RAI Labs occupies over 7,000 square feet of laboratory and office space. We are a part of the Department of Radiology in the Duke University School of Medicine, and also affiliated with the Duke Medical Physics Graduate Program. Duke University Health System, one of the largest private hospitals in the United States, is licensed for over 1,000 beds. My research is conducted in close collaboration with the Division of Breast Imaging, which performs over 450,000 exams each year, approximately 1,000 of which are sent to biopsy each year.