Logo image
Back
Registration of human breast mammography and ultrasonography for improved breast cancer detection
Thesis   Open access

Registration of human breast mammography and ultrasonography for improved breast cancer detection

Jenna Marie Schabdach
Master of Science (M.S.), Drexel University
May 2016
DOI:
https://doi.org/10.17918/etd-6905
pdf
Schabdach_Jenna_201614.04 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Electrical engineering Image registration Breast--Cancer--Diagnosis Breast--Radiography Ultrasonic imaging
Breast cancer is the second leading cause of death for women in the United States. In 2016 alone, over 250,000 new patients will be diagnosed and over 40,000 will die from some form of breast cancer. The current standard of care for breast cancer screening is primarily dependent on mammography. Magnetic resonance imaging (MRI) is extremely valuable for diagnosing patients who have dense or scarred breast tissue; however, the integration of MRI into breast cancer screening can be difficult. Radiographic alternatives to MRI are avoided to prevent risks associated with increased radiation exposure. Volumetric ultrasonic imaging, while inadequate for standalone screening, is a potential solution for complex screening cases if the modality can be fused with mammography scans. The additional information provided by volumetric ultrasound data allows for clearer discrimination between soft tissues possessing similar radiographic densities. This project aims to fuse volumetric ultrasound images with mammography scans by developing a new registration algorithm. This algorithm first reduces the dimensionality of the volumetric ultrasound to that of the target mammography projection. The ultrasound projections are then used to align the ultrasound volume and mammography images in three dimensional space and account for differences in deformation between the two modalities by framing the global and local registrations as an optimization problem.

Metrics

45 File views/ downloads
28 Record Views

Details

Logo image