Architectural Support for Medical Imaging
dc.contributor.author | Sampson II, Richard | |
dc.date.accessioned | 2017-06-14T18:35:01Z | |
dc.date.available | NO_RESTRICTION | |
dc.date.available | 2017-06-14T18:35:01Z | |
dc.date.issued | 2017 | |
dc.date.submitted | ||
dc.identifier.uri | https://hdl.handle.net/2027.42/137105 | |
dc.description.abstract | Advancements in medical imaging research are continuously providing doctors with better diagnostic information, removing the need for unnecessary surgeries and increasing accuracy in predicting life-threatening conditions. However, newly developed techniques are currently limited by the capabilities of existing computer hardware, restricting them to expensive, custom-designed machines that only the largest hospital systems can afford or even worse, precluding them entirely. Many of these issues are due to existing hardware being ill-suited for these types of algorithms and not designed with medical imaging in mind. In this thesis we discuss our efforts to motivate and democratize architectural support for advanced medical imaging tasks with MIRAQLE, a medical image reconstruction benchmark suite. In particular, MIRAQLE focuses on advanced image reconstruction techniques for 3D ultrasound, low-dose X-ray CT, and dynamic MRI. For each imaging modality we provide a detailed background and parallel implementations to enable future hardware development. In addition to providing baseline algorithms for these workloads, we also develop a unique analysis tool that provides image quality feedback for each simulation. This allows hardware designers to explore acceptable image quality trade-offs in algorithm-hardware co-design, potentially allowing for even more efficient solutions than hardware innovations alone could provide. We also motivate the need for such tools by discussing Sonic Millip3De, our low-power, highly parallel hardware for 3D ultrasound. Using Sonic Millip3De, we illustrate the orders-of-magnitude power efficiency improvement that better medical imaging hardware can provide, especially when developed with a hardware-software co-design. We also show validation of the design using a scaled-down FPGA proof-of-concept and discuss our further refinement of the hardware to support a wider range of applications and produce higher frame rates. Overall, with this thesis we hope to enable application specific hardware support for the critical medical imaging tasks in MIRAQLE to make them practical for wide clinical use. | |
dc.language.iso | en_US | |
dc.subject | Medical Imaging Hardware | |
dc.title | Architectural Support for Medical Imaging | |
dc.type | Thesis | en_US |
dc.description.thesisdegreename | PhD | en_US |
dc.description.thesisdegreediscipline | Computer Science & Engineering | |
dc.description.thesisdegreegrantor | University of Michigan, Horace H. Rackham School of Graduate Studies | |
dc.contributor.committeemember | Wenisch, Thomas F | |
dc.contributor.committeemember | Fessler, Jeffrey A | |
dc.contributor.committeemember | Kripfgans, Oliver Daniel | |
dc.contributor.committeemember | Mahlke, Scott | |
dc.subject.hlbsecondlevel | Computer Science | |
dc.subject.hlbtoplevel | Engineering | |
dc.description.bitstreamurl | https://deepblue.lib.umich.edu/bitstream/2027.42/137105/1/rsamp_1.pdf | |
dc.identifier.orcid | 0000-0002-8669-6584 | |
dc.identifier.name-orcid | Sampson, Richard; 0000-0002-8669-6584 | en_US |
dc.owningcollname | Dissertations and Theses (Ph.D. and Master's) |
Files in this item
Remediation of Harmful Language
The University of Michigan Library aims to describe library materials in a way that respects the people and communities who create, use, and are represented in our collections. Report harmful or offensive language in catalog records, finding aids, or elsewhere in our collections anonymously through our metadata feedback form. More information at Remediation of Harmful Language.
Accessibility
If you are unable to use this file in its current format, please select the Contact Us link and we can modify it to make it more accessible to you.