Intelligent Ultrasound and Photoacoustic Imaging Systems: Design, Development and Beyond
Open Access
- Author:
- Agrawal, Sumit
- Graduate Program:
- Bioengineering (PHD)
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- September 27, 2021
- Committee Members:
- Sri-Rajasekhar Kothapalli, Chair & Dissertation Advisor
William Higgins, Major Field Member
Nanyin Zhang, Major Field Member
Vishal Monga, Outside Unit & Field Member
Daniel Hayes, Program Head/Chair - Keywords:
- ultrasound imaging
photoacoustic imaging
artificial intelligence
spectral unmixing
artifact removal
low-cost devices
portable medical imaging device - Abstract:
- Over the past decade, photoacoustic (PA) imaging has evolved as a multi-scale imaging technology, enabling in vivo imaging from single cells to complete organs. PA imaging provides optical absorption based functional and molecular contrasts of deeper (> 3 mm) tissue with higher spatio-temporal resolutions than possible with conventional optical imaging methods. For example, in PA imaging, the differential optical absorption spectra of oxy- and deoxy- forms of hemoglobin enables unrivaled label-free imaging of vascular morphology, associated angiogenesis, blood oxygenation, oxygen metabolism, and blood flow. Such a rich multiparametric vascular information has found many applications in preclinical studies such as transcranial imaging of mouse brain activity as well as several clinical applications such as breast angiography, Crohn’s disease activity, prostate cancer screening and peripheral vascular disease. Despite such advancements, dual modality US and PA (USPA) imaging systems face the following long-standing technical challenges: (i) Requirement of a bulky class-IV high power tunable laser source limits portability and increases the overall cost and footprint of USPA systems, (ii) depth and wavelength dependent optical attenuation leads to spectral coloring, and (iii) the unknown optical and acoustic heterogeneities introduces artifacts and limit quantitative PA imaging performance in deep tissue regions. My Ph.D. work tackled above challenges and developed smart USPA imaging systems by integrating novel simulations, instrumentation, and artificial intelligence (AI) algorithms. This included developments of (i) a novel hybrid simulation platform capable of generating large scale, application-specific, and tissue-realistic US and PA imaging data, (ii) USPA domain-enriched AI networks for signal denoising, spectral unmixing and artifact correction in PA images of living mice and humans, (iii) affordable and portable USPA imaging and sensing systems by integrating several state-of-the-art technologies in optics, ultrasound and electronics, and (iv) a multimodal (US, PA, Doppler US, and Elastography) imaging approach capable of providing structural, functional and molecular contrasts for accurate diagnosis of cancer, cardiovascular and neurological diseases.