Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


Dissertation.pdf (8.18 MB)

ETD Abstract Container

Abstract Header

An Intelligent Analysis Framework for Clinical-Translational MRI Research

Yang, Kun
http://rave.ohiolink.edu/etdc/view?acc_num=case1592254585828664

Abstract Details

2020, Doctor of Philosophy, Case Western Reserve University, EECS - System and Control Engineering.
"Nothing beats seeing." Visualization disease is one of the most powerful tools in modern healthcare; it fundamentally changed the methodology in disease diagnosis and treatment. Among these different modalities, magnetic resonance imaging (MRI) is playing a critical role. Thus the novel algorithms and advanced techniques based on MRI are in dire need. With the unmet motivation and current clinical needs, this dissertation considers designing an end-to-end intelligent analysis framework for clinical-translational MRI research. MRI clinical applications historically have generated large amounts of data, driven by record keeping, data analysis, and regulatory requirements. In addition, such image analysis, including image reconstruction, registration, and perfusion quantification that performed "\textit{offline}" in the current protocol, is computationally-intensive and time-consuming. These limitations make some MRI analysis applications inappropriate for clinical timescale. Driven by the potential to improve the efficiency of MRI data analysis while reducing the costs, we develop a data streaming analysis framework that allows efficient "\textit{online}" processing. With optimized reconstruction and registration algorithms, as well as the integration of external computing resources, including Graphics Processing Units (GPUs) parallel computing techniques, the streaming framework achieved 180 times speed-up compared with the original protocol. Our Liver DCE and Crohn's Disease (CD) experiments showed significantly increased speed (Average 7.72 minutes total analysis time compared to 21.6 hours by original protocol) with minor differences in both image quality and perfusion quantification results. This framework also allows easy and direct deployment of clinical studies. With the data analysis framework, we further explore the potential to integrate artificial intelligence (AI) into our system. In Chapter 3, we introduce prostate cancer classification with the support of Magnetic Resonance Fingerprinting (MRF) and deep learning technique. The neural network classifier enables extracting quantitative features to increase the accuracy of cancer classification. Two classifiers have been designed, one classifier fed with quantitative biomarkers, the other fed with MRI and MRF images, the analysis framework allows end-to-end training. The cancer prediction results showed an AUC of 0.90 in the peripheral zone (PZ) and AUC of 0.89 in the transition zone (TZ). Both studies achieve better results than the clinical standard and previous works. This application largely reduces radiologists' efforts and time. More importantly, the ability to reproducibly and quantitatively analyze images could enable a more objective diagnosis. Almost all deep learning applications shared a critical challenge that neural networks require significant architecture engineering, which costs tremendous human efforts and indeed brings in human bias. In Chapter 4, we propose an efficient automatic neural network design method to classify prostate cancer based on neural architecture search (NAS). The infinite possibilities of architecture search space have been optimized in this study. The novel two-level design of search space and transfer-learning tremendously reduces the training time with friendly computation requirements (2.8 days, Nvidia 1080Ti). The experiment results demonstrate that our method outperforming state-of-art hand-crafted networks in prostate cancer classification. More promisingly, we test datasets from different medical centers, and results also show that the proposed method has excellent performance, which enables the potential for broader clinical deployment.
Pan Li (Advisor)
Kenneth Leparo (Committee Chair)
Xin Yu (Committee Member)
Daniel Saab (Committee Member)
80 p.
Biomedical Engineering; Biomedical Research; Computer Engineering; Computer Science; Medical Imaging; Systems Design

Recommended Citations

Citations

  • Yang, K. (2020). An Intelligent Analysis Framework for Clinical-Translational MRI Research [Doctoral dissertation, Case Western Reserve University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=case1592254585828664

    APA Style (7th edition)

  • Yang, Kun. An Intelligent Analysis Framework for Clinical-Translational MRI Research. 2020. Case Western Reserve University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=case1592254585828664.

    MLA Style (8th edition)

  • Yang, Kun. "An Intelligent Analysis Framework for Clinical-Translational MRI Research." Doctoral dissertation, Case Western Reserve University, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=case1592254585828664

    Chicago Manual of Style (17th edition)

case1592254585828664
40
© 2020, all rights reserved.
This open access ETD is published by Case Western Reserve University School of Graduate Studies and OhioLINK.