Investigators: Current Graduate Students: Current UnderGraduates: Recent Graduates:

Projects:

Recent Publications/Presentations:

Radonc/BME     Medical Image and
Computational Analysis Laboratory
Daniel Murff
MS in BME with Medical Physics concentration
Graduated: May 2015
Currently: medical physicist, private practice, Jacksonville FL

email link
  1. Resume (pdf)
  2. Research Overview

    Small tumor detection with 3D template matching

    Background and Significance
    Many groups have developed computer-aided detection (CAD) systems to aid in finding lung-nodules in medical images. However, a "ground-truth" method for validating those CAD system remains elusive. Some groups have attempted to validate their software using virtually created lung-nodule models, other groups have attempted to stockpile large medical image databases containing a variety of lung-nodules, and still others have attempted a "cut-and-paste" method for inserting tumors from one medical image into another image. Our group has chosen to validate our CAD system via a set of synthetically created lung nodules, which can then be seamlessly inserted into medical images such as CT's, and subsequently used for validation purposes.

          A problem with using a virtual tumor approach like the one we use is that we have used near-spherical shapes with homogenous textures as mathematical models to find and characterize tumor candidates. A successful solution to this problem would include creating a set of dynamically-shaped, non-solid synthetic tumors which could be scaled in shape and texture with unlimited variability. Such a set of virtual tumors could then be established as a ground-truth method for validating CAD systems. O'Dell et al have already addressed the issue of shape.

    Goal
    Modify the Ambrosini lung nodule set in order to create a set of non-solid tumor models.

    Scope
    The scope of this project encompasses the database of 17 synthetic tumors built by Ambrosini. Once the original set of 17 tumors has been modified to appear non-solid, they can then be scaled bigger or smaller, allowing for unlimited variation.

    Benefits
    Benefits following the successful completion of this project include:
    • non-solid tumors will allow for a more realistic tumor database than solid tumors, to be used for testing the ability of the CAD system to detect actual tumors
    • tumor database has the potential to serve as a ground-truth for training the clinician

    Deliverables
    Upon successful completion of this project, we will be able to provide a synthetic set of 17 non-solid tumors that more closely mimic actual tumors than the current set of solid tumors. This set of tumors will be scalable to any dimensions, allowing for unlimited variability.

    Research Plan
    1. In-depth literature review
      Looking through in-house older sources on the subject as well as new advances in this area in the past 5 -7 years. Relevant publications will be added to and cataloged in an internal on-line repository.
    2. Learn how to write and interpret MatLab
      This includes:
      • going through Ambrosini's MATLAB code and making sense of everything he did
      • reading applicable chapters in "Digital Image Processing with MATLAB" book
      • regular MATLAB practice sessions
      • weekly meetings with Dr. O'Dell to sift through the code and organize a plan for moving forward
    3. Modify Ambrosini code/ write MATLAB code to create non-solid tumors
      Modify the existing tumor database to create non-solid tumors.
    4. Write and submit manuscript to Medical Physics journal for publication detailing the advances on this project.
  3. Related MIACALab projects
    1. Early detection of breast cancer metastases

  4. Related Publications
    1. Ambrosini R, O'Dell WG
      Realistic simulated lung nodule dataset for testing CAD detection and sizing.
      Proceedings of SPIE Medical Imaging, San Diego, CA, Feb 2010
    2. Ambrosini R, Wang P, O'Dell WG
      Volume change determination of metastatic lung tumors in CT images using 3-D template matching.
      Proceedings of SPIE Medical Imaging [#7260-112], Orlando, FL, Feb 2009

  5. Close Research Collaborators
    1. W. O'Dell (Radiation Oncology Research)
    2. David Wymer, MD, Radiology
    3. Roger Shifrin, MD, CEO of RYSCI BIZNZ radiologist services