Last updated: 05/11/2018 6:00

The project goal is to create the framework for a learning health system that can that can identify potentially erroneous data with statistical anomaly detection. The system will allow the implementation of unique integrity check classes from the user.

• Students: Daniel Yuan, Vincent Qi
• Mentor(s): Dr. Todd McNutt, Pranav Lakshminarayanan

The final objective of the entire project is to improve quality of clinical data available to physicians in order to minimize the risk involved with radiotherapy for cancer patients. The target goal for the time period given is to improve the integrity of the contour model data using a learning health system by implementing tools that can help identify potentially erroneous data with statistical anomaly detection. We aim to build a framework that allows modular insertion of various detection rules in order to allow active approach using constantly updated clinical databases. Previous integrity checking focused on pre-analyzed methods across a set of patients for standardized integrity and treatments.

• Minimum: (April 15th, 2018)
  1. Working Framework that allows for modular insertion of new integrity checks
  2. Documented API to develop new integrity checks

• Expected: (April 29th, 2018)
  1. Implemented existing errant detection modules into working framework
  2. Implement new anomaly detection modules

• Maximum: (May 7th, 2018)
  1. Develop and implement numerous new integrity checks
  2. Implement compatibility packet to allow other programs access to results easily

We will be developing the framework in python. Below are diagrams of our design

• Dependency 1:
  • Access to clinical database
• Plan to resolve:
  • Coordinate with mentors
• Estimated Resolution Date:
  • February 26th
• Effect if not resolved:
  • Project cannot move forward until either this or alternative is done
• Alternatives:
  • Find other databases to work with
• Current Status:
  • RESOLVED

• Dependency 2:
  • Access to previous code
• Plan to resolve:
  • Coordinate with mentors
• Estimated Resolution Date:
  • February 26th
• Effect if not resolved:
  • No previous code to work off of, delaying progress and moving back timeline
• Alternatives:
  • Build own code base
• Current Status:
  • RESOLVED

• Dependency 3:
  • Access to computational power
• Plan to resolve:
  • Coordinate with mentors
• Estimated Resolution Date:
  • Unknown
• Effect if not resolved:
  • Cannot test code on larger datasets
• Alternatives:
  • Test on smaller datasets
• Current Status:
  • Using smaller datasets

1. Milestone name: Presentation
   • Planned Date: February 20th, 2018
   • Expected Date: February 20th, 2018
   • Status: Complete
2. Milestone name: Proposal
   • Planned Date: February 26th, 2018
   • Expected Date: February 26th, 2018
   • Status: Complete
3. Milestone name: Framework Design
   • Planned Date: March 15th, 2018
   • Expected Date: March 15th, 2018
   • Status: Complete
4. Milestone name: Existing module implementation
   • Planned Date: March 25th, 2018
   • Expected Date: March 25th, 2018
   • Status: Complete
5. Milestone name: Statistical module
   • Planned Date: April 7th, 2018
   • Expected Date: April 7th, 2018
   • Status: Complete
6. Milestone name: First new module
   • Planned Date: April 7th, 2018
   • Expected Date: April 7th, 2018
   • Status: Complete
7. Milestone name: More complex modules
   • Planned Date: April 23rd, 2018
   • Expected Date: April 23rd, 2018
   • Status: In-progress
8. Milestone name: Final presentation
   • Planned Date: May 11th, 2018
   • Expected Date: May 11th, 2018
   • Status: Complete

• Project Plan
  • Project plan presentation
  • Project plan proposal
• Project Checkpoint
  • midpoint_presentation.pdf
• Paper Seminar Presentations
  • McNutt et al. Paper
  • McNutt et al. Presentation
  • McNutt et al. Paper Report
  • Gaffney et al. Presentation
  • Gaffney et al. Paper Report
  • Gaffney et al. Paper
• Project Final Teaser
  • Teaser Slide
• Project Final Presentation
  • Final Poster
• Project Final Report
  • Final Report

* See presentations

Github repository (access requires permission from team members): https://github.com/necroteddy/oncotools-master