Morphomics

Project Information

Discipline

Statistics (403) 

Subdiscipline

14.05 Bioengineering and Biomedical Engineering 

Orientation

Research 

Cross-sectional images are used by surgeons for operative planning. These images contain vast amounts of additional data specific to that patient which are never assessed by clinicians. Analytic morphomics is a novel approach to clinical image analysis and may inform perioperative risk evaluations, adding objectivity to the subjective “eyeball test”. Our recent work has shown strong relationships between patient morphometric characteristics on preoperative imaging and surgical outcomes following major surgery. We propose to utilize these morphomics to predict various medical outcomes

This proposal leverages novel technologies and available data to communicate surgical risk. The surgical literature is replete with studies of the risk factors for poor surgical outcomes, but little effort has been made to bring this to the bedside and affect surgical decision-making. This proposal will extract data from preoperative images to inform clinical decisions. Over 90% of patients have cross-sectional imaging prior to major surgery. Within these images exist a robust amount of patient specific data well suited for preoperative risk assessment. Developing methods to optimally utilize these scans and to inform clinical decision-making builds upon our innovative work in these areas.

The decision to operate surgical procedures appears to be at the root-cause of many preventable peri-operative poor outcomes. Identifying and improving factors that enhance surgical decision-making would have an obvious value for patients. We will use novel morphomics to build such risk prediction models.

Project Contact

Project Lead

Peng Zhang (pczhang) 

Project Manager

Peng Zhang (pczhang) 

Project Members

Peng Zhang  

Resource Requirements

Hardware System

• I don't care (what I really need is a software environment and I don't care where it runs)

 

will be running statistical analysis on morphomics, mostly supervised learning, such as glm, random forrest, elastic net.

running analysis a few times per weeks. need more cpus to conduct it parallelly. (100 nodes and 30G memory)

Project Timeline

Submitted

03/25/2013 - 22:14