This repository contains templates for your final written report and GitHub repository. Follow the instructions below to clone this repository, and then turn in your final project's code via a pull request to this repository.
- To start, fork this BMIN503_Final_Project repository.
- Clone the forked repository to your computer.
- Modify the files provided, add your own, and commit changes to complete your final project.
- Push/sync the changes up to your GitHub account.
- Create a pull request on this, the original BMIN503_Final_Project, repository to turn in your final project.
Follow the instructions here if you are unsure what the above steps mean.
DUE DATE FOR FINAL VERSION: 12/12/18 11:59PM. This is a hard deadline. Turn in whatever you have by this date.
Project overview:
The objective of this project is to determine if morphological features of the placenta in a first-trimester 3D ultrasound exam can be used to predict low fetal birth weight. The primary outcome variable in this project is membership to the small-for-gestation-age "sga_10th" category (i.e., the 10th percentile of birth weight normalized with respect to gestational age). Feature selection is first performed to reduce the number of correlated placenta morphology measurements. Logistic regression and random forest are used to create a classifer that relates placenta measurements and maternal/fetal characteristics with the primary outcome variable. The models are evaluated with and without the maternal/fetal characteristics and are assessed by ROC analysis.
The R markdown file Pouch_final_project.Rmd contains the project writeup and R code. This file imports three dummy spreadsheets:
- The first spreadsheet ("non_placenta_measures_dummy") contains maternal and fetal characterstics (e.g., maternal age and fetal sex).
- The second ("vocal_measures_dummy") contains placenta volume measurements made using a commercial software package.
- The third ("placenta_measures_dummy") contains a collection of placenta measurements derived from custom 3D ultrasound image analysis.
Information from the three spreadsheets is merged based on a study identifier. Please refer to the final HTML document submitted to Canvas for the results generated using real clinical data. The results calculated with the dummy spreadsheets does not match the final analysis.