Image Detection of Red Colors(Blood) in Stool Images

Names:

• Nick DeCapite
• Jin Zhou

Timeline

Start Date: May, 2020

Process

• Images

  • Blood
  • Concerning
  • Good

• CSV Files

  • For Every Good and Blood image, the pixel concentration for each color in the ISCC system is recorded.
  • Done for all 267 colors and all 47 "red" colors.

• PCA Dimension Reduction

  • The color dimensions(267 or 47) are reduced to 10.

• KNN Analysis

  • Based on the reduced dimensions, the Good and Blood images are compared in order to create the the training and test models.
  • Currently, the highest accuracy at detecting the presence of blood is 72.22%.

Future Steps

I will be working on fine-tuning the input paramaters to have a more accurate model. In addition, I will be testing out other machine learning techniques on blood detection.

Other than blood detection, I will be working with the concerning color images in order to create a process for the detection of these images.