Download presentation
Presentation is loading. Please wait.
Published byFay Ward Modified over 9 years ago
1
Small Intestine Villi Cell Counting Meghan Olson & Jittapat Bunnag
2
Background/Motivation Research on Anemia (body can’t sufficiently distribute oxygen=> lack of iron) Count cells that absorb iron (blue) Currently counted manuallyCurrently counted manually Time consuming Inaccurate (different people count differently) Automated program would be ideal Standardized counting procedureStandardized counting procedure Less time consuming to researcherLess time consuming to researcher
3
Goal Count number of iron-absorbed (blue) cells Measure blue and red area, and calculate percentage
4
Complexity Cell size and color vary Count Cells only – not light transparent parts Inconsistency in image Brightness Contrast Color cast Sharpness Images are unique Hard to find algorithm that is compatible with all images
5
Algorithm Preprocessing Separate RGB channels set thresholdsSeparate RGB channels set thresholds Eliminate background and Villi edge Separate Red and Blue cells Count Blue Cells Count # of connected regionCount # of connected region Calculate Percentage Calculate area of Blue and Red cellsCalculate area of Blue and Red cells
6
Preprocessing Transform (stretch) image Better contrastBetter contrast Can separate blue and red cells from image using Green channel Problem: edgeProblem: edge Eliminate background & edge ~White=> R=B=G~White=> R=B=G Dilate & Subtract from original imageDilate & Subtract from original image EDGE Stretched Background extraction
7
Preprocessing Red Extraction Create binary templatesCreate binary templates Red>Blue Red>Green Combine binary templatesCombine binary templates “And” w/ No-background image“And” w/ No-background image Blue cells Red cells Blue Extraction Create binary templatesCreate binary templates Red<Blue Red<Green Combine binary templatesCombine binary templates “And” w/ No-background image“And” w/ No-background image
8
Works on Variety of images
9
Counting Convert the processed images into binary images Eliminate small regions and close small gaps that humans cannot distinguish Use built-in MATLAB functions to count the area, and the number of cells Area = 9246 Cells = 110 Area = 175018 Cells = 390
10
Algorithm Overview Background Extraction Binary image Stretching Transformation Blue pixels Red pixels Original Image Blue Extraction Red Extraction Calculate: Cell number Area Percentage Dilation/Erosion intersect dilate
11
Graphical User Interface User selects Magnification of Image Different algorithm for different magnificationDifferent algorithm for different magnification User selects area of image or entire image to be processed Results are displayed in graphical form Results include: Cell areaCell area Cell countCell count PercentagePercentage
12
Questions?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.