Download presentation
Presentation is loading. Please wait.
Published byAgatha Webster Modified over 8 years ago
1
A Neural Network Approach for classifying TACS By Mike Smith
2
Personal Background ECE Master’s student Research Assistant for the Laboratory for Optical and Computational Instrumentation (LOCI) –Design Control System for Laser Scanning Microscopes
3
Project Overview Classifying TACS (Tumor Associated Collagen Signatures) –Change in density and alignment during tumor development –Signatures can be seen before a tumor is formed allowing for early detection of cancer
4
Data Gathering Data is gathered using multiphoton laser scanning microscope Collagen produces a second harmonic effect naturally –Basically shine a laser on collagen, it will glow and we can capture that and form an image
5
Training/Testing Data -Classified data from images of tumors from a mouse mammary -Broken up into 32x32 discrete chunks
6
Current Techniques Classify intensity or average intensity of a section of data using artificial neural networks –Naive approach –Haven’t been happy with results –Gives baseline though Classify based on change in intensity and consistency with areas around it
7
Future Work Classify based on raw data, not images –Only 8 bit pixels, ADC provides 12 bit resolution Try to predict signatures before tumor is formed -Early Cancer Detection
8
References P. P. Provenzano, et al., "Collagen reorganization at the tumor-stromal interface facilitates local invasion". BMC Med. 4, 38 (2006). www.loci.wisc.edu
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.