Loading...

Improvement and Implementation of Bio-Inspired Model HMAX in order to Recognize Red Blood Cells Morphology

Arabshahi, Soroush | 2016

615 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 48377 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Vosoughi Vahdat, Bijan; Gholampour, Eiman
  7. Abstract:
  8. Blood is the most important liquid in human body. Red blood cells (RBCs) are the main part of blood. RBCs are rounded, but this shape would be changed during different disease. Blood tests taken to diagnose blood diseases. Current instruments do not provide any information about RBCs morphology, and an expert recognize the RBCs morphology by looking at blood smear under microscope. Regards to low accuracy and performance of this method, an automate method must be presented. Capturing images from blood smear under microscope is simple and low-cost. Although there is few methods on RBCs detection and very limited one on RBCs recognition on the literature, no general method to recognize most of the morphological abnormities for RBCs has been represented. Currently, for object recognition, bio-inspired methods achieved great success. HMAX model is one of the known methods on this manner. HMAX simulate the hierarchical structure of visual cortex on mammals’ brain based on biological facts. To use bio-inspired methods, like HMAX, on RBCs’ morphology recognition, the issue should alter to object recognition problem. Another way is to manipulate the model regarding the shape concerns. Shapes of RBCs are their identifier and it could vary in a same category. In addition, there is some peripheral distortion causing the RBCs’ morphology change. This will convert RBCs morphology recognition to object recognition problem. On this presentation, we merged Active Shape Model (ASM) ability for shape analysis and HMAX ability for recognition in order to classify RBCs morphology. Besides, we made a few manipulation on HMAX model algorithm and active shape model target point finder. Finally, we showed that improved method of HMAX and combination with active shape model will result in better performance and accuracy on RBCs’ morphology recognition, in comparison to other known feature extractor like Scaled Invarient Feature Transfomr and Histogram of Oriented Gardient . Also changing active shape model point finder will lead low usage of resources and make it possible to be used on embedded systems
  9. Keywords:
  10. Active Shape Model ; Feature Extraction ; Red Blood Cell ; Morphology ; RECOGNITION ; HMAX Model ; Visual Cortex ; Red Blood Cell Merphology

 Digital Object List

 Bookmark

No TOC