سال انتشار: ۱۳۹۰

محل انتشار: هفتمین کنفرانس ماشین بینایی و پردازش تصویر

تعداد صفحات: ۴

نویسنده(ها):

Sahar Jahani – Faculty of Biomedical Engineering, Science and Research Branch, Azad University, Tehran, Iran
S. Kamaledin Setarehdan – Control and Intelligent Processing Centre of Excellence, School of Electrical and Computer Eng.,Faculty of Engineering, University of Tehran, Tehran, Iran
Emadedin Fatemizadeh – Sharif University of Technology, School of Electrical Engineering

چکیده:

Karyotyping, is the process of classification of humanchromosomes within the microscopic images. This is a commontask for diagnosing many genetic disorders and abnormalities.Automatic Karyotyping algorithms usually suffer the poorquality of the images due to the non rigid nature of thechromosomes which makes them to have unpredictable shapesand sizes in various images. One of the main problems thatusually need operator’s interaction is the identification andseparation of the overlapping/touching chromosomes. Thispaper presents an effective algorithm for identification of anycluster of the overlapping/touching chromosomes together withthe number of chromosomes in the cluster, which is a very firststep towards the development of a fully automatic Karyotypingsystem. The proposed algorithm which is based on theextraction of the number of endpoints within the skeleton of theimage objects uses morphological operators. Two independentdatasets obtained from the Tesi-Imaging srl in Milan, Italy andthe Imam Hospital in Tehran, Iran was used to evaluate theperformance of the algorithm. An accuracy of %96 and %99were obtained on identification of the clusters ofoverlapping/touching chromosomes and single chromosomesrespectively by the proposed algorithm