Chromosome identification and pairing in optical microscopy images

From ISRWiki
Jump to navigation Jump to search
Paired chromosomes using the Lisbon Chromosome Dataset.
  • Supervisor: Prof. João Sanches
  • Co-Supervisor: Prof. Rodrigo Ventura
  • joint work with Prof. Carmo Fonseca

Objectives

In this work the data is the output of a karyotiping commercial package which, with medical assistance, allows the segmentation of the 46 chromosomes and a first paring guess. This first paring iteration usually leads to wrong paring results that must be corrected by the medical doctor. The core of this work in the improvement of this pairing task, by acting in two aspects: 1. Adaptative debluring - Pre-processing of the segmented chromosomes images, by adaptatively deblurr them in order to provide better data to the classifier used in the paring procedure. 2. Pairing - Features extraction from the pre-processed images to be used in the paring phase.

Description

Cell malignancy can be detected by observation of structural/morphologic and pattern abnormalities in the chromosomes during the metaphase of the mitosis. This powerful approach to detect genetic diseases is extensively used since the seventies when Seabright (1972) has introduced the G-banding, allowing the recognition of each individual chromosome. Since that time a significant effort has been done to develop automatic procedures to recognize pair and classify the 46 chromosomes observed during the metaphase. Along this phase they are more compacted, concentrated and visible, presenting, each one, a pattern characteristic that can be used to their recognition. Well known diseases detected by this method are the several trisomies, the most known is the trisomy 21, monosomy and several types of acute leukaemia. This procedure, however, is difficult and must be performed with an intensive human assistance because the images are blurred by a space varying PSF along the image and superposition and occlusions occurred very often. The paring procedure, in particular, is one of the most difficult steps in the whole processing pathway because homologous chromosomes may present very different pattern due distortion or blurring, and sometimes, due genetic diseases.

Links