Automated identification of analyzable metaphase chromosomes depicted on microscopic digital images

Abstract

Visual search and identification of analyzable metaphase chromosomes using optical microscopes is a very tedious and time-consuming task that is routinely performed in genetic laboratories to detect and diagnose cancers and genetic diseases. The purpose of this study is to develop and test a computerized scheme that can automatically identify chromosomes in metaphase stage and classify them into analyzable and un-analyzable groups. Two independent datasets involving 170 images are used to train and test the scheme. The scheme uses image filtering, threshold, and labeling algorithms to detect chromosomes, followed by computing a set of features for each individual chromosome as well as for each identified metaphase cell. Two machine learning classifiers including a decision tree (DT) based on the features of individual chromosomes and an artificial neural network (ANN) using the features of the metaphase cells are optimized and tested to classify between analyzable and un-analyzable cells. Using the DT based classifier the Kappa coefficients for agreement between the cytogeneticist and the scheme are 0.83 and 0.89 for the training and testing datasets, respectively. We apply an independent testing and a 2-fold cross-validation method to assess the performance of the ANN-based classifier. The area under and receiver operating characteristic (ROC) curve is 0.93 for the complete dataset. This preliminary study demonstrates the feasibility of developing a computerized scheme to automatically identify and classify metaphase chromosomes.

Figure 1

Digital images of two metaphase chromosome cells in which (a) is considered un-analyzable cells that will be deleted and (b) is an analyzable cell that will be selected to perform karyotyping.

Figure 2

A flow diagram of a computerized scheme to segment chromosomes and classify metaphase cells into analyzable and un-analyzable cells.

Figure 3

A five-feature based DT for recognizing analyzable and un-analyzable metaphase chromosome cells. Note: F1 – Average size of each region; F2 – Circularity of each region; F3 – Average gray value of each region; F4 – Radial length of each region; Th1 - Number of regions is between N_min = 19 and N_max = 46; Th2 – Average size of each region is between S_min = 90 and S_max = 5000; Th3 – Circularity of each region is < C_T = 0.9; Th4 - Average gray value of each region is ≥ I_T = 75; Th5 - Number of regions is either < N_min = 19 or > N _max = 46; Th6 - Average size of each region is either <S_min = 90 or >S _max = 5000; Th7 - Circularity of each region is ≥ C_T = 0.9; Th8 - Average gray value of each region is < I_T = 75; Th9 - Radial length of each region is ≥ L_k (the maximum radial length – standard deviation of radial length of all labeled regions); and Th10 - Radial length of each region is < L_k.

Figure 4

A scatter diagram between two features of 100 training samples including 35 analyzable (“positive”) and 65 un-analyzable (“negative”) cells.

Figure 5

A scatter diagram between two features of 70 testing samples including 37 analyzable (“positive”) and 33 un-analyzable (“negative”) cells.

Figure 6

The distribution of the computed performance points for two data subsets (“training” and “test” subsets) and the complete dataset. A ROC-type performance curve was generated based on fitting of the complete dataset using ROCFIT program.