Computer Assisted Analysis of the Hepatic Spheroid Formation

Abstract

Computer assisted techniques could enable the use of morphological characteristics of hepatic spheroids as surrogate for their response to various stimuli. The aim of this work is to develop an automatic analysis procedure able to correctly acquire all the important morphological parameters of the hepatic spheroids under static and after stimulation conditions. Within the datasets, several issues can occur related to the non-uniform illumination background and inherent limitation in counting the exact object number when two or more are adjacent. Some of the images include patterns with intensity comparable to the spheroid intensity, such as extended grooves, and they are filtered out based on their eccentricity values. Traditional methods such as Otsu threshold does not segment the spheroid images truthfully due to their energy minimization based approach. To circumvent this limitation initially background removal is applied as a preprocessing step. Filters applied for this depend on the relative size of the spheroids in order not to diminish the image quality. Therefore, we propose a guided automatic threshold value that can discriminate between the background and the spheroids more accurately by finding the critical peak on the pixel intensity histogram. Pixel intensity histograms are composed of three modes and the local minimum after the peak at the lowest values is the threshold value. After applying the new guided thresholding technique, watershed algorithm is used in order to determine the separating nodes between objects that are contiguous to each-other. These two techniques are compared with Gabor filters based methods that are shape based filters. Employing the three methods spheroid parameters including the number, area and the perimeter were determined and their performance and robustness are discussed.