Our study is focused on the development of a new method for the automatic analysis of cell images. We focused on neurons (cells line SH-SY5Y) treated/untreated with ultrasound and stained with Haematoxylin-Eosin. The aim of the algorithm is the automatic detection of the cell body as well as the determination of the number and the length of neuron elongations. Starting point of the algorithm was the convolution of an image with a bank of rotating Gaussian kernels and the construction of a module map. Then several strategies were implemented to detect cell bodies and to detect and extract data about cell elongations. We have also realized a graphical user interface allowing the loading, saving and processing of images. Results show that this method is able to properly and efficiently detect cell contours and elongations. The automated evaluation is in strong agreement with manual evaluation performed by an expert operator, with an average error of 11% with most parameter combinations. This tool constitutes an important support in biological research activities, where operators need to analyze a large number of images to investigate about cell morphology before and after a treatment.
Morphological analysis of neurons: Automatic identification of elongations
BRANCA, JACOPO JUNIO VALERIO;MORUCCI, GABRIELE;
2015-01-01
Abstract
Our study is focused on the development of a new method for the automatic analysis of cell images. We focused on neurons (cells line SH-SY5Y) treated/untreated with ultrasound and stained with Haematoxylin-Eosin. The aim of the algorithm is the automatic detection of the cell body as well as the determination of the number and the length of neuron elongations. Starting point of the algorithm was the convolution of an image with a bank of rotating Gaussian kernels and the construction of a module map. Then several strategies were implemented to detect cell bodies and to detect and extract data about cell elongations. We have also realized a graphical user interface allowing the loading, saving and processing of images. Results show that this method is able to properly and efficiently detect cell contours and elongations. The automated evaluation is in strong agreement with manual evaluation performed by an expert operator, with an average error of 11% with most parameter combinations. This tool constitutes an important support in biological research activities, where operators need to analyze a large number of images to investigate about cell morphology before and after a treatment.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
