Title: AI tools in medical image analysis: efficacy of ANN for oestrogen receptor status assessment in immunohistochemical staining of breast cancer

Authors: Mohamed Ali Cherni; Mounir Sayadi; Farhat Fnaiech

Addresses: ESSTT, University of Tunis, 5 Av. Taha Hussein, B.P. 56, 1008 Tunis, Tunisia ' ESSTT, University of Tunis, 5 Av. Taha Hussein, B.P. 56, 1008 Tunis, Tunisia ' ESSTT, University of Tunis, 5 Av. Taha Hussein, B.P. 56, 1008 Tunis, Tunisia

Abstract: Evaluating oestrogen receptor status in immunohistochemical staining of breast cancer is so complicated. This process is done subjectively and is so much time consuming. In fact, the studied images present many characteristics such as the non uniformity in the intensity of the organic tissue and the cells, and also the variability of the size and the form of cells which make their processing so difficult. So, given all these complexities, conventional methods are unable to solve the problem. In this work, we study the ability of artificial intelligence as a modern and an unconventional technique to automatically classify breast cancer cells. This step is considered as primary in assessing oestrogen receptor status. Three intelligent techniques are presented, applied and compared: fuzzy c-means (FCM), artificial neural network (ANN) and genetic algorithm (GA). The statistical analysis demonstrates the efficacy of the artificial neural network by recording an average rate of sensitivity, specificity and accuracy exceeding 97% versus a lower average rate of each evaluation criteria for each of the two other techniques.

Keywords: artificial intelligence; artificial neural networks; ANNs; breast cancer; evolutionary algorithms; fuzzy c-means; genetic algorithms; granulometry; mathematical morphology; medical images; image analysis; watershed; oestrogen receptor status; immunohistochemical staining; cancer cell classification.

DOI: 10.1504/IJBET.2013.056285

International Journal of Biomedical Engineering and Technology, 2013 Vol.12 No.1, pp.60 - 83

Received: 29 Nov 2012
Accepted: 25 Jun 2013
Published online: 27 Sep 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article