Computer Methods and Programs in Biomedicine (2005) 79, 197—208 An automated procedure to properly handle digital images in large scale Tissue Microarray experiments Rossana Dell’Anna a,* , Francesca Demichelis a,c , Mattia Barbareschi b , Andrea Sboner a,c a ITC-irst, SRA Division, Bioinformatics Group, Via Sommarive 18, 38050 Povo, Trento, Italy b Department of Surgical Pathology, Santa Chiara Hospital, L.go Medaglie D’Oro 9, 38100 Trento, Italy c Department of Information and Communication Technology, Trento University, Via Sommarive 16, 38050 Povo, Trento, Italy Received 20 September 2004; received in revised form 19 April 2005; accepted 28 April 2005 KEYWORDS Automation; Image processing; Tissue Microarray Summary Tissue Microarray (TMA) methodology has been recently developed to enable ‘‘genome-scale’’ molecular pathology studies. To enable high-throughput screening of TMAs automation is mandatory, both to speed up the process and to improve data quality. In particular, in acquiring digital images of single tissues (core sections) a crucial step is the correct recognition of each tissue position in the array. In fact, further reliable data analysis is based on theexactassignmentofeachtissuetothecorrespondingtumor.Asmostofthetimes tissue alignment in the microarray grid is far from being perfect, simple strategies to perform proper acquisition do not fit well. The present paper describes a new solution to automatically perform grid location assignment. We developed an ad hoc image processing procedure and a robust algorithm for object recognition. Algorithm accuracy tests and assessment of working constraints are discussed. Our approach speeds up TMA data collection and enables large scale investigation. © 2005 Elsevier Ireland Ltd. All rights reserved. 1. Introduction Characterization of gene-expression profiles with DNA microarrays provides a new powerful tool to discover disease-related genes, particularly in the ∗ Corresponding author. Tel.: +39 0461 405 312; fax: +39 0461 405 372. E-mail address: dellanna@itc.it (R. Dell’Anna). case of cancer [1—3]. In situ expression analysis of most discriminating genes is a further development whichcanleadtotheidentificationofnewbiomark- ers, which in turn can be valuable for diagnostic or prognostic purposes or which can be used to predict the response to therapy [4]. It is known that traditional in situ molecu- lar analysis on individual tissue sections needs enormous efforts in terms of time and costs to 0169-2607/$ — see front matter © 2005 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.cmpb.2005.04.004