Sensors and Actuators B 111–112 (2005) 293–298 A model to predict fish quality from instrumental features A. Macagnano a , M. Careche b , A. Herrero b , R. Paolesse a,c , E. Martinelli d , G. Pennazza d , P. Carmona e , A. D’Amico a,d , C. Di Natale a,d,∗ a CNR–Institute of Microelectronics and Microsystems, Via Fosso del Cavaliere, 00133 Rome, Italy b Istituto del Frio, CSIC, Av. Jose Antonio Novais, 10 Ciudad Universitaria, 28040 Madrid, Spain c Department of Chemical Science and Technology, University of Rome “Tor Vergata”, Via della Ricerca Scientifica, 00133 Rome, Italy d Department of Electronic Engineering, University of Rome “Tor Vergata”, Via della Ricerca Scientifica, 00133 Rome, Italy e Istituto de Estructura de la Materia, CSIC, Ciudad Universitaria, Serrano 121, 28006 Madrid, Spain Available online 16 August 2005 Abstract Sensorial evaluation of fishes, using a well-defined scheme including several qualitative attributes, can give a reliable quantitative evaluation of freshness (quality index method, QIM). The introduction of artificial instruments, mimicking human senses, seems a promising approach to obtain a comparable judgment with trained panels one. The outputs of colour, texture and electronic nose measurements can be compared and combined by data fusion, to construct an artificial quality index (AQI), describing the quality of fish at least as well as the QIM predicted. In this paper, the application of such approach to build a fish freshness indicator of sardine fishes is illustrated. © 2005 Published by Elsevier B.V. Keywords: Quality index method (QIM); Artificial quality index (AQI); Partial least square (PLS) 1. Introduction A current regulation of the European Community (1996) establishes the principles to control and certify the qual- ity warranty in the seafood field [1]. Precisely, the quality attributes include a series of parameters related to safety, nutritional quality, availability, freshness and edibility [2], which may be affected mainly by handling, processing and storage procedures from the catch to the consumers. Prac- tically, physical, chemical, biochemical and microbiological changes occurring post-mortem in fishes, result in a progres- sive lost of food characteristics in terms of taste and a general concept of quality [3]. Both the passed time and the temper- ature storage of fish are key factors for the final quality of the product. Fish spoilage depends in fact, mostly on temper- ature, which controls to a large extent the bacterial and the autolytic breakdown. Moreover, the rate of spoilage depends on several factors such as the kind of fish species, the sanitary conditions on board, and the amount of food in the guts. Fish ∗ Corresponding author. Tel.: +39 06 7259 7348; fax: +39 06 2020 519. E-mail address: dinatale@uniroma2.it (C.D. Natale). freshness is the most required property from the consumers because of its strong relationship to the taste. A number of sensorial inspection procedures have been introduced to point the state of freshness. These procedures involve the use of sight (to evaluate the skin appearance and the colour and the global aspect of eyes), tactile (to test the flesh firmness and elasticity) and olfaction (to smell the gill odour). A trained panel can evaluate all those sensory attributes depending on fish freshness, i.e. appearance, texture, smell, colour and gill status. Each attribute is translated in a numeric scale using a demerit score system [2]. All the attributes are merged together by a simple summation of the demerit scores. All the method has been dubbed as quality index method (QIM). QIM was introduced for codfishes; nonetheless, the same approach can be utilized for other species. It is con- structed in a way that it shows a linear evolution with the number of storage days in ice [4]. However, trained panels are generally expensive, time consuming, and not always available along the different steps of the fishery chain. Consequently, to satisfy the need for qual- ity measurements in the fish industry, instrumental methods are needed. 0925-4005/$ – see front matter © 2005 Published by Elsevier B.V. doi:10.1016/j.snb.2005.06.028