Result 1: Beef Spoilage VOC database

Development of sensors sensitive to whole groups or classes of chemicals will be preferred in TOXDTECT project. Their global response will be related to packed meat characteristics and will allow further adaptations to other types of meat and meat products. Experimental analysis will be carried out in order to identify classes of metabolites representative of beef degradation in MAP conditions. In order to obtain useful conclusions from the expected large number of experimental measurements in intricate systems such as food degradation, we will perform MVA (multivariable analysis) of the data sets. The Matlab® analysis program will be used for pattern recognition. Among various MVA techniques, Principal Component Analysis (PCA) algorithms will be used to transform variable inputs into small numbers of non-correlated variables, further allowing 2D and 3D graphical representations of the different principal components. In order to classify the measurements and to ensure correlation between the data collected and the time or day of their collection, the Adaptive Resonance Theory will be used.

We will determine aging and product spoilage using Gas/VOC monitoring of the head space of the packaged product. Variations in oxygen or carbon dioxide concentration of MAP are markers of bacterial respiration whilst volatile alcohols or aldehydes evidence the oxidation of fibrous meat tissues. Indeed biogenic amines such as triethylamines or other volatile amines such as putrescine (1,4-diaminobutane) as well as ATP degradation markers have been demonstrated as pertinent markers of meat product degradation.

In order to study the specific VOCs related to meat spoilage, we will examine the compounds released by the following microorganisms (SSO) associated with beef spoilage: Salmonella spp, L. Monocytogenes, Mesophilic aerobes, Clostridium spp, E. coli, S. aureusand Pseudomonas spp, yeast and molds. As a result, we will advance the State of the Art in predictive models for meat spoilage by specifically developing a database of volatile biomarkers for beef (quantitative and qualitative) and by establishing correlations between the expression of these markers, bacterial population growth and packaging conditions.

Result 2: An innovative Beef Package consisting of a functionalized multilayer film containing a printed array of multi-analyte sensors, selected after a deep analysis of the most representative classes of metabolites associated with beef spoilage. An electronic nose sensing system will be developed advancing in the State of the Art by allowing the characterization of complex attributes of the whole beef sample rather than the determination of a singular analyte. To this end, printed conducting polymers and their functionalisation with metal oxide nanoparticles will be investigated. Polyaniline will be studied for RH determination, polythiophene derivatives for aldehyde and ketone determinations, and immobilization of diamine oxidase onto membrane electrodes for alkyldiamine determination.

The developed film will comply with current food legislations and will have the properties indicated in table 6below. The packaging will be designed to ease the interface between the sensors and external reader. In order to ease the use of the end product, the packaging film will be designed with a protruding flip strip on which connectors will be available to allow soft contact between the sensors and the external reader.

Result 3: An external reading device that will collect the information provided by the sensors and will interpret the data to give a predictive shelf life. The system will be supported by 2 components: (i) an intelligent decision-making technology based on Artificial Neural Networks (ANNs) and (ii) the external reading device. The electrical change registered by the sensors will be read by an external reading device that will interpret the sensors signal and will display the product´s remaining shelf life by means of an electronic decision making system based on predictive algorithms.

  • Predictive Data Mining Models: We will apply data mining techniques to identify correlations between input parameters and the output of the intelligent system. Input to the classification system will be; the relative abundances of gases in the meat samples, presence of metabolites and information about the temperature, pH and moisture. The output of the classification system will be the current quality state of the meat and predictions about its shelf-life.
  • Reading Device components: This includes the interface with the package sensors, the conditioning electronics of the sensors, the microcontroller running the decision software (SW), the display, the control SW of the device and the user interface of the application. All these elements must be miniaturised so that they can be embedded into a small device similar to a bar code reader gun. The same platform will be initially developed for a portable reader but further adaptation for a fixed reading station will also be possible.
  • Development of Intelligent Decision software: Once the data mining models have been designed, the extracted knowledge (i.e. the correlation links) is used to implement an intelligent system that will measure, infer and control in real-time the input parameters to provide the user with a meat quality estimation. In this task, the components of the system are defined and the software implementation is completed. Our initial approach will use Artificial Neural Networks. Other software architectures may be investigated as well.

The proposed system will be designed to comply with both the requirements of the SME-AG
Communities represented in TOXDTECT (guaranteed product safety and recovery of lost beef sales due to fear to food diseases; innovative product, low production cost packaging film) and with current EC regulations (food contact materials and product traceability) and will satisfy the demand of end users to ensure product quality.


Identification of target metabolites representative of beef meat spoilage. Among existing indicators of meat spoilage (putrescine, triethylamines, ATP degradation markers, O2, CO2.)


Design of an electronic decision making system (software) for meat quality assessment based on predictive algorithms Implementation and successful testing of the predictive system with real meat in packages.