ORIGINAL PAPER Hyperspectral imaging for the investigation of quality deterioration in sliced mushrooms (Agaricus bisporus) during storage A. A. Gowen Æ C. P. O’Donnell Æ M. Taghizadeh Æ E. Gaston Æ A. O’Gorman Æ P. J. Cullen Æ J. M. Frias Æ C. Esquerre Æ G. Downey Received: 17 December 2007 / Accepted: 25 February 2008 / Published online: 14 March 2008 Ó Springer Science+Business Media, LLC 2008 Abstract In this study, the potential application of hyperspectral imaging (HSI) for quality prediction of white mushroom slices during storage at 4 °C and 15 °C was investigated. Mushroom slice quality was measured in terms of moisture content, colour (CIE Lightness, L* and yellowness, b*) and texture (hardness, H and chewiness, Ch). Hyperspectral images were obtained using a push- broom line-scanning HSI instrument, operating in the wavelength range of 400–1,000 nm with spectroscopic resolution of 5 nm. Multiple linear regression (MLR) and Principal Component Regression (PCR) models were developed to investigate the relationship between reflec- tance and the various quality parameters measured. 20 optimal wavelengths for quality prediction were selected after performing an exhaustive search for the best subsets of predictor variables on a calibration set of 84 samples. PCR applied to the set of optimal wavelengths gave the best performance as compared to MLR and PCR on the entire wavelength range. When applied to an independent validation set of samples, PCR models developed on the calibration set were capable of predicting moisture content with RMSEP of 0.74% w.b. and R 2 of 0.75, L* with RMSEP of 0.47 and R 2 of 0.95, b* with RMSEP of 0.66 and R 2 of 0.75, H with RMSEP of 0.49 N and R 2 of 0.77 and Ch with RMSEP of 0.27 N and R 2 of 0.72. Virtual images showing the distribution of moisture content on the mushroom surface were generated from the estimated PCR model. Results from this study could be used for the development of a non-destructive monitoring system for prediction of sliced mushroom quality. Keywords Mushroom  Slice  Agaricus  Bisporus  Hyperspectral Introduction Hyperspectral Imaging, also known also as Chemical or Spectroscopic Imaging, is an emerging technique that integrates conventional imaging and spectroscopy to attain both spatial and spectral information from an object. Because of this integrated feature of imaging and spec- troscopy, hyperspectral imaging can expand our capability for detecting some chemical constituents and intrinsic characteristics of an object as well as their spatial distri- butions. It was originally developed for remote sensing applications [1] but has since found application in such diverse fields as astronomy [2, 3], agriculture [4–6], phar- maceuticals [7–9], food [10–12] and medicine [13–15]. Hyperspectral images, known as hypercubes, are three dimensional blocks of data, comprising of two spatial and one wavelength dimension. Each pixel of the hypercube has an associated spectrum; equally, each plane of the hypercube represents the hyperspectral image at a partic- ular wavelength. Hypercubes are data rich; for example, the hyperspectral imaging system employed in this study, which operates in the wavelength range of 400–1,000 nm, with spatial A. A. Gowen (&)  C. P. O’Donnell  M. Taghizadeh Biosystems Engineering, School of Agriculture, Food Science and Veterinary Medicine, University College Dublin, Dublin 4, Ireland e-mail: aoife.gowen@ucd.ie E. Gaston  A. O’Gorman  P. J. Cullen  J. M. Frias School of Food Science and Environmental Health, Dublin Institute of Technology, Cathal Brugha Street, Dublin 1, Ireland C. Esquerre  G. Downey Teagasc Ashtown Food Research Centre, Ashtown, Dublin 15, Ireland 123 Sens. & Instrumen. Food Qual. (2008) 2:133–143 DOI 10.1007/s11694-008-9042-4