Traditional methods for detection and enumeration of microbial growth in food stuff are very time consuming, destructive, invasive, complex, expensive and risky especially in case of pathogenic microbes. Therefore, the experimental work comprehensively detailed in this study was carried with the aim of non-destructive detection and quantification of fungal growth in bread using spectral analysis as one of the most promising techniques. For a period of seven consecutive days, spectral images in the near infrared (NIR) range were acquired for freshly-backed bread samples. Concurrently, the corresponding mould growth was monitored and assessed with the standard plating methods. Spectral data extracted from the images of bread samples and their reference mould counts during the storage period were modelled using multivariate statistical models. The principal component analysis (PCA) indicated that bread samples at the first four days consistently had similar spectral fingerprint and projected at the same location in the principal component plot. Starting from the fifth day, bread samples exhibited extraordinary spectral behaviour. Moreover, results demonstrated good prediction of mould counts in calibration and validation sets of bread samples (= 0.97 and  = 0.94). The results presented in this work revealed that the biochemical fingerprints during fungal invasion conveyed by NIR spectral images in combination with the appropriate multivariate analysis strategy have significant potential for rapid assessment of bread spoilage.