Rapid assessment of fish freshness for multiple supply-chain nodes using multi-mode spectroscopy and fusion-based artificial intelligence

Authors: ZADEH, HOSSEIN - Safetyspect Inc; HARDY, MIKE - University Of North Dakota; SUEKER, MITCHELK - University Of North Dakota; LI, YICONG - University Of North Dakota; TZOUCHAS, ANGELIS - Safetyspect Inc; MACKINNON, NICHOLAS - Safetyspect Inc; BEARMAN, GREGORY - Safetyspect Inc; HAUGHEY, SIMON - University Of North Dakota; AKHBARDEH, ALIZEREZA - Safetyspect Inc; BAEK, INSUCK - Orise Fellow; HWANG, CHANSONG - University Of Maryland; Qin, Jianwei - Tony Qin; TABB, AMANDA - Chapman University; HELLBERG, ROSALEE - Chapman University; ISMAIL, SHEREEN - University Of North Dakota; REZA, HASSAN - University Of North Dakota; VASEFI, FARTASH - Safetyspect Inc; Kim, Moon; TAVAKOLIAN, KOUHYAR - University Of North Dakota; ELLIOTT, CHRISTOPHER - Thammasat University

Submitted to: Sensors
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/19/2023
Publication Date: 5/28/2023
Citation: Zadeh, H., Hardy, M., Sueker, M., Li, Y., Tzouchas, A., Mackinnon, N., Bearman, G., Haughey, S., Akhbardeh, A., Baek, I., Hwang, C., Qin, J., Tabb, A.M., Hellberg, R., Ismail, S., Reza, H., Vasefi, F., Kim, M.S., Tavakolian, K., Elliott, C.T. 2023. Rapid assessment of fish freshness for multiple supply-chain nodes using multi-mode spectroscopy and fusion-based artificial intelligence. Sensors. 23:5149. https://doi.org/10.3390/s23115149.
DOI: https://doi.org/10.3390/s23115149

Interpretive Summary: Fresh fish is a highly perishable product with more than 20% wasted at retail level every year. One reason for such waste is that early fish decay is not easily detectable by human senses. There is a need for sensing techniques that allow for onsite inspection of fish freshness in a rapid, cost-effective, and nondestructive manner. This study developed a multimode spectroscopy method for rapid assessment of the fish freshness. Three types of spectral data (i.e., visible and near infrared reflectance, short wave infrared reflectance, and fluorescence) were collected with time from fish fillets of four species (i.e., farmed Atlantic salmon, wild coho salmon, Chinook salmon, and sablefish). Machine learning methods using the fusion of the three spectral data types achieved 95% accuracy to classify fish from fresh to spoiled condition. The data analysis and classification methods developed in this research can be used to assist development of an easy-to-use handheld device to estimate remaining shelf life of the fish fillets, which can lead to major waste reduction by allowing dynamic sales management for the seafood industry.

Technical Abstract: Fish is highly perishable, with 21.3% of fish and 24.1% of shellfish wasted at retail level. This research is directed towards developing a fast, non-destructive, and easy-to-use handheld multimode spectroscopic system for fish quality assessment. It uses the fusion of visible near infra-red (VIS-NIR), short wave infra-red (SWIR) reflectance and fluorescence (FL) spectroscopy data features to classify fish from fresh to spoiled condition. Farmed Atlantic Salmon Salmo salar, wild Coho Salmon Oncorhynchus kisutch, Chinook Salmon Oncorhynchus tshawytscha and Sablefish Anoplopoma fimbria fillets were measured with time. Different machine learning techniques such as principal component analysis, self-organized maps, linear/quadratic discriminant analyses, k-nearest neighbors, random forest, support vector machine, linear regression and ensemble and majority voting methods, were used to explore data and train classification models to predict freshness. Results show multimode spectroscopy achieves 95% accuracy, improving the accuracies of the FL, VIS-NIR and SWIR single-mode spectroscopies by 26, 10 and 9%, respectively.