Automated Detection of Porcine Gelatin Using Deep Learning-Based E-Nose to Support Halal Authentication

Abstract

Authenticating gelatin sources is essential for consumers, particularly those with dietary restrictions or religious concerns regarding pork-derived ingredients. Porcine gelatin, widely used in food and pharmaceutical products, poses considerable challenges for authentication due to its prevalence and the difficulty of detecting it, especially in processed products. In this study, we developed and evaluated an integrated electronic nose (e-nose) system with a Recurrent Neural Network (RNN) to detect and classify gelatin type based on their sources. The e-nose system utilized an array of gas sensors to capture the unique volatile organic compounds (VOCs) associated with each gelatin type, which was subsequently classified by the RNN. The classification performance of the integrated 7-module e-nose system showed promising results based on time points after sample preparation, with accuracy, sensitivity, and AUC of 96.3%, 96.6%, and 98.2% at the 0-hour point, respectively, rising to 99.1% for all three metrics at 2-hour point. The sensitivity of the system also showed an increase over time for single gelatin samples, from 100%, 97.8%, and 91.9% to 98.6%, 99.3%, and 99.3% for pig-derived, cow-derived, and fish gelatin, respectively. For mixed gelatin samples, the system maintained high accuracy, sensitivity, and AUC at 98.2%, 97.9%, and 98.1%, respectively. In conclusion, the integrated e-nose system demonstrates the potential for robust performance in gelatin authentication, paving the way for more efficient and reliable methods of halal food authentication.