A novel characterization of starch gelatinization using microscopy observation with deep learning methodology

作者：Niu, Yahui; Zheng, Yong; Fu, Xiong; Zeng, Delu^*; Liu, Hongsheng^*

来源：Journal of Food Engineering, 2022, 327: 111057.

DOI：10.1016/j.jfoodeng.2022.111057

• Birefringence
• Computer vision
• Gelatinization
• Microscopy observation
• Starch granules

Starch gelatinization under microscopy shows a continuous change of granule swelling and birefringence number whereas is yet inspected by eyes with inevitable subjective errors. In this study, a novel microscopy observation was developed for characterizing starch swelling capacity (SC) and its degree of gelatinization (DG) through a computer graphic analysis combined with deep learning techniques. SC and DG present a similar trend which was not seen a significant change before 60 degrees C followed a dramatical increase within 60-70 degrees C. The significant increase of DG was slower than SC, so that there were still high ungelatinized degrees at 68 degrees C when granules had disintegrated completely. In the case, the birefringence of small granules was efficiently identified and a high accuracy of 95% was achieved by using improved Starch-SSD. It may provide a fast and high throughput characterization of gelatinization merely based on a machine learning analysis of the photographical evidence.