Accurate species identification of food-contaminating beetles with quality-improved elytral images and deep learning

Authors: BISGIN, HALIL - University Of Michigan; BERA, TANMAY - Us Food & Drug Administration (FDA); WU, LEIHONG - Us Food & Drug Administration (FDA); DING, HONGJIAN - Us Food & Drug Administration (FDA); BISGIN, NESLIHAN - University Of Michigan; LIU, ZHICHAO - Us Food & Drug Administration (FDA); PAVA-RIPOLL, MONICA - Us Food & Drug Administration (FDA); BARNES, AMY - Us Food & Drug Administration (FDA); Campbell, James - Jim; VYAS, HIMANSU - Us Food & Drug Administration (FDA); FURLANELLO, CESARE - Hk3 Lab; TONG, WEIDA - Us Food & Drug Administration (FDA); XU, JOSHUA - Us Food & Drug Administration (FDA)

Submitted to: Frontiers in Artificial Intelligence
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/22/2022
Publication Date: 8/12/2022
Citation: Bisgin, H., Bera, T., Wu, L., Ding, H., Bisgin, N., Liu, Z., Pava-Ripoll, M., Barnes, A., Campbell, J.F., Vyas, H., Furlanello, C., Tong, W., Xu, J. 2022. Accurate species identification of food-contaminating beetles with quality-improved elytral images and deep learning. Frontiers in Artificial Intelligence. https://doi.org/10.3389/frai.2022.952424.
DOI: https://doi.org/10.3389/frai.2022.952424

Interpretive Summary: Food samples are routinely screened for food-contaminating beetles due to their adverse impacts, and if beetle fragments are found they are analyzed to identify the species responsible for the contamination. At present this identification is done through manual examination using a microscope and is based on the unique patterns on the hardened forewing (elytra) of the beetles. In our earlier studies, we demonstrated that automated species identification of beetles is feasible through elytral pattern recognition, but due to poor image quality prediction accuracies of more than 80% were not obtained. In this study, higher quality elytral images were evaluated to determine how accurately 27 different species of beetles could be identified. A convolutional neural network (CNN) model was used to compare performance between two different image sets for various beetle species. Improved image quality did lead to better prediction accuracy, but multiple high-quality images were also necessary, especially for species with high variation in elytra patterns. Automating the pattern recognition process through machine learning would enable more efficient and consistent identification of beetle species and our results provide a direction toward achieving the ultimate goal of automated species identification system.

Technical Abstract: Food samples are routinely screened for food-contaminating beetles (i.e., pantry beetles) due to their adverse impact on the economy, environment, public health and safety. If found, their remains are subsequently analyzed to identify the species responsible for the contamination; each species poses different levels of risk, requiring different regulatory and management steps. At present, this identification is done through manual microscopic examination since each species of beetle has a unique pattern on its elytra (hardened forewing). Our study sought to automate the pattern recognition process through machine learning. Such automation will enable more efficient identification of pantry beetle species and could potentially be scaled up and implemented across various analysis centers in a consistent manner. In our earlier studies, we demonstrated that automated species identification of pantry beetles is feasible through elytral pattern recognition. Due to poor image quality, however, we failed to achieve prediction accuracies of more than 80%. Subsequently, we modified the traditional imaging technique, allowing us to acquire high-quality elytral images. In this study, we explored whether high-quality elytral images can truly achieve near-perfect prediction accuracies for 27 different species of pantry beetles. To test this hypothesis, we developed a convolutional neural network (CNN) model and compared performance between two different image sets for various pantry beetles. Our study indicates improved image quality indeed leads to better prediction accuracy; however, it was not the only requirement for achieving good accuracy. Also required are many high-quality images, especially for species with a high number of variations in their elytral patterns. The current study provided a direction toward achieving our ultimate goal of automated species identification through elytral pattern recognition.