A deep learning-based system for automatic detection of emesis with high accuracy in Suncus murinus

• DOI: 10.1038/s42003-025-07479-0
• Publication Date: 2025-02-10T09:33:49Z
• Authors: Zengbing Lu; Yimeng Qiao; Xiaofei Huang; Dexuan Cui; Julia Y. H. Liu; Man Piu Ngan; Luping Liu; Zhixin Huang; Zi-Tong Li; Lingqing Yang; Aleena Khalid; Yingyi Deng; Sze Wa Chan; Longlong Tu; John A. Rudd
• Keywords: Suncus

Abstract Quantifying emesis in Suncus murinus ( S. ) has traditionally relied on direct observation or reviewing recorded behaviour, which are laborious, time-consuming processes that susceptible to operator error. With rapid advancements deep learning, automated animal behaviour quantification tools with high accuracy have emerged. In this study, we pioneere the use of both three-dimensional convolutional neural networks and self-attention mechanisms develop Automatic Emesis Detection (AED) tool for , achieving an overall 98.92%. Specifically, motion-induced videos as training datasets, validation results demonstrating 99.42% emesis. our model generalisation application studies, assess AED using various emetics, including resiniferatoxin, nicotine, copper sulphate, naloxone, U46619, cyclophosphamide, exendin-4, cisplatin. The prediction accuracies these emetics 97.10%, 100%, 98.97%, 96.93%, 98.91%, 98.41%, respectively. conclusion, employing learning-based automatic analysis improves efficiency mitigates human bias errors. Our study provides valuable insights into development learning network models aimed at automating behaviours potential applications preclinical research drug development.