Publication Database

Towards an automated facial expression analysis in mice using deep learning (2022)

Art

Vortrag

Autoren

Hohlbaum, K. (WE 11)

Andresen, N.

Wölhaf, M.

Lewejohann, L. (WE 11)

Helwich, O.

Thöne-Reineke, C. (WE 11)

Belik, V. (WE 16)

Kongress

15th FELASA Congress 2022

Marseille, France, 13. – 16.06.2022

Quelle

Laboratory animals

Bandzählung: 56

Heftzählung: 1 Suppl.

Seiten: 71 – 72

ISSN: 0023-6772

Sprache

Englisch

Verweise

URL (Volltext): https://journals.sagepub.com/doi/full/10.1177/00236772221103950

DOI: 10.1177/00236772221103950

Kontakt

Institut für Veterinär-Epidemiologie und Biometrie

Königsweg 67
14163 Berlin
+49 30 838 56034
epi@vetmed.fu-berlin.de

Abstract / Zusammenfassung

The Mouse Grimace Scale (MGS) is a coding system for facial expression analysis of pain in mice and is widely accepted as welfare indicator. To facilitate and improve the use of the MGS, we aimed to develop a facial expression recognition software for mice. To this end, we utilized an image dataset of adult male and female C57BL/6JRj mice, that were either untreated, anesthetized (with isoflurane or ketamine/xylazine) or castrated (under isoflurane anaesthesia, meloxicam, lidocaine/prilocaine). The dataset was divided into two categories, i.e. “post surgical/anaesthetic effects present” and “no post-surgical/anaesthetic effects present”. Abinary classifier was trained to differentiate between the two categories. We used three convolutional neural network (CNN) architectures (two pre-trained state of the art deep CNN: ResNet50 and InceptionV3; one CNN of our own design without pre-training). When the network was provided multiple images per mouse, an accuracy of up to 99% was achieved. A feature visualization technique (Deep Taylor decomposition) indicated that the decision of the network was mainly based on image areas depicting the mouse faces. Our first steps towards a fully automated facial expression recognition software contributes to refining pain and stress assessment in laboratory mice.