Dataset Information

Fully automated deep learning models with smartphone applicability for prediction of pain using the Feline Grimace Scale.

ABSTRACT: This study used deep neural networks and machine learning models to predict facial landmark positions and pain scores using the Feline Grimace Scale^© (FGS). A total of 3447 face images of cats were annotated with 37 landmarks. Convolutional neural networks (CNN) were trained and selected according to size, prediction time, predictive performance (normalized root mean squared error, NRMSE) and suitability for smartphone technology. Geometric descriptors (n = 35) were computed. XGBoost models were trained and selected according to predictive performance (accuracy; mean square error, MSE). For prediction of facial landmarks, the best CNN model had NRMSE of 16.76% (ShuffleNetV2). For prediction of FGS scores, the best XGBoost model had accuracy of 95.5% and MSE of 0.0096. Models showed excellent predictive performance and accuracy to discriminate painful and non-painful cats. This technology can now be used for the development of an automated, smartphone application for acute pain assessment in cats.

SUBMITTER: Steagall PV

PROVIDER: S-EPMC10703818 | biostudies-literature | 2023 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Fully automated deep learning models with smartphone applicability for prediction of pain using the Feline Grimace Scale.

Steagall P V PV Monteiro B P BP Marangoni S S Moussa M M Sautié M M

Scientific reports 20231207 1

This study used deep neural networks and machine learning models to predict facial landmark positions and pain scores using the Feline Grimace Scale<sup>©</sup> (FGS). A total of 3447 face images of cats were annotated with 37 landmarks. Convolutional neural networks (CNN) were trained and selected according to size, prediction time, predictive performance (normalized root mean squared error, NRMSE) and suitability for smartphone technology. Geometric descriptors (n = 35) were computed. XGBoost ...[more]

PMID: 38062194

Similar Datasets

Project description:BackgroundThe Feline Grimace Scale (FGS) is a facial expression-based scoring system for acute pain assessment in cats with reported validity using image assessment. The aims of this study were to investigate the clinical applicability of the FGS in real-time when compared with image assessment, and to evaluate the influence of sedation and surgery on FGS scores in cats.MethodsSixty-five female cats (age: 1.37 ± 0.9 years and body weight: 2.85 ± 0.76 kg) were included in a prospective, randomized, clinical trial. Cats were sedated with intramuscular acepromazine and buprenorphine. Following induction with propofol, anesthesia was maintained with isoflurane and cats underwent ovariohysterectomy (OVH). Pain was evaluated at baseline, 15 min after sedation, and at 0.5, 1, 2, 3, 4, 6, 8, 12 and 24 h after extubation using the FGS in real-time (FGS-RT). Cats were video-recorded simultaneously at baseline, 15 min after sedation, and at 2, 6, 12, and 24 h after extubation for subsequent image assessment (FGS-IMG), which was performed six months later by the same observer. The agreement between FGS-RT and FGS-IMG scores was calculated using the Bland & Altman method for repeated measures. The effects of sedation (baseline versus 15 min) and OVH (baseline versus 24 h) were assessed using linear mixed models. Responsiveness to the administration of rescue analgesia (FGS scores before versus one hour after) was assessed using paired t-tests.ResultsMinimal bias (-0.057) and narrow limits of agreement (-0.351 to 0.237) were observed between the FGS-IMG and FGS-RT. Scores at baseline (FGS-RT: 0.16 ± 0.13 and FGS-IMG: 0.14 ± 0.13) were not different after sedation (FGS-RT: 0.2 ± 0.15, p = 0.39 and FGS-IMG: 0.16 ± 0.15, p = 0.99) nor at 24 h after extubation (FGS-RT: 0.16 ± 0.12, p = 0.99 and FGS-IMG: 0.12 ± 0.12, p = 0.96). Thirteen cats required rescue analgesia; their FGS scores were lower one hour after analgesic administration (FGS-RT: 0.21 ± 0.18 and FGS-IMG: 0.18 ± 0.17) than before (FGS-RT: 0.47 ± 0.24, p = 0.0005 and FGS-IMG: 0.45 ± 0.19, p = 0.015).ConclusionsReal-time assessment slightly overestimates image scoring; however, with minimal clinical impact. Sedation with acepromazine-buprenorphine and ovariohysterectomy using a balanced anesthetic protocol did not influence the FGS scores. Responsiveness to analgesic administration was observed with both the FGS-RT and FGS-IMG.

Project description:ObjectivesThe aim of the present study was to investigate the construct validity, responsiveness and reliability of the Feline Grimace Scale (FGS) in kittens.MethodsA total of 36 healthy female kittens (aged 10 weeks to 6 months) were included in a prospective, randomized, blinded study. Video recordings of all kittens were made before and 1 and 2 h after ovariohysterectomy using an opioid-free injectable anesthetic protocol with or without multimodal analgesia. Additional recordings were taken before and 1 h after administration of rescue analgesia (buprenorphine 0.02 mg/kg IM) to painful kittens. Screenshots of facial images were collected from the video recordings for FGS scoring. Four observers blinded to treatment groups and time points scored 111 randomized images twice with a 5-week interval using the FGS. Five action units (AUs) were scored (ear position, orbital tightening, muzzle tension, whiskers position and head position; 0-2 each). Construct validity, responsiveness, and inter- and intra-rater reliability were evaluated using linear models with Benjamini-Hochberg correction, Wilcoxon signed-rank test and single intraclass correlation coefficients (ICCsingle), respectively (P <0.05).ResultsFGS total ratio scores were higher at 1 and 2 h after ovariohysterectomy (median [interquartile range, IQR]: 0.30 [0.20-0.40] and 0.30 [0.20-0.40], respectively) than at baseline (median [IQR]: 0.10 [0.00-0.30]) (P <0.001). FGS total ratio scores were lower after the administration of rescue analgesia (median [IQR] before and after rescue analgesia) 0.40 [0.20-0.50] and 0.20 [0.10-0.38], respectively (P <0.001). Inter-rater ICCsingle was 0.68 for the FGS total ratio scores and 0.35-0.70 for all AUs considered individually. Intra-rater ICCsingle was 0.77-0.91 for the FGS total ratio scores and 0.55-1.00 for all AUs considered individually.Conclusions and relevanceThe FGS is a valid and responsive acute pain-scoring instrument with moderate inter-rater reliability and good to excellent intra-rater reliability in kittens.

Dataset Information

Fully automated deep learning models with smartphone applicability for prediction of pain using the Feline Grimace Scale.

Publications

Fully automated deep learning models with smartphone applicability for prediction of pain using the Feline Grimace Scale.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets