Table 3

An example of variables collected in data extraction table

Study information
 Study yearYear of the study published
 Author informationLast name of author, whether clinical practitioners participated in the study
 Type of studyProspective cohort study or study that used a published database
 Journal nameJournal name
 PICO/PECO elementsPICO/PECO elements in summary
Database information
 Database nameName of the database used for modelling
 Host organisationName of the hosting organisation of the database
 Sample sizeSample size of the database (image or video)
 SponsorshipThe funding or sponsorship information
Patient demographic information
 GenderGender of infants (both, only boy, only girl)
 AgeAge distribution
 RaceRace/country of participants
 Disease diagnosisDisease diagnosis
 Medical proceduresProcedure categories
Machine learning method information
 Model nameThe name of the model
 Model typeMachine learning model type
 Model taskClassification, regression or both
 Objective functionThe objective function for modelling
 Optimisation algorithmThe optimisation method for modelling
 Format of input featureFrame, sequence or image
 Positive/negative size inputThe size of positive and negative for modelling
 Feature extraction methodThe methods of feature extraction
 Type of extracted featurePixel feature, AU, landmark or transformed feature
 Model performancePerformance metrics and score of performance
 Computational efficiency and costComputational efficiency (speed, cloud space, etc) and cost related to the algorithm (eg, require GPU resources, large cluster, etc)
  • AU, action unit; PICO/PECO, population, intervention (exposure), control, outcome.