Table 2.

Study characteristics

Author (year)	Country	Classification	Total images	Model	Reference standard	Comparison
Fatima et al.[12] (2017)	Pakistan	Normal fundus Papilledema	160 images from 2 different data sets; 90 from STARE and 70 from local data set acquired from AFIO 90: Normal 70: Abnormal	Robust method to detect the location of the optic disk SVM-RBF classifier to detect and grade papilledema	STARE annotations, then re-annotated by two ophthalmologists + local data set annotated by two ophthalmologists	None
Akbar S et al.[13] (2017)	Pakistan	Normal fundus Papilledema Grading Mild papilledema Severe papilledema	160 images from 2 data sets STARE database Local database collected by AFIO	Robust method to detect the location of the optic disk SVM-RBF classifier to detect and grade papilledema	Classification by two ophthalmologists using Friesen grading rule	None
Milea D et al.[9] (2020)	Singapore	Normal fundus Papilledema Disk with other abnormalities	15.846 photographs from 7.532 patients  Training and validation: 14.341 photographs from 6.769 patients  External testing: 1.505 photographs	BONSAI-DLS U-Net to detect the location of the optic disk DenseNet to classify the optic disk	Classification by the study steering committee	None
Biousse V et al.[10] (2020)	Singapore, USA	Normal fundus Papilledema Disk with other abnormalities	This study compared the performances of previously trained and validated systems versus expert neuro-ophthalmologists using 800 photographs from 454 patients	BONSAI-DLS U-Net to detect the location of the optic disk DenseNet to classify the optic disk	Diagnosis provided by each center	Two expert neuro-ophthalmologists (without clinical information given)
Saba T et al.[14] (2021)	Saudi Arabia	Normal fundus Papilledema Grading Mild papilledema Severe papilledema	Training: 500 images Testing: 60 papilledema and 40 normal fundus from the STARE dataset	DLS: DenseNet to classify optic disc as papilledema or normal U-Net to grade the papilledema	Dataset annotation and ophthalmologist’s classification	None
Vasseneix C et al.[11] (2021)	Singapore	Grading Mild/moderate papilledema Severe papilledema	Training: 2.103 photographs of 965 patients (1.052 mild to moderate papilledema, 1.051 severe papilledema) Testing: 214 photographs from 111 patients	DLS: U-Net to detect the location of the optic disk VGGNet to grade the papilledema	Consensus by panel of expert consisting of 2 experts, with additional 2 neuro-ophthalmologists	Three independent neuro-ophthalmologist (masked to clinical information)

RBF=Radial basis function, SVM=Support vector machine, BONSAI=Brain and optic nerve study with artificial intelligence, DLS=Deep learning systems, AFIO=Armed forces institute of ophthalmology