Table 4.

BIAS analysis of the nine selected datasets: Using an adapted version of BIAS tools the quality of dataset description was evaluated.

	BRIXIA	ML HANNOVER	COVID-19-AR	HM HOSPITALES	ACTUALMED	BIMCV-COVID 19	COVIDGR	RICORD	AIforCOVID	Coverage (%)
Principal objective	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	88
Team description	Yes	No	Yes	No	No	Yes	Yes	Yes	Yes	66
Oficial website or platform	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	100
Open or on request	On request	Open	Open	On request	Open	On request	Open	Open	On request	100 (55)
Include potential limitations	Yes	No	No	No	No	No	No	Yes	Yes	33
Ethical approval	Yes	No	No	No	No	Yes	Yes	Yes	Yes	55
Usage agreement specification	Yes	Yes	Yes	Yes	No	Yes	No	Yes	Yes	77
General image information	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	100
General patient information	Yes	Yes	Yes	Yes	No	Yes	No	Yes	Yes	77
Acquisition device	Yes	No	Yes	Yes	No	Yes	No	Yes	Yes	66
Acquisition protocol	Yes	No	Yes	Yes	No	Yes	No	Yes	Yes	66
Post-procesing tecniques	Yes	No	Yes	Yes	No	Yes	Yes	Yes	Yes	77
Regional origin of data	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	88
Whole or training/test division	Whole *	Whole	Whole	Whole	Whole	Whole	Whole	Whole	Whole	100 (Whole)
Different population distribution explanation	No	No	Yes	No	No	No	No	No	Yes	22
Description of annotation methods	Yes	No	Yes	No	No	Yes	Some	Yes	Yes	66(55)
If human annotations, description	Yes	–	–	–	–	Yes	No	Yes	–	75% When apply
If multiple annotations merge, description	Yes	–	–	–	–	–	–	Yes	–	100% When apply
Indication of potential error sources	Yes	No	No	No	No	Yes	No	No	Yes	33
Quantification of potential error sources	Some	No	No	No	No	Some	No	No	Some	33(0)

*The dataset was realised as a whole, but the annotation method for training and test differ, hence in other parts of the paper training and test subsets are analysed separately.