Autism risk classification using placental chorionic surface vascular network features

Table 1 The first five principal components (PCs) of the data retain approximately 88% of the data variability

Boruta ranking	Vascular features (variability captured)	PC1 (35.27%)	PC2 (22.57%)	PC3 (17.20%)	PC4 (7.79%)	PC5 (5.80%)
1	MeanThickness	− 0.1582	− 0.4747	0.1035	0.0651	− 0.0089
2	MeanTortuosity	0.0002	0.0575	0.5347	− 0.0979	0.0013
3	MurrayL1FitError	− 0.256	− 0.3903	0.0438	0.0139	0.0397
4	StdThickness	− 0.1566	− 0.4762	0.0701	− 0.0046	0.0196
5	StdDevTortuosity	0.0029	0.0812	0.5912	− 0.0641	0.1449
6	MaxTortuosity	0.0948	0.0724	0.5459	− 0.0264	0.1709
7	MeanAngle	− 0.0611	0.0704	0.2028	0.2135	− 0.936
8	NumEndPoints	0.4251	− 0.0298	− 0.0132	0.0153	− 0.005
9	ArcLength	0.3773	− 0.1259	− 0.0035	− 0.0163	0.0116
10	NumBranchPoints	0.4254	− 0.0301	− 0.0125	0.0146	− 0.0038
11	MurrayBranchesUsed	0.4254	− 0.0301	− 0.0125	0.0146	− 0.0038
12	Volume	0.1444	− 0.4823	0.065	0.0502	− 0.0368
13	NumGenerations	0.3182	− 0.0237	0.014	0.2178	− 0.0619
14	MeanDistEndPointToPerim	0.0055	− 0.0323	0.0545	0.905	0.2124
15	VesselToDiscPercent	0.255	− 0.3502	0.0031	− 0.2561	− 0.1457

The absolute value of the attributes within each PC gives a measure of contribution. The higher the value, the bigger the contribution. Specifically, NumEndPoints, NumBranchPoints, and MurrayBranchesUsed contributed the most to PC1, Thickness, StdThickness, and Volume contributed the most to PC2, MeanTortuosity, StdDevTortuosity, MaxTortuosity contributed the most to PC3, MeanDistEndPointToPerim contributed most to PC4, and MeanAngle contributed most to PC5

ISSN: 1472-6947