Algebraic topology-based machine learning using MRI predicts outcomes in primary sclerosing cholangitis

European Radiology Experimental

Table 1 Baseline characteristics of patients

	Derivation cohort/training (n = 54)	Validation cohort (n = 115)	p value
Females	21 (38.9%)	39 (33.9%)	0.619
Inflammatory bowel disease presence	49 (90.7%)	94 (81.7%)	0.166
Age (years)	38.5 [24.0, 52.8]	49.7 [32.3, 58.9]	0.019
PSC duration (years)	3.26 [0.41, 11.28]	0.32 [0.00, 4.69]	0.002
SAP/ULN	2.50 [1.26, 4.24]	1.72 [1.03, 3.07]	0.039
Total bilirubin (mg/dL)	0.90 [0.50, 2.90]	0.80 [0.50, 1.86]	0.624
Platelets (× 10⁹/L)	267 [169, 326]	238 [167, 338]	0.800

Data are given as median [interquartile range] with the exception of the number and percentage of females and of patients with inflammatory bowel disease. For some patients, the MRI/MRCP which established their diagnosis of PSC was used in this study. The date of the MRI used in the study was time zero or the baseline. Therefore, the duration of PSC for some subjects was 0 years
MRI Magnetic resonance imaging, MRCP Magnetic resonance cholangiopancreatography, PSC Primary sclerosis cholangitis, SAP/ULN Serum alkaline phosphatase relative to the upper limit of normal within three months of the baseline magnetic resonance imaging