Skip to main content
. 2024 Oct 29;8(6):zrae127. doi: 10.1093/bjsopen/zrae127

Table 2.

Survival analyses on the ground-truth and NIC-classified HGP

Desmoplastic versus non-desmoplastic Non-desmoplastic 5-year OS (95% c.i.) Desmoplastic 5-year OS (95% c.i.) Desmoplastic versus non-desmoplastic
Univariable HR (95% c.i.) Multivariable HR (95% c.i.)*
Development cohort (n = 932)
 Ground-truth HGP 40% (36,44) 63% (57,70) 0.57 (0.47,0.70) 0.63 (0.50,0.79)
 NIC-classified HGP 40% (37,44) 60% (54,67) 0.61 (0.50,0.75) 0.64 (0.51,0.79)
Validation cohort (n = 294)
 Ground-truth HGP 64% (58,71) 80% (70,91) 0.51 (0.30,0.86) 0.40 (0.22,0.75)
 NIC-classified HGP 66% (60,72) 73% (63,84) 0.64 (0.41,1.02) 0.48 (0.28,0.83)

*Corrected for age, sex, primary tumour location, pT-stage, pN-stage, disease-free interval, number of CRLM, diameter of largest CRLM, preoperative CEA, and extrahepatic disease. CEA, carcinoembryonic antigen; CRLM, colorectal liver metastasis; HGP, histopathological growth pattern; NIC, neural image compression; OS, overall survival.