Table 3 Seven different prognostic models for early prediction of OS in patients with mCRC.

From: Deep learning for the prediction of early on-treatment response in metastatic colorectal cancer from serial medical imaging

Model name (short name)

Cox proportional hazards model

Predictors (data type)

Time points used

RECIST

Univariate

RECIST criteria (dichotomous)

Baseline and the 1st follow-up

Tumor Burden (TB)

Univariate

Measurement of TB (continuous)

Baseline

Early Tumor Shrinkage (ETS)

Univariate

Measurement of TB (continuous)

Baseline and the 1st follow-up

DL Baseline Score (DL-BS)

Univariate

Prediction score by DL network (continuous)

Baseline

DL Prediction Score (DL-PS)

Univariate

Prediction score by DL network (continuous)

Baseline and the 1st follow-up

Size Nomogram (Size-Nomo)

Multivariate

RECIST + TB + ETS

Baseline and the 1st follow-up

DL Nomogram (DL-Nomo)

Multivariate

DL-PS + RECIST + TB + ETS

Baseline and the 1st follow-up

  1. Note: ā€˜+’ In the table indicates combination of predictors.