output risk for computing timeROC

[EQUIWDH]

Hello, I am considering about increasing evaluation for deepsurv. I am wondering it's reasonable that use output risk $h_{\theta}(x)$ to calculate survival lines just imitating breslow and KM estimation method because they come from full likelihood of CoxPH. such the base line cumulative hazard function $$\Lambda_0 (t) = \sum_{i=1}^{N} \frac{\delta_i}{ \sum_{j \in R( T_{(i)} ) } \exp(h_{\theta}(x))} I( t \geq T_{(i)} )$$ and then, calculate timeROC. Thanks!