The IceCube Neutrino Observatory instruments a cubic-kilometer of glacial ice and has been the first experiment to identify high-energy astrophysical neutrinos.
There are two main morphologies of IceCube events: tracks and cascades.
Tracks result from muons, while cascades result from particle showers induced by in-ice interactions.
The directional reconstruction of cascades is less precise than that of tracks, which limits the sensitivity of astrophysical neutrino analyses with cascade events.
In order to improve the directional reconstruction of cascade events, accurate ice modeling is essential.
However, potential biases might exist in data stemming from unconstrained systematic uncertainties.
In this work, feasibility studies to better understand the ice using a data-driven approach are performed, where photons that are likely to have been induced by the hadronic cascade part of muon neutrino charged-current interactions are categorized using probability density functions in time, distance and angle, and the reconstructed direction with this {\it pseudo-cascade} is compared with the track direction.
In this proceedings, methodology and results are detailed and a path towards better understanding of the ice is discussed.