Tumor Regression Grading (TRG) schemes for neoadjuvantly treated gastrointestinal (GI) tumors provide crucial prognostic information for treatment planning. However, TRG systems suffer from well-known interobserver variability, especially when used by less experienced pathologists. A more precise disease characterization, building upon existing classification schemes and executed through computer-assisted digital pathology (DP) on fully digitized slides, can enhance patient prognosis stratification and ultimately improve treatment planning and patient outcomes.

Our goal is to leverage an existing unique multi-reader dataset with our established deep learning (DL) technology to deterministically validate commonly used grading systems such as Mandard and Becker, and subsequently extend them through a clinically deployable computer-assisted diagnostic tool (CDx), referred to as CDx-TRG+.