P6 Casanova et al.

Identification of morphologic parameters affecting the prognosis of lung squamous cell carcinoma patients by computer-based morphometric analysis

Ruben Casanova, Undine Rulle, Bart Vrugt, Holger Moch, Alex Soltermann
Department of Pathology and Molecular Pathology, University Hospital, Zurich, Switzerland

Background: In addition to staging, morphological features may bring additional criteria to stratify patients into relevant prognostic groups. In our study, we have used an image-based computational method on histologic sections to identify morphological features of lung squamous cell carcinoma (LSCC) with prognostic relevance.

Methods: Three patients’ cohorts of surgically resected lung SCC tumors were investigated (tissue microarray (TMA): n=208, whole sections (WS): n=99). TMA and whole sections cohorts were immuno-histochemically stained with pan-cytokeratin.  Color-based segmentation followed by morphologic features screening were performed. Tumor fragmentation (TF), defined as tumor clusters >800µm2 (circa >5 cells) was further evaluated. An external validation TCGA (The cancer genome atlas) cohort (n=335, H&E stained) was scored by eye for TF.

Results: We identified tumor fragmentation (TF) as a morphologic feature significantly associated with poor outcome, independent from tumor stage, in two independent clinical cohorts. This was confirmed using a similar human-based scoring system in the TCGA cohort.

Conclusion: In summary, by using a computer-based morphometric approach, we identified TF as an independent prognostic marker for chemotherapy-naïve LSCC patients, which could serve to refine current tumor grading systems.