Computer-assisted Histomorphological Analysis of Head and Neck Carcinomas and its Relevance in the Oncological Outcome

Introduction Head and neck carcinomas (HNSCC) display a heterogeneous histological appearance. However, there are only a few prognostic relevant histological parameters to be evaluated routinely. Therefore, our aim was to obtain clinically relevant morphological markers by using digital pathology tools on hematoxylin and eosin-(HE)-stained sections.

Material and Methods 100 HE-stained HNSCC slides were assessed. Among various parameters (including tumor growth and infiltration pattern [compact vs. discontinuous], infiltration depth, proportion of different tissue components), the lymphocytic stroma (IZRS%) in particular, characterized by the density of tumor-infiltrating lymphocytes (TIL), was evaluated both by an experienced pathologist (TIL%) and in a computer-assisted way (absTIL) using neural networks. Then, we evaluated the impact of the histomorphological parameters on the disease-specific survival.

Results Our cohort included 76 male and 24 female patients with an average age of 62±13 years, as for tumor stage n=37 pT1, n=32 pT2, n=18 pT3 and n=13 pT4 cases. TIL measurements obtained by the pathologist showed a significant correlation with respect to the computer-assisted

Results TIL% vs. absTIL [R=0.82; p<0.05]; TIL% vs. IZRS% [R=0.86; p<0.05], IZRS% vs. absTIL [R=0.76; p<0.05]. An infiltration depth of < 5 mm, a compact infiltration pattern, and an absTIL value > 4450/mm2 were found to be significantly favorable prognostic factors.

Conclusion Additional prognostically relevant information can be derived from HE-stained routine sections in a standardized and reproducible manner using artificial intelligence methods. These may contribute to the personalized oncotherapy of HNSCC in the future.

Keine finanzielle Unterstützung vorhanden.

Publication History

Article published online:
12 May 2023