Conventional and semi-automatic histopathological analysis of tumor cell content for multigene sequencing of lung adenocarcinoma

Daniel Kazdal, Eugen Rempel, Cristiano Oliveira, Michael Allgäuer, Alexander Harms, Kerstin Singer, Elke Kohlwes, Steffen Ormanns, Ludger Fink, Jörg Kriegsmann, Michael Leichsenring, Katharina Kriegsmann, Fabian Stögbauer, Luca Tavernar, Jonas Leichsenring, Anna Lena Volckmar, Rémi Longuespée, Hauke Winter, Martin Eichhorn, Claus Peter HeußelFelix Herth, Petros Christopoulos, Martin Reck, Thomas Muley, Wilko Weichert, Jan Budczies, Michael Thomas, Solange Peters, Arne Warth, Peter Schirmacher, Albrecht Stenzinger, Mark Kriegsmann

Research output: Contribution to journalArticlepeer-review

9 Scopus citations


Background: Targeted genetic profiling of tissue samples is paramount to detect druggable genetic aberrations in patients with non-squamous non-small cell lung cancer (NSCLC). Accurate upfront estimation of tumor cell content (TCC) is a crucial pre-analytical step for reliable testing and to avoid false-negative results. As of now, TCC is usually estimated on hematoxylin-eosin (H&E) stained tissue sections by a pathologist, a methodology that may be prone to substantial intra- and interobserver variability. Here we the investigate suitability of digital pathology for TCC estimation in a clinical setting by evaluating the concordance between semi-automatic and conventional TCC quantification. Methods: TCC was analyzed in 120 H&E and thyroid transcription factor 1 (TTF-1) stained high-resolution images by 19 participants with different levels of pathological expertise as well as by applying two semi-automatic digital pathology image analysis tools (HALO and QuPath). Results: Agreement of TCC estimations [intra-class correlation coefficients (ICC)] between the two software tools (H&E: 0.87; TTF-1: 0.93) was higher compared to that between conventional observers (0.48; 0.47). Digital TCC estimations were in good agreement with the average of human TCC estimations (0.78; 0.96). Conventional TCC estimators tended to overestimate TCC, especially in H&E stainings, in tumors with solid patterns and in tumors with an actual TCC close to 50%. Conclusions: Our results determine factors that influence TCC estimation. Computer-assisted analysis can improve the accuracy of TCC estimates prior to molecular diagnostic workflows. In addition, we provide a free web application to support self-training and quality improvement initiatives at other institutions.

Original languageEnglish
Pages (from-to)1666-1678
Number of pages13
JournalTranslational Lung Cancer Research
Issue number4
StatePublished - Apr 2021
Externally publishedYes


  • Digital pathology
  • Lung adenocarcinoma (lung ADC)
  • Molecular pathology
  • Next-generation sequencing (NGS)
  • Tumor cell content (TCC)


Dive into the research topics of 'Conventional and semi-automatic histopathological analysis of tumor cell content for multigene sequencing of lung adenocarcinoma'. Together they form a unique fingerprint.

Cite this