ECP 2017 SY-03: Speaker: Stephan Borkenfeld, Germany - Cognitive programming and digital pathology

How to Cite: Gian Kayser, Germany - Conference Report. Diagnostic Pathology, Apr. 2019. ISSN 2364-4893. Available at: Date accessed: 01 apr. 2017.


Conference Report:

PD Dr. Gian Kayser, Institute of Surgical Pathology, University of Freiburg, Germany

Impressions from the 30th European Congress of Pathology in Bilbao, Spain 08. – 12. September 2018

This year, the 30th European Congress of Pathology (ECP), organized by the European Society of Pathology (ESP), took place in Bilbao from September 8 to 12. The main theme of this event was titled „Pathology: Path to Precision Medicine“, a play-on-words which expresses the central role pathology plays in guiding today´s personalized targeted therapy. This central theme resonated around two rapidly evolving fields: molecular pathology and predictive testing for targeted therapies in oncology, and technological advancements, esp. digital pathology. Thus, this report will focus on these topics. That said, the oral free paper sessions at ECP 2018 offered innumerous contributions of highest scientific quality across various fields of study. Additionally, excellent diagnostic sessions drew crowds of attendees that by far exceeded the auditorium capacity, especially in the areas of gastro- and lung-pathology.

Molecular pathology

Throughout the last decade, the possibilities of targeted therapy have not only grown by numbers due to the development of new precision medicines by the pharmaceutical industry, but also novel targets, methods and technologies for predictive biomarker testing (CDx).

External quality assurance (EQA): A prerequisite for CDx testing

very year, pathologists are confronted with a steadily increasing range of quality assurance schemes. These may not be limited any more to single targets, but cover various targets in a single cancer type. Additionally, EQA schemes are increasingly “method-agnostic”, i.e. the technology used for result determination is chosen by the participating institute.

The evolution of BRAF CDx in melanoma

Analyses of the British EQA provider UK NEQAS were presented in an industry symposium. For BRAF-testing in malignant melanoma, Sandi Deans reported a clear increase in institutes utilizing next-generation sequencing (NGS). Alternative methods like Sanger- and pyrosequencing are declining. Surely, this trend is caused by the panel-testing now required for many cancer types. The increase in actionable molecular alterations has rendered the sequential testing of individual biomarkers in a reasonable, patient-friendly time frame nearly impossible. In this context, an interesting international multi-center study (Hartmann et al.,) was presented, investigating the concordance of BRAF-testing at eight institutes. In this study, test results showed a congruence of 95 %. Remarkably, of all methods used for BRAF analysis, the highest accuracy was observed for PCR-based systems.

The testing need for BRAF mutations is continuing to evolve: Selective BRAF-inhibitors, like vermurafenib, have been approved for the treatment of stage IV malignant melanoma for several years now. Three years ago, the combination treatment of the BRAF-inhibitor dabrafenib with a MEK-inhibitor (trametinib) was approved for advanced BRAF- mutated m alignant melanomas (– due to a significant clinical benefit. In September 2018, the approval of this combination therapy was extended to include the adjuvant setting for stage III melanoma patients, which now requires pathologists to detect BRAF mutations in the lymph node (micro-) metastases of these locally advanced tumors – or the primary melanoma itself.

Challenges in immuno-oncologic biomarkers: The quest continues…

Another major topic when it comes to predicting therapeutic response is immuno-oncology. For malignant melanoma, as well as non-small-cell lung carcinoma (NSCLC), checkpoint inhibitors were initially approved for second line without the necessity of a predictive test (CDx). Later, cut-offs of 1% for were required for adenocarcinoma. Finally this year, pembrolizumab was approved as first line therapy for PD-L1-positive non-metastasized NSCLC. Here, PD-L1 positivity is defined as the expression of PD-L1 in more than 50 % of tumor cells. PD-L1 expression as therapy predictor has been a point of discussion for several years now. Reasons for debate were not only the heterogeneous study results and varying threshold/cut-off values, but also the mounting number of scoring algorithms which, depending on tumor type, required interpretation of PD-L1 expression in tumor cells alone, both tumor and stroma cells, or the proportion of all PD-L1-positive cells divided by the number of tumor cells

A number of harmonizing studies have shown that the choice of antibody assay plays only a minor role in determining PD-L1 positivity. It is known that there exists a spatial as well as temporal heterogeneity in the expression of this antigen within the tumor – this might account for the aforementioned discrepancies observed in different studies. With that in mind, the results of using tumor mutational burden (TMB) as predictor of therapy response for checkpoint inhibitors were highly anticipated.

The Checkmate 227 study had already indicated that the response to combination therapy of ipilimumab and nivolumab is better in patients with tumors harboring a high mutational burden (> 10 mutations/megabase). In this study, TMB was determined using the FoundationOne panel. The harmonization of TMB testing is becoming a major task for the pathologist community nationally, EU-wide, as well as internationally. Here, not only the established use of different analysis techniques, but especially the following issues should be addressed: The minimum number of DNA base pairs, coding and/or non-coding, TMB analysis has to cover in order to achieve a valid assessment. How to deal with driver mutations, since it is known that tumors carrying such mutations respond less well to immune-checkpoint blockade. And, last but not least, how to bioinformatically compute TMB analysis, taking into account non-coding segments and polymorphisms. Given that TMB testing/analysis is so technically and bioinformatically complex, its adoption into routine diagnostics as a predictive test will be an exciting endeavor.

Digital pathology

The increasing interest in digital pathology is reflected by the fact that slide- scanner manufacturers have been prominently featured at pathology conventions over the recent years.

Telepathology represents a classic area of digital pathology, and remains of interest today. Region-wide pathology coverage is of particular interest to rural, sparsely populated areas as encountered, for example, in Scandinavia or Canada. Telepathology also presents an appealing option for institutional networks, to make diagnosis a more efficient process. While the use of telepathology for frozen-section analysis has been gaining international acceptance, it remains to be seen where German pathologists stand on the subject. Even though the question of legal responsibility concerning the use of digital pathology for consultation and online telepathology seems to have been settled, personal opinions continue to clash. The Federal Association of German Pathologists (Bundesverband Deutscher Pathologen) has published the guideline „Digital Pathology“, which is internationally available in the digital journal „“ (

Yet, digital pathology comprises a lot more than just telepathology. Several talks centered on the implementation of digital pathology in routine diagnostics. Here, the integration of digitalization into the workflow of individual institutes emerges as a central challenge – this in particular is THE link to the lab-/pathology-information system (LIS) and therefore to routine diagnostics. Despite existing communication standards (such as HL7), each institute still requires individual solutions. To achieve extensive, region-wide implementation of digital diagnostics, harmonization of these communication standards between scanner manufacturers and LIS providers will be essential.

For years now, pathologists have held high expectations for IT-assisted diagnostics of digitized slides. Rapid development in the areas of computer processors and storage/memory capacity for the consumer market have enabled affordable solutions for automated analyses and diagnostic aids. Internationally, many working groups are working on suitable approaches based on so-called “deep-learning”. This artificial intelligence (AI) consists of more or less complex, multi-layered convolutional neuronal networks (CNNs), which statistically analyze and weight individual image characteristics according to pre-defined classifications. The advantage of these approaches lies, for one, in the free availability of CNN-libraries (e.g. tensorflow, theano, caffe). And secondly, in their relatively easy implementation, since CNNs carry out image analysis for classification autonomously. It remains exciting to see, when these developments will be ready for application in routine diagnostics and if how they may accelerate the adoption of digital pathology into daily routine. A harmonization of communication standards as well as image formats would be highly desirable, if not crucial, to facilitate inter-institutional exchange and to enable institutes to freely choose and combine systems.

In 2019, the ECP will be hosted in Nice. The author is already looking forward to more exciting and interesting contributions that will broaden the scientific and diagnostic horizon in pathology.


This report was created with support of Novartis Pharma GmbH

404 Not Found