What is the future of pathology

Algorithms will not replace pathologists

BERLIN. How is machine learning changing medicine? There are different views on this, but almost everyone agrees that self-learning algorithms will find their way across the board, at least in the diagnostic subjects, especially radiology and pathology.

"In the first step, we will certainly talk about increasing efficiency," said Professor Frederick Klauschen from the Institute for Pathology at Charité Berlin at an event organized by the Federal Association of German Pathologists (BDP) in Berlin.

Klauschen, who is particularly active in the field of oncology and is not only a doctor but also a graduate physicist, sees a good chance that algorithms will relieve pathologists of some rather "annoying" tasks such as counting cell nuclei in the relatively short term. This accelerates the pathological diagnosis, and the pathologist gains more time for those difficult questions for which humans will often be better suited than machines in the future too.

Klauschen, who is researching artificial intelligence in pathology at the Berlin Center for Machine Learning funded by the Federal Ministry of Research, sees not only a quantitative, but also a qualitative benefit of self-learning algorithms. The search for biomarkers for immuno-oncology could possibly benefit from artificial intelligence. Because here there are increasing indications that complex biomarker patterns are better suited than individual biomarkers to predict therapy response. And precisely that, the recognition of patterns, is one of the great strengths of self-learning algorithms.

The question is how it will change the professional profile of pathologists when more and more of the originally medical activities can be shifted to machines. Is pathology abolishing itself through machine learning research?

BDP President Professor Karl-Friedrich Bürrig from the Institute for Pathology Hildesheim considers this thesis to be absurd: “Data need interpretation. Artificial intelligence alone will not make any diagnoses. Making diagnoses means dealing with probabilities, and I don't want to leave that to a machine. "

The BDP President is convinced that the professional profile of the pathologist will change: "My future vision is the hybrid pathologist who acts as a diagnostic whisperer using all digital methods." In this vision, artificial intelligence algorithms become a kind of assistance system established diagnostic specialists. As representatives of a cross-sectional subject, pathologists are better suited than others to take on the role of such an information manager, says Bürrig. To what extent the weighting of the diagnostic subjects will shift or new demands are placed on the training is still a long way off: "In the medium term, the disciplines will probably move closer together."

The BDP clearly opposes all tendencies towards centralization in pathology. Pathology should not become a second laboratory medicine even if day-to-day tissue diagnostics business should become much more efficient through machine learning algorithms. In this context, Bürrig referred to the tumor boards, which, like the entire oncological care, are organized decentrally in Germany, from which the patients benefit enormously. The BDP President considers it counterproductive to take pathologists out of this decentralized, interdisciplinary world of care, in which the actors know and trust each other. Rather, the goal must be to use the possibilities of digitization to build functional, decentralized data infrastructures. (gvg)