Digital Pathology

Klaus Kayser MD PhD, Professor of Pathology and Epidemiology, Dr. rer.nat. (Physics), Dr. med, Dr. honoris causae mult. headed the Institute of Pathology, Thoraxklinik, University of Heidelberg until 2005. The former faculty member of Heidelberg and Berlin University (now Charite Berlin) is a pioneer in electronic medical communication, research on image ana- lysis, lectins, structural entropy, and lung cancer. 5 New study strongly encourages the use of digital image analysis in Breast Cancer – the most common cause of cancer death among women worldwide In a study of 436 breast cancer cases with 28 years of survival data histo-pathologists from Karolinska Institute, Stockholm, Sweden shows that protein tissue-based biomarker data are at least as good as gene expression assays. costs will be a thing of the past. While developments in terms of data transmission are stagnating – whether the hoped-for revamping of the internet with fibre op- tics, or the use of satellite telephones. for example in Africa, will spell progress remains to be seen – the ima- ging market is immense with hardware and software solutions becoming more and more affordable.What’s really missing is an umbrella organisation bringing toge- ther pathologists and lab and IT experts. The wait-and- see attitude of the industry is a definite obstacle: many companies have excellent IT staff who potter about wit- hout understanding the work of the pathologist. On the other hand there are many specialist physicians who are highly interested in IT but lack the necessary knowledge. Communication does not really happen! In the now defunct GDR mathematicians and physicists frequently worked in institutes of pathology alongside their colleagues in medicine and to a large extent it was this direct access that enabled them to develop innovations. This very effective cooperation was abo- lished and today in Germany there is a slew of insti- tutes – Fraunhofer, Max-Planck and the German Cancer Research Centre – all of which work in different and dis- tinct areas. However, when it comes to application-spe- cific issues, close spatial proximity is the non plus ultra. The almost philosophical contemplation of the relati- onship between structure and function is a topic only very few pathologists are interested in. In biology it is a matter of the inside and the outside. When observed long enough, a structure will turn into a function. This approach opens a different view on pathology: Today, no theory, be it energy balance, metabolism, or any one of the common physical-chemical concepts, can explain why a cancer lesion of 2 cm diameter can destroy the entire system and kill the patient. ‘Cardiovascular failu- re’ is nothing but a catch-all phrase because, in the end, structure-associated functions defined on the gene level determine what does not function and why. Which para- meter is it that triggers a domino effect that causes the human system to collapse? Digital pathology may well help to understand the construction of ‘life as such’. RALF MATEBLOWSKI When summarizing the study from Karolinska Institute, Stockholm, Sweden, manual assessment of the biomarkers ER, PR, HER2, and Ki67, with an emphasis on the latter, is in most aspects an inferior alternative to Digital Image Analysis (DIA). When comparing sensiti- vity and specificity to PAM50 subtype classification and prognostic power DIA outperforms manual scoring. Being able to reduce time to diagnosis as well as the overall cost for breast cancer testing are good news. What makes it even more convincing is that all subtypes including the Luminal B, which is widely considered being the most challenging distinction in surrogate sub-classification, produced a slightly better concordance and Cohen’s agreement. The cost of combining automated DIA with biomarker analysis is a fraction of the cost of a gene expression test, which is time consu- ming to conduct and often has to be done outside the histopathology lab. Combining biomarker analysis with DIA creates value for all par- ties involved: patients, pathologists and payers. For patients it is to receive a faster treatment, for pathologists an opportunity to reduce time consumption and allocate precious resources to more qualified tasks and for payers a much more cost efficient diagnostic method. An operator of the Visiopharm DIA system has the option to manually evaluate the digitized slide image and perform the diagnosis based on this, but DIA also allows the computer to conduct an automated analysis avoiding subjective assessments of biomarker positivity. This implicates that an approach like DIA to count full tumor cross-sections or to make automated ‘hot spot’ analysis could free up time for hi- stopathologists, biomedical scientists or other laboratory personnel. “The desire for faster and more cost-effective solutions for the hi- stopathology laboratories need no longer be a dream. In complex heterogeneous tumors DIA in combination with immunohistochemi- stry has shown to be a viable competitive option for laboratories lacking in-house gene expression testing,” says Helle Fisker, CMO at Visiopharm. The DIA system used in the study was the Visiopharm DIA software VIS 4.6.3 Wednesday, May 24th • 5.30 PM–7 PM • Room: Hörsaal History of the European conference series on digital pathology: memories and perspectives K. Kayser, Heidelberg, Germany Digital Image Analysis (DIA) a viable and competitive, if not superior, alternative for gene expression testing in breast cancer

Pages Overview