Digital Pathology

8 Challenges in digital pathology We expect to see some changes Strictly speaking, digital pathology has not yet resulted in any groundbreaking changes for clinical diagnostics. The conventional light microscope introduced to pathology around 100 years ago continues to be the most important tool for pathologists. Nevertheless, in the future, according to private lecturer Dr Frederick Klau- schen, Head of the Molecular and Systems Pathology Group and Consultant at the Institute for Pathology at the Charité Clinic Berlin, we can expect to see some changes from the introduction of digital technology. ‘Digital pathology is supporting us already in making information from tissue sections more easily quantifiable with the help of computer assisted systems,’ says patholo- gist Dr Frederick Klauschen. ‘However, when it comes to pattern recognition, and there- fore tumour typing and classification into malignant or non-malignant tumours, the pathologist will remain superior to the com- puter for the foreseeable future.’ Let the computer do the counting To date, the biggest innovation is a more objective and standardised view and quan- tification of certain tissue characteristics fa- cilitated by image analysis procedures. ‘One example of this is the measurement of the proliferation index,’ Klauschen says. This can be determined via the immunohistoche- mical detection of the protein Ki67, which is found in the cell nucleus of proliferating cells. The result normally shows some indivi- dual cell nuclei in the normal histological co- lour (blue) and other cell nuclei where Ki67 is detected in brown or red shades. Counting the frequency of the brownish colours was once something the patholo- gist had to do ‘manually’. Now, however, the counting can be done with the help of a computer, using representative image regions. ‘We have developed a specific pro- gramme for this purpose, the so-called Ki67 Quantifier. This software supports us with the counting and determination of the pro- liferation index. It facilitates standardised, automated and precise quantification,’ Klau- schen explains. ‘We work very closely with the German Bre- ast Group with regards to breast cancer, for the validation of such procedures.’ Tumour samples that arrive, from all over Germany and other countries, are examined in the Institute for Pathology at the Charité Clinic, which acts as a pathology reference centre. The above-mentioned software is then utili- sed in the context of these studies and vali- dated based on clinical data. Digital procedures are currently being deve- loped and tested to examine different types of tissue characteristics and markers. ‘Ho- wever, many of these procedures are not yet ready for use,’ the pathologist points out, adding: ‘although some Institutes are already using these software solutions it will take some time before all areas of diagnos- tics will benefit from them.’ Common problem – lack of standards One big problem with the practical applica- tion of image analysis procedures is the lack of standards. Although there is a dialogue and exchange between the various areas of pathology, the analysis of tissue slides from different labs is difficult, the specia- list regrets. There are no unified standards regarding histological colourings. ‘Each la- boratory produces slightly different depths of colours, so the sections differ from one another. Pathologists are not normally wor- ried by a stronger pink or blue – a computer, however, then may have problems with the classification. The objective is the develop- ment of procedures which are independent of colour,’ he emphasises. Although previ- ously not an important topic in pathology, these types of standardisation are works in progress, having gained importance only with the introduction of digitisation. Klauschen sees a further difficulty in the practical implementation of digitisation in day-to-day pathology. ‘Although we have the technology available, scans of sections are only compiled for special studies, confe- rences, or for research projects. The actual pathology workflow is not digitised.’ One reason for this is the vast number of tissue sections produced at the Institute every day and the related expenditure of time for the digitisation of each section. The data volume of the images still repre- sents a big problem as pathological images have much larger file sizes than those in ra- diology, for instance. ‘We have amounts of data which simply represent big challenges for the IT infrastructure. There is room for development here, particularly regarding the archiving of all this data.’ Klauschen is awa- re of only two institutes where the workflow 1. Ki57 Quantifier: Determination of the Ki67 proliferation index in breast cancer. The image shows the original immunohistological image (brown: Ki67-positive (proliferating) cells, blue: Ki67-negative (non-proliferating cells)

Pages Overview