Cellular diagnosis for the clinicians
Attila Tárnok · Cytometry Part A · 2017
Taking advantage of cells for advanced clinical diagnosis is a long-standing aim in quantitative single cell sciences. For years, clinicians have realized that there is an urgent need for making cellular diagnosis (i.e., cytopathology and histopathology) reproducible, less influenced by the level of experience of the human observer and less time consuming. However, it turns out that it is long and stony road to make computers understand what cells and subcellular compartments are in a grey scale or in a color image. This year's special issue on “Computer-aided Diagnostics in Digital Pathology” guest edited by Bengtsson et al. 1 further focused on innovations in this area. Recently, computational approaches like deep learning became popular for identifying and quantifying biological structures in tissues as investigated for breast cancer 2. These approaches harbor the promise to strengthen reproducibility in cellular diagnostics 3. The present issue contains several articles addressing clinical questions to be answered by the cells and their images using computing methods to increase reliability. The important thing about these articles is that they are jointly authored by practitioners (physiologists) from specific clinical areas, basic scientists and information scientists, demonstrating how essential collaboration between specialists is herein. Among these collaborations is the work by Bombrun and colleagues (this issue, page 1068), highlighted by a Commentary from Doan (this issue, page 1051), and was selected by me as one of two Editor's Choice articles of the month. Here, the authors demonstrate a new high-content analysis approach to identify and quantitate lipid droplets that play an important role in adipogenesis in health and disease. Peikari and colleagues (this issue, page 1078) developed an automated method to analyze material from surgical breast tumors and determine their cellularity. This parameter is of high relevance for clinical decision making in breast cancer. The new method that works with whole slide images has a comparable performance to pathology experts when tested on real-world specimens, but clearly reduces their workload. Do and colleagues (this issue, page 1088) used Imaging Flow Cytometry to visualize chromosomal deletions and cell sorting for further molecular biological analysis. Here, the combination of (flow) imaging that led to improved detection and isolation, and is of high relevance for applying individualized therapy. Finally, the second Editor's Choice is by Leers and colleagues (this issue, page 1059). They investigated the relevance of the major subgroups of circulating monocytes in different types of coronary heart diseases. Although they applied slightly different gating approaches to identify classical, non-classical and intermediate monocytes than described in an earlier article by Zawada et al. 4, they could confirm previous findings on the correlation of coronary heart disease monocyte subset abundance. However, this study goes beyond and shows the correlates of antigen expression and NSTEMI and STEMI type of cardiac infarction, opening for new and more specific diagnosis. This editorial was drafted on my flight from Singapore, and still under the strong impressions of CYTOAsia in Singapore; the first large CYTO conference and education event in Asia. In a nutshell, it was the consensus that CYTOAsia was a great success and demonstrates the impressive activity and engagement particularly from young scientists in this part of the world. Read more about it in an upcoming issue of Cytometry Part A.