The approach of lymphocyte (CD4+, CD3+) testing proposed in this project is one step further in the direction to develop an alternative to the flowcytometric cell-analysis, termed FACS, which is today used in larger facilities, e.g. hospitals, for leukemia and infectious disease diagnostics. A promising alternative to FACS would be a low-cost cell analysis, for de-centralized laboratories in addition to the larger hospitals and centralized laboratories. Here we aimed at developing simpler assays than the current state-of-art. In one approach the cells analyzed had to be lysed and then CD4 analyzed. Similarly in further approach the DNA expression of CD4 was analyzed in cell lysates. In turn, further rapid assays where no further steps like cell lysis in the sample preparation is needed are warranted. Our initial idea to use smart derivatized surfaces for CD4 lymphocyte trapping proved unsuccessful. Trapping by derivatized latex beads followed by a read-out using increased scattering caused by latex agglutination was found to be feasible. Interestingly this approach has never been described as such in the literature, although it has many advantages in terms of simplicity. For example, the analysis is run rapidly within a few minutes.
The objective of the malaria detection was to automatize the detection of infected blood cells by malaria and to estimate the ratio of infected cells to the total number of red blood cells. This ratio is an important indicator of how to treat the patient and current practice let a technician perform this analysis, with a very high error rate. No generic solution exists for now, as the feature extraction is far from easy and very dependent on the optical setup. The images used during the project were acquired at CHUV by Professor Guy Prod’hom at the end of 2014 and showed a high degree of illumination inhomogeneity, even the color is not homogeneous in the left and right part of the field.
Two domains were investigated for quantification of infected cells: standard image processing and learning-based algorithms. The result is a software able to learn thanks to a supervised learning algorithm and to automatically calculate the ratio of infected cells. This unique tool is now available on GitHub with no equivalent in the open source community.