This association is carried out if the cell in the subsequent

frame happens to be within a threshold distance r. However, erroneous associations may occur depending on the value of this threshold distance, especially Selleck BMS 907351 at high densities of cells or in crowded regions. In the case of TIAM, tracking accuracy is quite robust to changes in the value of threshold distance r, at least at the density of cells present in the benchmark experiments ( Fig. 3b, Table 1). Finally, we compared the overall performance of TIAM with some of the other well-known tools such as DYNAMIK (Jaeger et al., 2009), Icy (de Chaumont et al., 2012), Imaris (from Bitplane), and Volocity (from PerkinElmer). SFDA and ATA provide a direct way for such comparisons as they offer a single, comprehensive measure of accuracy

of detection and tracking, respectively. SFDA and ATA were computed for results from all the tools on both the benchmark experiments. selleck screening library TIAM performed better than the other tools both in detection and tracking (Table 1, Videos S1 and S2). Extraction of features from the multi-channel image series and integration of these features with tracking results is a unique capability of TIAM. Whereas tools such as Volocity, CellProfiler and TACTICS can report on additional channels based on the mask created by global thresholding of the primary channel, TIAM handles every channel separately and performs local segmentation in each one of them. We sought to assess how well TIAM is able to perform in segmenting transmitted light, reflection, and fluorescence images and in extracting information on polarity, contact area, and mean fluorescence intensity, respectively. We again did this by comparing against ground truth that was established manually based on personal Amobarbital expertise. Outlines of cells in DIC, reflection and fluorescence images drawn by TIAM were in good agreement with those from the ground truth (Video S3, Video S4 and Video S5). Measurement of aspect ratio as a readout of morphological polarity from outlines

in DIC image series was reasonable, but not very good (Fig. 4a, Fig. S10). We have nonetheless decided to include it as part of TIAM due to its potential value for interpretation on the biology being studied. The contact area and mean pixel intensity of cells measured from outlines of cells from reflection and fluorescence images, respectively, were in good agreement with the ground truth (Fig. 4b and c). The median absolute error in measurements was below 10% for both (Fig. S10). The systematic bias towards higher values in reporting mean fluorescence intensity was due to higher threshold values chosen by the Otsu’s method used for local segmentation (Video S5). Along with the accuracy of calculations, processing time is also crucial to the end-user’s considerations.