Excursus Particle Sizes - Konnsy/REAML2022-hackathon GitHub Wiki

You can extend the system created for Task 2 to determine a size distribution offering additional information on the sample at hand. Since the antibody layer used in the instrument reacts only to certain particles, knowledge about them can be combined with the detections to achieve verifiability of the results and to make statements about possible infectivity. After calibration of the system, with reference samples whose size distributions are known, detected size distributions can be used to check whether the predicted sizes are within the range plausible for a pathogen. In this way, outliers can be detected as false detections. The determined counts can also be linked to domain knowledge for plausibility checks and can help decide if critical thresholds for the infectivity are reached. This can lead to a more informative and a more trustworthy system.

The calculation of the relative particle sizes can be considered as an extension. Since one trace is seen as one particle, we can determine particle sizes as intensity differences between the times before and after the particle attaches.

To accomplish this, we consider the area in which the particle of interest appears spatially and average this value for each frame. We can then look at these values over time and detect a step in the time series:

The difference between the mean value after and the mean value before the particle attachment is proportional to the actual particle size. By determining this value for each trace, we can eventually produce a size distribution that lists how often each particle size occurs in the sample under study.

Below you can see the extension of the example pipeline from Task 2 which contains an additional module to estimate the sizes of contained particles:

Here you can see example distributions visualized in histograms showing particle counts belonging to specific intensity intervals: