Point set registration for combining fluorescence microscopy methods
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (Arcadia Science)
Abstract
Implementation of combined microscopy methods provides valuable information across various scientific applications. However, aligning the datasets and finding the correct point correspondence poses a challenge, especially for large, randomly distributed point sets that are subject to positional errors and missing points. Here, we provide a three-step procedure to perform point set registration, which can be applied to datasets with millions of points and stays robust even when only 10% of the points correspond. In the first global step, the scaling and rotation parameters for the imaging systems are determined once on a smaller calibration dataset using a geometric hashing algorithm. When the global transformation is known, full experimental datasets can be registered by performing step two: a course registration using cross-correlation, and step three: a precise registration to fine-tune the transformation. After these three steps, point correspondence is determined by setting a distance threshold based on a statistical model of random point sets that additionally provides the matching error. We have demonstrated its successful implementation in coupling fluorescence and sequencing methodologies. To enable wide application of these point set registration and correspondence algorithms we provide a python library called MatchPoint.
Article activity feed
-
only one tile can be matched at a time
Sorry, but this also confuses me. How can this be considered a global transformation when only one tile is matched at a time for translation? I would have assumed that the tiles are roughly in place from using e.g. stage coordinates as a crude estimate of tile position.
-
Although the algorithm can be used to only find the scaling and rotation parameters, the use of additional parameters will result in a clearer cluster, as the incorrectly matched pairs will be distributed over more dimensions. Therefore, it is beneficial to also add translation
Why do you say the algorithm can be used to "only find the scaling and rotation parameters" when you immediately add translation? Do you mean only scaling and rotation are solved for in the original implementation of the algorithm and your adaptation adds translation?
-
When examining the transformation parameters that belong to pairs of resembling quadruplets, the correctly matched quadruplet pairs will cluster, as they all have similar transformations; on the other hand, the incorrectly matched pairs will be spread randomly across a larger space.
Your geometric hashing approach sounds similar to RANSAC. I am curious if you compared the two and what the tradeoffs between time and accuracy look like?
-
since the imaging parameters of the sequencer were unknown
Hello, I am not very familiar with this imaging modality but it seems rather strange that the imaging parameters are unknown. Is this typically the case?
-