What is an analysis model?
When you onboard assays to Reshape, our CX team will help you get the right analysis model configured. An analysis model makes sure we correctly detect and interpret the right things on your plates. The key components that make up an analysis model:
A custom name that makes it clear for everyone in your organization which plates and assays to apply this to. This can be further made clear in the use case description.
A segmentation model (optional): these are our computer vision models that detect individual objects on a plate and quantify, for instance, their color and size.
A classification model (optional): these are our computer vision models that analyze the entire image, and not individual objects in the image. Examples of this could be if there’s salmonella present on the plate or if the plate is too numerous to count,
Metrics: the data points the analysis model outputs for each plate or well, shown in the side panel when you open that specific plate or well.
Metric filters: most metrics can be filtered by color or size, which enables you to, for instance, only count colonies of a specific color or above a certain size.
How to apply an analysis to a job*
There are several ways to apply the right analysis to your job:
Choose the right analysis when creating a job
Inherit the analysis from the project the job is a part of
Apply an analysis after you’ve created a job
How to remove an analysis from a job*
To remove an assay analysis from a job, hit the little trash can icon next to the applied analysis.
*Note: Only users in the same group as the job, can add or remove analyses
Analysis versions
Reshape constantly refines its analyses to be more accurate. It’s up to you to choose how often you want to update your analyses. Your CX team can help you evaluate the performance of a new analysis based on your validation set (see how to validate a new analysis model here). If the new version has a higher accuracy on your validation set, you can apply it to the segmentation or classification model version in your analysis library. From that point, all new jobs will be run using the latest versions. To update to a new version in existing jobs, you can do so through the job overview.
Legacy & archived models explained
During the second half of 2025, Reshape has been rolling out a new type of analysis. These analyses are more flexible and can thus be tailored more to your specific assays, only providing the data points you are looking for.
The analyses used prior to this time will be displayed as legacy models, which will have a grey icon to make them distinct from the newer ones. As long as they are visible in your library, they can still be used.
Under the “Archived models” section, you’ll find previous analyses that have been archived. You will no longer be able to use these analyses when starting new jobs.