Image atlas systems and methods
In some embodiments, a process and system are provided for generating a user interface for classification of a sample image of a cell that includes receiving a sample image of a sample particle from a biological sample and selecting reference images that each portray a reference particle of a biological sample. The reference images can be ordered based on similarity and the reference images can be selected based on the order. The first selected reference image can be aligned with the sample image and expanded such that the adjacent edges of the reference image and sample image are the same. The expanded image can be dynamically filled. The sample image and the expanded reference image can be displayed in a user interface.
1. A method, comprising:
receiving, at a computer system, a sample visual image of a sample particle from a biological sample;
selecting, by the computer system, a plurality of reference images from a database of reference images, each reference image of the plurality of reference images being a reference image of a reference particle from a plurality of reference particles from one or more biological samples;
determining, by the computer system, an order of the plurality of reference images based on a similarity of each reference image to the sample image;
selecting, by the computer system, a first reference image from the plurality of reference images based on the determined order;
physically aligning, by the computer system, the first reference image with the sample image;
dynamically filling, by the computer system, a physical background image of the aligned first reference image; and
displaying, by the computer system, the sample image and the dynamically filled and aligned first reference image in a user interface;
wherein the determining the order of the plurality of reference images comprises:
comparing, by the computer system, the sample image with each reference image of the plurality of reference images; and
determining, by the computer system, the similarity of each reference image to the sample image;
wherein the comparing the sample image with each reference image of the plurality of reference images comprises:
generating, by the computer system, a feature vector for each of the reference images of the plurality of reference images and the sample image; and
calculating, by the computer system, a distance metric for each reference image of the plurality of reference images by comparing the feature vectors of each of the plurality of reference images with the feature vectors of the sample image.
2. The method of claim 1 , wherein features used to generate the feature vector for each of the reference images of the plurality of reference images and the sample image include at least one of a patch size, a particle size, a particle shape, average pixel intensity, and an image mean grayscale value.
3. The method of claim 1 , wherein aligning the first reference image with the sample image comprises:
determining, by the computer system, a first display location for displaying the sample image in the user interface, the first display location being within a first cell of a grid of cells;
determining, by the computer system, a second display location adjacent to the first display location for displaying the first reference image in the user interface, the second display location being within a second cell of the grid of cells; and
aligning, by the computer system, a centroid of the first reference image within the second display location with a centroid of the sample image within the first display location.
4. The method of claim 3 , wherein dynamically filling the background of the aligned first reference image comprises:
expanding, by the computer system, the background of the first reference image within the second cell to fill the second cell;
determining, by the computer system, a background color of the first reference image; and
filling, by the computer system, the expanded background of the first reference image with the background color.
5. The method of claim 4 , wherein determining the background color of the first reference image comprises:
determining, by the computer system, four corner pixels of the first reference image; and
calculating, by the computer system, a mean color value of the four corner pixels.
6. The method of claim 1 , wherein the plurality of reference images are selected based on a tag associated with each reference image of the plurality of reference images.