System and method of classification of biological particles
A method and system for classification of cells and particles in a biological sample using an automated image-based feature extraction and classification architecture. A method operates by applying a mask or series of masks to an image, extracting features from the unmasked portions of the image based on the content and location of colored pixels, selecting a subset of the extracted features, and mapping the subset of the extracted features into a classifier architecture. In a majority of cases, the first level model architecture provides an accurate identification of the cell or particle. In a minority of cases, the classification of the cell or particle requires a second level step requiring the use of numerical or categorical values from the first level in combination with a second level model.
1. A method of determining a classification of a particle in a biological sample comprising:
acquiring an image of the particle with a digital camera;
receiving the image of the particle at a processor system;
utilizing the processor system to execute computer executable code stored on a non-transitory computer readable medium, the computer executable code comprising instructions that, when executed on the processor system, cause the processor system to:
apply a first mask to the image;
acquire a first set of pixels from the image based on applying the first mask;
apply a second mask to the image, wherein applying the second mask reveals different pixels than the first mask;
acquire a second set of pixels based on applying the second mask;
extract a plurality of features from the first and second set of pixels; and
determine the classification based on at least a subset of the extracted features;
wherein the application of each mask reveals different pixels concentrically outside a center of the image.
2. The method of claim 1 , further comprising defining a center of the image.
3. The method of claim 1 , further comprising normalizing the image to a size of the first mask.
4. The method of claim 1 , further comprising applying the first mask substantially to a center of an image normalized to the size of the first mask.
5. The method of claim 1 , wherein the first mask and the second mask are circular or ring-shaped.
6. The method of claim 1 , wherein the extracting comprises clustering the first set of pixels into a group, creating a color palette from the clustered group of pixels, determining a label for the image based in part on the color palette, or any combination thereof.
7. The method of claim 1 , wherein the extracting comprises normalizing the first mask to a unit magnitude, using a chosen color space, including red-green-blue (RGB) hue-saturation-value (HSV), hue-saturation-lightness (HSL), or hue-saturation-brightness (HSB), or any combination thereof.
8. The method of claim 1 , further comprising: using a first level model to compare the subset of the extracted features to a previously stored data set, identifying a preliminary classification based on the comparison of the subset of the extracted features to the previously stored data set, calculating a probability value that the preliminary classification is correct using a first level model, determining the classification based on the preliminary classification when the probability value is at or above a threshold value.
9. The method of claim 8 , wherein the first level model is a machine learning model.
10. The method of claim 8 , further comprising using a second level model to determine the particle classification when the probability value is below the threshold value.
11. The method of claim 10 , wherein the second level model is a machine learning model.
12. The method of claim 11 , wherein the second level model further comprises:
receiving the probability value at the second level model;
creating a sorted list of values according to a classification performance in relation to a particle category;
combining the probability value and the sorted list to create a second level probability value;
using the first level probability value and the second level probability value to determine the particle classification.
13. The method of claim 1 , wherein the subset of the extracted features comprises training features, validation features, or testing features.
14. The method of claim 1 , wherein the subset of the extracted features is mapped into a cascade classifier architecture.
15. The method of claim 1 , wherein the particle comprises a member selected from the group consisting of a neutrophil, a lymphocyte, a monocyte, an eosinophil, a basophil, an immature white blood cell, a reticulocyte, a nucleated red blood cell, an erythrocyte, an epithelial cell, a bacterium, a yeast, or a parasite.
16. The method of claim 1 , further comprising an extraction routine and a mapping routine both executed by the processor system, wherein the extracting the plurality of features from the first and second set of pixels is part of the extraction routine and a mapping at least the subset of the extracted features into a classifier to determine the classification is part of the mapping routine.
17. A non-transitory computer-readable storage medium including program instructions executable by one or more processors that, when executed, causes the one or more processors to perform operations, the operations comprising:
acquiring an image of a particle in a biological sample;
receiving, at a processor system, the image of the particle;
executing, using a processor system, computer executable code stored on a non-transitory computer readable medium, the computer executable code comprising instructions that, when executed on the processor system, cause the processor system to:
apply a first mask;
acquire a first set of pixels from the image based on applying the first mask;
apply a second mask to the image, wherein applying the second mask reveals different pixels than the first mask;
acquire a second set of pixels based on applying the second mask;
extract a plurality of features from the first and second set of pixels; and
determine the classification based on at least a subset of the extracted features;
wherein the application of each mask reveals different pixels concentrically outside a center of the image.
18. The method of claim 17 , further comprising: using a first level model to compare the subset of the extracted features to a previously stored data set, identifying a preliminary classification based on the comparison of the subset of the extracted features to the previously stored data set, calculating a probability value that the preliminary classification is correct using a first level model, determining the classification based on the preliminary classification when the probability value is at or above a threshold value.
19. The method of claim 17 , further comprising an extraction routine and a mapping routine both executed by the processor system, wherein the extracting the plurality of features from the first and second set of pixels is part of the extraction routine and a mapping at least the subset of the extracted features into a classifier to determine the classification is part of the mapping routine.