Systems and methods for detecting disease using mobile device
A video processing method improves the repeatability of point-of-care (POC) mean pixel intensity (MPI) analysis, for example by rejecting low-quality input data using image features. Using these features, the method selects high-quality frames and achieves an improvement in correlation and standard error for the MPI-absorbance relationship. In another example, a mobile device (e.g., smartphone) based image processing method produces an MPI output with smaller variances, lower limits of detection (LODs), and a higher dynamic range than existing methods. This method has improved performance in the presence of additive and multiplicative ambient light noise.
1. A smartphone programmed classification method to reject low-quality input frames, comprising:
collecting analyte sample image data representing frames of a video with a smartphone;
extracting region of interest data of the analyte sample from the image data;
tracking the region of interest data from frame to frame throughout the video with the smartphone;
automatically identifying and rejecting video frames having problematic images thereon with the smartphone;
automatically selecting non-rejected video frames; and
outputting the selected video frames for analyte imaging time-averaged mean pixel intensity analysis.
2. The method of claim 1 , wherein the non-rejected video frames have a higher correlation and lower standard error than the rejected video frames.
3. The method of claim 1 , wherein the collected image data includes video of a well plate having an analyte sample within the well plate.
4. The method of claim 1 , wherein the extracting region of interest data includes using Shi-Tomasi corner detection to obtain a set of corner points, determining an optical flow for each corner point, and updating a location of a region of interest center of each frame based on the nearest corner points.
5. The method of claim 1 , wherein the problematic images include blurred images, images with excessive white-balancing, images having excessive reflection.
6. A smartphone programmed color space transform method for analyzing intensity change assays, comprising:
collecting image data of an analyte sample with a smartphone;
extracting region of interest data of the analyte sample from the image data, the region of interest data including red, green, and blue data;
converting the region of interest data to processed region of interest data with the smartphone, the processed region of interest data including hue-saturation-value color space data having a saturation channel corresponding to the chromatic intensity of the region of interest data; and
quantifying the saturation channel of the processed region of interest data to determine absorbance and concentration of the analyte sample.
7. The method of claim 6 , wherein the extracting region of interest data of the analyte sample includes selecting the region of interest data from the image data and isolating the selected region of interest data in a well plate or any other colorimetric assay.
8. The method of claim 6 , wherein the converting the region of interest data to processed region of interest data includes transforming the red, green, and blue coordinates to hue-saturation-value color space data.
9. The method of claim 8 , wherein the converting the region of interest data to processed region of interest data includes determining the arithmetic mean of pixel intensities in the saturation channel as a mean pixel intensity, and the mean pixel intensity is used in the quantifying step to quantify the analyte sample.
10. The method of claim 6 , wherein the hue-saturation-value color space data includes non-normalized hue-saturation-value color space data.
11. A smartphone programmed color space transform method for analyzing intensity change assays, comprising:
collecting analyte sample image data representing frames of a video with a smartphone;
extracting region of interest data from the image data with the smartphone, the region of interest data including red, green, and blue data;
tracking the region of interest data from frame to frame throughout the video with the smartphone;
automatically identifying and rejecting video frames having problematic images thereon with the smartphone;
automatically selecting non-rejected video frames;
converting the region of interest data to processed region of interest data with the smartphone, the processed region of interest data including hue-saturation-value color space data having a saturation channel corresponding to the chromatic intensity of the region of interest data; and
quantifying the saturation channel of the processed region of interest data to determine absorbance and concentration of the analyte sample.
12. The method of claim 11 , wherein the non-rejected video frames have a higher correlation and lower standard error than the rejected video frames.
13. The method of claim 11 , wherein the collected image data includes video of a well plate having an analyte sample within the well plate.
14. The method of claim 11 , wherein the extracting region of interest data includes using Shi-Tomasi corner detection to obtain a set of corner points, determining an optical flow for each corner point, and updating a location of a region of interest center of each frame based on the nearest corner points.
15. The method of claim 11 , wherein the problematic images include blurred images, images with excessive white-balancing, images having excessive reflection.
16. The method of claim 11 , wherein the extracting region of interest data of the analyte sample includes selecting the region of interest data from the image data and isolating the selected region of interest data in a well plate or any other colorimetric assay.
17. The method of claim 11 , wherein the converting the region of interest data to processed region of interest data includes transforming the red, green, and blue coordinates to hue-saturation-value color space data.
18. The method of claim 17 , wherein the converting the region of interest data to processed region of interest data includes determining the arithmetic mean of pixel intensities in the saturation channel as a mean pixel intensity, and the mean pixel intensity is used in the quantifying step to quantify the analyte sample.
19. The method of claim 11 , wherein the hue-saturation-value color space data includes non-normalized hue-saturation-value color space data.