Machine learning systems and techniques for multispectral amputation site analysis
Certain aspects relate to apparatuses and techniques for non-invasive and non-contact optical imaging that acquires a plurality of images corresponding to both different times and different frequencies. Additionally, alternatives described herein are used with a variety of tissue classification applications including assessing the presence and severity of tissue conditions, such as necrosis and small vessel disease, at a potential or determined amputation site.
1. A tissue classification system comprising:
at least one light detection element configured to collect light reflected from a tissue region;
one or more processors in communication with the at least one light detection element and configured to:
identify at least one patient health metric value corresponding to a patient having the tissue region,
use the at least one patient health metric value to select a classifier from among a plurality of classifiers, each of the plurality of classifiers trained from a different subset of a set of training data, wherein the classifier is selected based on having been trained with a subset of the set of the training data including data from other patients having the at least one patient health metric value;
receive a plurality of signals from the at least one light detection element, a first subset of the plurality of signals representing non-laser light emitted at a plurality of wavelengths and reflected from the tissue region;
generate, based on at least some of the plurality of signals, an image having a plurality of pixels depicting the tissue region;
for each pixel of the plurality of pixels depicting the tissue region:
determine, based on the first subset of the plurality of signals, a reflectance intensity value at the pixel at each of the plurality of wavelengths, and
determine a classification score of the pixel associated with tissue healing potential by inputting the reflectance intensity value into the classifier; and
generate, based on the classification score of each pixel, an overall score associated with tissue healing potential for the plurality of pixels depicting the tissue region.
2. The system of claim 1 , wherein the one or more processors are further configured to determine, based on the classification score of each pixel, a recommended location of an amputation within the tissue region.
3. The system of claim 1 , wherein the one or more processors are further configured to associate each pixel of the plurality of pixels into one of a plurality of tissue categories comprising a healthy tissue category, a necrotic tissue category, or a vascularly diseased tissue category, and to generate, based on the association of each pixel, a mapping of the plurality of tissue categories over the plurality of pixels depicting the tissue region.
4. The system of claim 3 , wherein generating the mapping comprises identifying regions of the plurality of pixels depicting tissue associated with at least one tissue category of the healthy tissue category, the necrotic tissue category, or the vascularly diseased tissue category.
5. The system of claim 4 , wherein the one or more processors are configured to output the image for display, wherein the plurality of pixels depicting the tissue regions are displayed in different colors according to the associated tissue category.
6. The system of claim 1 , wherein the one or more processors are further configured to associate each pixel of the plurality of pixels into one of a plurality of tissue categories comprising viable tissue, necrotic tissue category, or tissue having small vessel disease, based on the classification score of each pixel.
7. The system of claim 1 , wherein the one or more processors are further configured to identify at least one second patient health metric value corresponding to the patient, and wherein the classification score of each pixel is determined by inputting the reflectance intensity value and the at least one second patient health metric value into the classifier.
8. The system of claim 1 , wherein the one or more processors are configured to determine, based on the plurality of signals, a melanin index of the tissue region.
9. The system of claim 8 , wherein the one or more processors are configured to select the classifier based on at least the melanin index.
10. The system of claim 1 , wherein the plurality of signals includes a second subset sequentially captured at a number of times, the one or more processors further configured, for each pixel of the plurality of pixels depicting the tissue region, to:
determine, based on the second subset of the plurality of signals, a photoplethysmography (PPG) amplitude value at the pixel; and
determine the classification score of the pixel based additionally on inputting the PPG amplitude value into the classifier.
11. A system for identifying a recommended location of an amputation, the system comprising:
at least one light detection element configured to collect light reflected from a tissue region in need of amputation; and
one or more processors in communication with the at least one light detection element and configured to:
control the at least one light detection element to capture data representing a plurality of images of the tissue region, the data representing the plurality of images including a first subset each captured using light of a different one of a number of different wavelengths emitted as non-laser light and reflected from the tissue region;
generate, based on at least one of the plurality of images, an image having a plurality of pixels depicting the tissue region;
for each pixel of the plurality of pixels depicting the tissue region:
determine, based on the first subset of the data representing the plurality of images, a reflectance intensity value at the pixel at each of the plurality of wavelengths, and
determine a classification score of the pixel associated with tissue healing potential by at least inputting the reflectance intensity value into a classifier; and
identify, based on the classification score of each pixel, the recommended location of the amputation within the tissue region.
12. The system of claim 11 , wherein the one or more processors are further configured to generate, based on the classification score of each pixel, a mapping of a plurality of tissue categories over the plurality of pixels depicting the tissue region.
13. The system of claim 12 , wherein the plurality of tissue categories includes at least a healthy tissue category, a necrotic tissue category, or a vascularly diseased tissue category.
14. The system of claim 12 , wherein the one or more processors are further configured to output a visual representation of the mapping to a user, wherein the visual representation includes a visual identification of the recommended location for the amputation.
15. The system of claim 11 , wherein the one or more processors are further configured to generate, based on the classification score of each pixel, an overall score associated with tissue healing potential for the plurality of pixels depicting the tissue region.
16. The system of claim 11 , wherein the one or more processors are further configured to associate each pixel of the plurality of pixels into one of a plurality of tissue categories comprising viable tissue, necrotic tissue category, or tissue having small vessel disease, based on the classification score of each pixel.
17. The system of claim 11 , wherein the one or more processors are further configured to identify at least one patient health metric value corresponding to a patient having the tissue region.
18. The system of claim 17 , wherein the one or more processors are further configured to use the at least one patient health metric value to select the classifier from among a plurality of classifiers.
19. The system of claim 18 , wherein the one or more processors are further configured to identify at least one second patient health metric value corresponding to a patient having the tissue region, wherein the classification score of each pixel is determined by inputting the reflectance intensity value and the at least one second patient health metric value into the classifier.
20. The system of claim 11 , wherein the data representing the plurality of images includes a second subset sequentially captured at a number of times, the method further comprising, for each pixel of the plurality of pixels depicting the tissue region:
determining, based on the second subset of the data representing the plurality of images, a photoplethysmography (PPG) amplitude value at the pixel; and
determining the classification score of the pixel based additionally on inputting the PPG amplitude value into the classifier.