REFLECTIVE MODE MULTI-SPECTRAL TIME-RESOLVED OPTICAL IMAGING METHODS AND APPARATUSES FOR TISSUE CLASSIFICATION
Certain aspects relate to apparatuses and techniques for non-invasive 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 burns and other wounds.
1 . A multispectral imaging system for analyzing wound tissue, the multispectral imaging system comprising:
at least one light emitter configured to emit each of first and second wavelengths of light to illuminate patient tissue;
a light detection element configured to collect light emitted from the at least one light emitter and reflected from the patient tissue;
one or more processors in communication with the at least one light emitter and the light detection element and configured to:
control the at least one light emitter to emit each of the first and second wavelengths of light toward a tissue region of a patient, the tissue region including at least a portion of a wound;
receive multispectral image data from the light detection element, the multispectral image data including at least a first image corresponding to light emitted at the first wavelength of light reflected from the tissue region and at least a second image corresponding to light emitted at the second wavelength of light reflected from the tissue region;
input the first and second images into a machine learning model trained to evaluate wound bed tissue;
generate, using the machine learning model, an output representing a characteristic of a wound bed of the wound; and
based on the output of the machine learning model, output information identifying viable wound bed tissue within the wound bed.
2 . The multispectral imaging system of claim 1 , wherein the machine learning model is trained to quantify a healing potential of wound tissue, and wherein the output represents a healing potential of the at least a portion of the wound.
3 . The multispectral imaging system of claim 1 , wherein the machine learning model is trained to classify areas of imaged tissue into one or more tissue classes, wherein the output represents at least one tissue class associated with the at least a portion of the wound, and wherein the information includes an area of the at least a portion of the wound.
4 . The multispectral imaging system of claim 3 , wherein the wound comprises a burn, and wherein the one or more processors are configured to determine a percentage burned surface area of the patient based on the area of the burn.
5 . The multispectral imaging system of claim 4 , wherein the information includes a classified image representing the percentage burned surface area of the patient.
6 . The multispectral imaging system of claim 4 , wherein the one or more processors are configured to:
determine an additional percentage burned surface area of an additional patient based on an additional output of the machine learning model when provided with additional images captured at least at the first and second wavelengths of an additional tissue region of the additional patient; and
output information for performing mass-casualty burn care triaging of at least the patient and the additional patient based on the percentage burned surface area of the patient and the additional percentage burned surface area of the additional patient.
7 . The multispectral imaging system of claim 4 , wherein the one or more processors are configured to determine a treatment for the patient based at least partly on the percentage burned surface area of the patient, wherein the information includes an indication of the treatment.
8 . The multispectral imaging system of claim 7 , wherein the one or more processors are configured to determine an amount of fluid to administer to the patient based at least partly on the percentage burned surface area of the patient, wherein the treatment includes the amount of fluid.
9 . The multispectral imaging system of claim 1 , wherein the one or more processors are configured to:
control the at least one light emitter and the detector to capture photoplethysmographic data representing changes in blood volume in the tissue region over a period of time; and
additionally input the photoplethysmographic data into the machine learning model.
10 . A method for analyzing wound tissue, the method comprising:
controlling at least one light emitter to emit each of first and second wavelengths of light to illuminate a tissue region of a patient, the tissue region including at least a portion of a wound;
receiving multispectral image data from a light detection element configured to collect light emitted from the at least one light emitter and reflected from the patient tissue, using the multispectral imaging system of claim 2 , wherein the multispectral image data includes at least a first image corresponding to light emitted at the first wavelength of light reflected from the tissue region and at least a second image corresponding to light emitted at the second wavelength of light reflected from the tissue region;
inputting the first and second images into a machine learning model trained to evaluate wound bed tissue;
generating, using the machine learning model, an output representing a characteristic of a wound bed of the wound; and
identifying viable wound bed tissue within the wound bed based on the output of the machine learning model.
11 . The method of claim 10 , further comprising determining a treatment for the patient.
12 . The method of claim 11 , further comprising determining, based on the output of the machine learning model, that the patient requires a surgical procedure.
13 . The method of claim 10 , wherein the machine learning model is trained to classify areas of imaged tissue into one or more tissue classes, wherein the output represents a number of pixels associated with at least one tissue class, the method further comprising determining an area of the tissue region based on the number of pixels.
14 . The method of claim 13 , wherein the tissue region comprises a wound, the method further comprising determining a percentage surface area of the wound of the patient based on the area of the tissue region.
15 . The method of claim 14 , further comprising outputting a classified image representing the percentage surface area of the wound of the patient.
16 . The method of claim 14 , further comprising:
determining an additional percentage surface area of a wound of an additional patient based on an additional output of the machine learning model when provided with additional images captured at least at the first and second wavelengths of an additional tissue region of the additional patient; and
performing triaging of at least the patient and the additional patient based on the percentage surface area of the wound of the patient and the additional percentage surface area of the wound of the additional patient.
17 . The method of claim 14 , further comprising determining a treatment for the patient based at least partly on the percentage surface area of the wound of the patient.
18 . The method of claim 17 , wherein said wound is a burn and further comprising determining an amount of fluid to administer to the patient based at least partly on the percentage burned surface area of the patient, wherein the treatment includes the amount of fluid.
19 . The method of claim 10 , wherein the tissue region comprises a wound and wherein the machine learning model is trained to quantify a healing potential of wound tissue, the method further comprising determining the healing potential of the wound based on the output of the machine learning model.
20 . The method of claim 19 , wherein the wound is one of a diabetic foot ulcer or an amputation site, the method further comprising predicting the healing potential of the diabetic foot ulcer or the amputation site.