SPECTRAL IMAGING SYSTEMS AND METHODS FOR HISTOLOGICAL ASSESSMENT OF WOUNDS
The present disclosure relates to systems and methods for assessing or predicting the status of wounds such as burns. Systems can include at least one light detection element and one or more processors configured to receive a signal from the at least one light detection element representing light reflected from a tissue region, generate an image having a plurality of pixels depicting the tissue region, and determine a burn status of a subset of pixels of the image using one or more deep learning algorithms. Systems can further be configured to generate a classified image of the tissue region and/or determine a predictive score associated with healing of the wound.
1 . A system for assessing or predicting wound status, the system comprising:
at least one light detection element configured to collect light of at least a first wavelength after being reflected from a tissue region comprising a burn; and
one or more processors in communication with the at least one light detection element and configured to:
receive a signal from the at least one light detection element, the signal representing light of the first wavelength reflected from the tissue region;
generate, based on the signal, an image having a plurality of pixels depicting the tissue region;
determine, based on the signal, a reflectance intensity value at the first wavelength for each pixel of at least a subset of the plurality of pixels;
determine, using at least one deep learning (DL) algorithm, a burn status corresponding to each pixel of the subset of pixels depicting the tissue region; and
generate a classified image based at least in part on the image and the determined burn status corresponding to each pixel of the subset of pixels depicting the tissue region.
2 . The system of claim 1 , wherein the classified image comprises pixels having different visual representations based on the burn status corresponding to each pixel.
3 . The system of claim 1 , wherein the one or more processors are further configured to cause a visual display of the classified image.
4 . The system of claim 1 , wherein the burn status corresponding to each pixel is selected from a non-healing burn status and a healing burn status.
5 . The system of claim 1 , wherein the burn status corresponding to each pixel is a status associated with burn depth.
6 . The system of claim 5 , wherein the burn status corresponding to each pixel is selected from a first degree burn status, a superficial second degree burn status, a deep second degree burn status, and a third degree burn status.
7 . The system of claim 1 , wherein the burn status corresponds to necrosis of adnexal structures within at least a portion of the burn.
8 . The system of claim 7 , wherein determining the burn status corresponding to each pixel of the subset of pixels depicting the tissue region comprises identifying a percentage of necrotic adnexal structures within the at least a portion of the burn.
9 . The system of claim 8 , wherein a non-healing burn status corresponds to necrosis of greater than 50.0% of the adnexal structures.
10 . The system of claim 8 , wherein a non-healing burn status corresponds to necrosis of greater than 0.0% of the adnexal structures.
11 . The system of claim 1 , wherein the at least one DL algorithm comprises a convolutional neural network.
12 . The system of claim 11 , wherein the convolutional neural network comprises a SegNet.
13 . The system of claim 1 , wherein the at least one DL algorithm comprises an ensemble of a plurality of DL algorithms.
14 . The system of claim 13 , wherein the at least one DL algorithm comprises a weighted averaging ensemble.
15 . The system of claim 13 , wherein the at least one DL algorithm comprises a TPR ensemble.
16 . The system of claim 1 , wherein the at least one DL algorithm is trained using a wound database.
17 . The system of claim 16 , wherein the wound database comprises a burn database.
18 . The system of claim 1 , wherein the at least one DL algorithm is trained based at least in part on a plurality of ground truth masks, wherein at least some of the ground truth masks are generated based at least in part on the presence of necrotic adnexal structures in burn tissue biopsies.
19 . The system of claim 1 , wherein the one or more processors are further configured to determine, based at least in part on the burn status corresponding to each pixel of the subset of pixels depicting the tissue region, a predictive score associated with healing of the burn over a predetermined time interval following generation of the image.
20 . The system of claim 19 , wherein the predictive score corresponds to a probability of healing without surgery or skin grafting.
21 . The system of claim 19 , wherein the predetermined time interval is 21 days.
22 . A method of detecting cellular viability or damage, collagen denaturation, damage to adnexal structures or adnexal structure necrosis and/or damage to blood vessels of a subject after a wound, preferably a burn comprising:
selecting a subject having a wound, preferably a burn;
imaging a region of the wound, preferably a burn, using the multispectral image system of claim 1 ;
evaluating the image data using a DL algorithm trained with a wound, preferably a burn, database;
displaying whether cells of the wound are viable or damaged, collagen is denatured, adnexal structures are damaged or necrotic and/or blood vessels are damaged within the imaged region of the wound, preferably a burn; and optionally, providing a predictive score for healing of the wound, preferably a burn, over a set time period, preferably 21-30 days, without advanced care such as surgery or skin grafting.
23 . The method of claim 22 , wherein the damaged adnexal structures evaluated comprise hair follicles, sebaceous glands, apocrine glands and/or eccrine sweat glands.
24 . The method of claim 22 , wherein the cell viability or damage, collagen denaturation, damage to adnexal structures or adnexal structure necrosis and/or damage to blood vessels of the subject are evaluated in the papillary region of the skin.
25 . The method of claim 22 , wherein the cell viability or damage, collagen denaturation, damage to adnexal structures or adnexal structure necrosis and/or damage to blood vessels of the subject are evaluated in the reticular dermis of the skin.
26 . The method of claim 22 , wherein the cell viability or damage, collagen denaturation, damage to adnexal structures or adnexal structure necrosis and/or damage to blood vessels of the subject are evaluated deeper than the reticular dermis of the skin.
27 . The method of claim 22 , wherein hyalinzed collagen or lack of detectable individual collagen fibers is detected.
28 . The method of claim 22 , wherein the cellular damage is cell swelling, cytoplasmic vacuolization, or nuclear pyknosis.
29 . The method of claim 22 , wherein when 50% or greater of the adnexual structures analyzed is identified as being damaged or necrotic, a predictive score of non-healing burn is provided and, optionally said subject is provided guidance to receive advanced care such as skin grafting or surgery or said subject is provided skin grafting or surgery.
30 . The method of claim 22 , wherein the DL algorithm was trained using stochastic gradient descent with a momentum optimizer and cross-entropy loss.
31 . The method of claim 22 , wherein the DL algorithm is selected from SegNet, SegNet with filter-bank regularization, SegNet with auxiliary loss, U-Net, Dilated fully connected neural network (dFCN), Averaging Ensemble, TPR-ensemble, or Weighted Averaging Ensemble.
32 . The method of claim 22 , wherein the DL algorithm is SegNet.