Real-time Burn Classification using Ultrasound Imaging

Abstract

This article presents a real-time approach for classification of burn depth based on B-mode ultrasound imaging. A grey-level co-occurrence matrix (GLCM) computed from the ultrasound images of the tissue is employed to construct the textural feature set and the classification is performed using nonlinear support vector machine and kernel Fisher discriminant analysis. A leave-one-out cross-validation is used for the independent assessment of the classifiers. The model is tested for pair-wise binary classification of four burn conditions in ex vivo porcine skin tissue: (i) 200 °F for 10 s, (ii) 200 °F for 30 s, (iii) 450 °F for 10 s, and (iv) 450 °F for 30 s. The average classification accuracy for pairwise separation is 99% with just over 30 samples in each burn group and the average multiclass classification accuracy is 93%. The results highlight that the ultrasound imaging-based burn classification approach in conjunction with the GLCM texture features provide an accurate assessment of altered tissue characteristics with relatively moderate sample sizes, which is often the case with experimental and clinical datasets. The proposed method is shown to have the potential to assist with the real-time clinical assessment of burn degrees, particularly for discriminating between superficial and deep second degree burns, which is challenging in clinical practice.

A deep learning model for burn depth classification using ultrasound imaging

“…The superposition of scattering acoustic echoes from the regions of contrasting acoustic impedance produces an intricate interference pattern that is manifested as speckles in the B-mode images (White, 2005). Our previous study shows an increase in speckle pattern for the thickness of the skin and subcutaneous tissues with increasing burn depth (Lee et al, 2020).…”

“…The performance of the BurnNet is benchmarked against well-known classifiers for the binary classification of deep partial-thickness and the rest of the burn depths. The benchmark models include (i) linear discriminant analysis (LDA) (Mika et al, 1999), (ii) support vector machine (SVM) with radial basis function (RBF) kernel (Lee et al, 2020), and classifiers based on the stateof-the-art (SOTA) CNN such as (iii) VGG16 (Simonyan and Zisserman, 2014), (iv) ResNet (He et al, 2016), and (v) DenseNet (Huang et al, 2017). The fully connected layers of the SOTA classifiers are replaced by a global average pooling layer concatenated with a fully connected layer activated by a softmax function for classification.…”

“…Second-order statistical measures are calculated from a GLCM and used to characterize the specific texture features of an image. A list of textural features used in this study is provided in our previous work (Lee et al 2020). Our previous study showed that the GLCM based textural features could differentiate between four burn depths with an average accuracy of 93% (Lee et al, 2020).…”

“…However, these studies rely on subjective evaluation of burn depths based on A-or B-mode signals and are often limited in accuracy (Ye and De, 2017). In recent work, we have shown that the engineered textural features obtained from B-mode images of ex vivo porcine skin can classify burn depths with an average accuracy of 93% (Lee et al, 2020). However, our preliminary study indicates that this approach yields an average accuracy of 88% in classifying burn depths in freshly euthanized pig in situ skin samples.…”

Identification and Verification of Five Potential Biomarkers Related to Skin and Thermal Injury Using Weighted Gene Co-Expression Network Analysis

“…Clinical diagnosis by visual and tactile examination remains the current standard for determining the depth of a patient’s burn injury. This method has a serious lag in determining the patient’s condition and does not provide timely information regarding the patient’s progress due to the rapid progression of burn injury ( Lee et al, 2020 ). Therefore, the development of an effective molecular diagnostic burn technique is necessary to improve early burn care, reduce complications, and decrease treatment-associated costs.…”