Deep Learning Classification for Diabetic Foot Thermograms

Abstract

According to the World Health Organization (WHO), Diabetes Mellitus (DM) is one of the most prevalent diseases in the world. It is also associated with a high mortality index. Diabetic foot is one of its main complications, and it comprises the development of plantar ulcers that could result in an amputation. Several works report that thermography is useful to detect changes in the plantar temperature, which could give rise to a higher risk of ulceration. However, the plantar temperature distribution does not follow a particular pattern in diabetic patients, thereby making it difficult to measure the changes. Thus, there is an interest in improving the success of the analysis and classification methods that help to detect abnormal changes in the plantar temperature. All this leads to the use of computer-aided systems, such as those involved in artificial intelligence (AI), which operate with highly complex data structures. This paper compares machine learning-based techniques with Deep Learning (DL) structures. We tested common structures in the mode of transfer learning, including AlexNet and GoogleNet. Moreover, we designed a new DL-structure, which is trained from scratch and is able to reach higher values in terms of accuracy and other quality measures. The main goal of this work is to analyze the use of AI and DL for the classification of diabetic foot thermograms, highlighting their advantages and limitations. To the best of our knowledge, this is the first proposal of DL networks applied to the classification of diabetic foot thermograms. The experiments are conducted over thermograms of DM and control groups. After that, a multi-level classification is performed based on a previously reported thermal change index. The high accuracy obtained shows the usefulness of AI and DL as auxiliary tools to aid during the medical diagnosis.

Deep Learning for Diabetes: A Systematic Review

“…Their results show an excellent localization performance for the quantification and the potential to be adopted in clinical settings. In [104], a customized CNN was designed to detect plantar ulcers on the thermography of diabetic foot with a publicly accessible dataset [114]. Moreover, Yamada et al investigated the incidence of cardiovascular disease among three anti-diabetic drugs, using a DMLP model that achieved better results than conventional LR analysis [103].…”

“…Moreover, Yamada et al investigated the incidence of cardiovascular disease among three anti-diabetic drugs, using a DMLP model that achieved better results than conventional LR analysis [103]. In summary, it is noted that most of the studies focused on microvascular complications, including DR [91]- [101], diabetic foot [104], and diabetic neuropathy [105], while there is only one study focusing on macrovascular complications (cardiovascular diseases) [103].…”

“…The unsupervised pre-training in DBNs is calculated layer by layer [47] to find good initial weights for the discriminative fine-tuning, which is used by [59]. Alternatively, data augmentation is a way to improve model performance with limited data, which can be found in [98], [104]. These studies used a series of image manipulation, such as shifting, rotating, and flipping, to transform the available images and expand the training sets.…”

Assessment of Registration Methods for Thermal Infrared and Visible Images for Diabetic Foot Monitoring

“…Liu et al [ 30 ] apply a rigid transformation in order to correct the distortions introduced by the subject distance but an analysis on the performance degradation with distance together with other transformation methods is not carried out. Other studies focus only on the thermal images, and do not perform registration between thermal and visible spectrum images, limiting the registration to contralateral thermal images comparisons, thus not confronting the challenges of using heterogeneous sensors and its matching [ 31 , 32 ]. To the best of the authors’ knowledge, a consistent evaluation of different registration methods in order to assess the best choice for the diabetic foot monitoring application has not yet been carried out.…”

Predictive Models for the Characterization of Internal Defects in Additive Materials from Active Thermography Sequences Supported by Machine Learning Methods

“…Within the machine learning (ML) approaches, regression learners study the relationship between one or more explanatory random variables and their responses [ 15 ]. Specifically, the artificial neural network (ANN) has been applied for regressions in various investigations with thermography to estimate the depth of the defects [ 4 , 16 ] or for biomedical applications [ 17 ]. ANN can be applied using visualization approaches that provide information about its behavior and structure [ 18 ].…”