Publication
Diabetic foot monitoring using dynamic thermography and AI classifiers
| dc.contributor.author | Ricardo Vardasca, PhD, ASIS, FRPS | |
| dc.date.accessioned | 2021-12-29T14:26:40Z | |
| dc.date.available | 2021-12-29T14:26:40Z | |
| dc.date.issued | 2019 | |
| dc.description.abstract | Diabetes Mellitus (DM) is one of the most growing burdens in healthcare, one of its impacting consequences is Diabetic Foot Ulcers (DFU), which will affect at least 1 in each 4 DM patients in their lifetime. If not identified early, DFU can become chronic and in more severe cases lead to amputations affecting seriously the quality of life of patients and increase the healthcare costs. Infrared thermal (IRT) imaging has been used as a research method to early identification of DFU, since an elevation of skin temperature is a sign of inflammation, and a reduction a sign of poor vascularization. There are two main types of DFU: neuro-ischemic and ischemic. A database with dynamics IRT plantar foot examination images of 39 active DFU patients was built, the images were analyzed through measuring mean temperature of regions of interest (ROI), which correspond to most frequent documented locations of DFU. Statistics showed that there was no evidence of significant differences between thermal asymmetry values and thermal recovering differences in all ROI, apart from the one located at the medial forefoot. The ROIs were assessed in both feet and the value of thermal asymmetry was taken in consideration per each ROI. Using the database with the analysis results, a decision support system was built implementing machine learning algorithms such as: Artificial Neural Networks (ANN), Support Vector Machines (SVM) and k-Nearest Neighbour (k-NN), to classify the data and assess the correct identification of the type of DFU. The best overall result achieved (Table 1) was with k-NN of 5 neighbors with 81.25% accuracy, 80% specificity and 100% sensitivity. These results are promising for DFU early identification and expected to improve with a larger sample. | pt_PT |
| dc.description.version | info:eu-repo/semantics/publishedVersion | pt_PT |
| dc.identifier.doi | 10.21611/qirt.2019.027 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10400.26/38507 | |
| dc.language.iso | eng | pt_PT |
| dc.title | Diabetic foot monitoring using dynamic thermography and AI classifiers | pt_PT |
| dc.type | conference object | |
| dspace.entity.type | Publication | |
| person.familyName | Vardasca | |
| person.givenName | Ricardo | |
| person.identifier | R-001-FFR | |
| person.identifier.ciencia-id | 9F17-FD5F-E767 | |
| person.identifier.orcid | 0000-0003-4217-2882 | |
| person.identifier.rid | J-4948-2013 | |
| person.identifier.scopus-author-id | 24491279800 | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | conferenceObject | pt_PT |
| relation.isAuthorOfPublication | 33602b11-6c79-40f9-a768-d7c792bc2d57 | |
| relation.isAuthorOfPublication.latestForDiscovery | 33602b11-6c79-40f9-a768-d7c792bc2d57 |
Files
Original bundle
1 - 1 of 1