Quantitative interpretation of infrared images of lower limbs in individuals with and without type 2 diabetes mellitus

Abstract

AbstractBackgroundThe quantitative interpretation of the radiometric information extracted from infrared (IR) images in individuals with and without type 2 diabetes mellitus (DM2) is an open problem yet to be solved. This is of particular value given that DM2 is a worldwide health problem and onset for evolution toward diabetic foot disease (DFD). Since DM2 causes changes at the vascular and neurological levels, the metabolic heat distribution on the outer skin is modified as a consequence of such alterations. Of particular interest in this contribution are those alterations displayed over the skin's heat patterns at the lower limbs. At the core of such alterations is the deterioration of the vascular and neurological networks responsible for procuring systemic thermoregulation. It is within this context that IR imaging is introduced as a likely aiding tool to assist with the clinical diagnosis of DM2 at stages early enough to prevent the evolution of the DFD.MethodsIR images of lower limbs are acquired from a cohort of individuals clinically diagnosed with and without DM2. Additional inclusion criteria for patients are to be free from any visible wound or tissue‐related trauma (e.g., injuries, edema, and so forth), and also free from non‐metabolic comorbidities. All images and data are equally processed and analyzed using indices that evaluate the spatial and temporal evolution of temperature distribution in lower limbs. We studied the temporal response of individuals’ legs after inducing an external stimulus. For this purpose, we combine the information of the asymmetry and thermal response index (ATR) and the thermal response index (TRI), computed using images at different times, improving the results previously obtained individually with ATR and TRI.ResultsA novel representation of the information extracted from IR images of the lower limbs in individuals with and without DM2 is presented. This representation was built using the ATR and TRI indices for the anterior and posterior views (PVs), individually and combining the information from both views. In all cases, the information of each index and each view presents linearity properties that allow said information to be interpreted quantitatively in a well‐defined and limited space. This representation, built in a polar coordinate space, allows obtaining sensitivity values of 86%, 97%, and 97%, and specificity values of 83%, 72%, and 78% for the anterior view (AV), the PV, and the combined views, respectively. Additionally, it was observed that the angular variable that defines this new representation space allows to significantly (p < 0.01) differentiate the groups, while correlating with clinical variables of interest, such as glucose and glycated hemoglobin.ConclusionThe linearity properties that exist between the ATR and TRI indices allow a quantitative interpretation of the information extracted from IR images of the lower extremities of individuals with and without DM2, and allow the construction of a representation space that eliminates possible ambiguities in the interpretation, while simplifying it, making it accessible for clinical use.