Feature Ranking by Variational Dropout for Classification Using Thermograms from Diabetic Foot Ulcers.

Diabetes mellitus presents a high prevalence around the world. A common and long-term derived complication is diabetic foot ulcers (DFUs), which have a global prevalence of roughly 6.3%, and a lifetime incidence of up to 34%. Infrared thermograms, covering the entire plantar aspect of both feet, can be employed to monitor the risk of developing a foot ulcer, because diabetic patients exhibit an abnormal pattern that may indicate a foot disorder. In this study, the publicly available INAOE dataset composed of thermogram images of healthy and diabetic subjects was employed to extract relevant features aiming to establish a set of state-of-the-art features that efficiently classify DFU. This database was extended and balanced by fusing it with private local thermograms from healthy volunteers and generating synthetic data via synthetic minority oversampling technique (SMOTE). State-of-the-art features were extracted using two classical approaches, LASSO and random forest, as well as two variational deep learning (DL)-based ones: concrete and variational dropout. Then, the most relevant features were detected and ranked. Subsequently, the extracted features were employed to classify subjects at risk of developing an ulcer using as reference a support vector machine (SVM) classifier with a fixed hyperparameter configuration to evaluate the robustness of the selected features. The new set of features extracted considerably differed from those currently considered state-of-the-art but provided a fair performance. Among the implemented extraction approaches, the variational DL ones, particularly the concrete dropout, performed the best, reporting an F1 score of 90% using the aforementioned SVM classifier. In comparison with features previously considered as the state-of-the-art, approximately 15% better performance was achieved for classification.

Segmentation Approaches for Diabetic Foot Disorders.

Thermography enables non-invasive, accessible, and easily repeated foot temperature measurements for diabetic patients, promoting early detection and regular monitoring protocols, that limit the incidence of disabling conditions associated with diabetic foot disorders. The establishment of this application into standard diabetic care protocols requires to overcome technical issues, particularly the foot sole segmentation. In this work we implemented and evaluated several segmentation approaches which include conventional and Deep Learning methods. Multimodal images, constituted by registered visual-light, infrared and depth images, were acquired for 37 healthy subjects. The segmentation methods explored were based on both visual-light as well as infrared images, and optimization was achieved using the spatial information provided by the depth images. Furthermore, a ground truth was established from the manual segmentation performed by two independent researchers. Overall, the performance level of all the implemented approaches was satisfactory. Although the best performance, in terms of spatial overlap, accuracy, and precision, was found for the Skin and U-Net approaches optimized by the spatial information. However, the robustness of the U-Net approach is preferred.

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

This work presents a revision of four different registration methods for thermal infrared and visible images captured by a camera-based prototype for the remote monitoring of diabetic foot. This prototype uses low cost and off-the-shelf available sensors in thermal infrared and visible spectra. Four different methods (Geometric Optical Translation, Homography, Iterative Closest Point, and Affine transform with Gradient Descent) have been implemented and analyzed for the registration of images obtained from both sensors. All four algorithms' performances were evaluated using the Simultaneous Truth and Performance Level Estimation (STAPLE) together with several overlap benchmarks as the Dice coefficient and the Jaccard index. The performance of the four methods has been analyzed with the subject at a fixed focal plane and also in the vicinity of this plane. The four registration algorithms provide suitable results both at the focal plane as well as outside of it within 50 mm margin. The obtained Dice coefficients are greater than 0.950 in all scenarios, well within the margins required for the application at hand. A discussion of the obtained results under different distances is presented along with an evaluation of its robustness under changing conditions.

Performance Assessment of Low-Cost Thermal Cameras for Medical Applications.

* Correspondence: evilla@iac [...].

Radiation treatment monitoring with DCE-US in CWR22 prostate tumor xenografts.

BACKGROUND: Longitudinal monitoring of potential radiotherapy treatment effects can be determined by dynamic contrast-enhanced ultrasound (DCE-US). PURPOSE: To assess functional parameters by means of DCE-US in a murine subcutaneous model of human prostate cancer, and their relationship to dose deposition and time-frame after treatment. A special focus has been placed to evaluate the vascular heterogeneity of the tumor and on the most suitable data analysis approach that reflects this heterogeneity. MATERIAL AND METHODS: In vivo DCE-US was acquired 24 h and 48 h after radiation treatment with a single dose of 7.5 Gy and 10 Gy, respectively. Tumor vasculature was characterized pixelwise using the Brix pharmacokinetic analysis of the time-intensity curves. RESULTS: Longitudinal changes were detected ( P < 0.001) at 24 h and 48 h after treatment. At 48 h, the eliminating rate constant of the contrast agent from the plasma, kel, was correlated ( P ≤ 0.05) positively with microvessel density (MVD; rτ = 0.7) and negatively with necrosis (rτ = -0.6) for the treated group. Furthermore, Akep, a parameter related to transcapillary transport properties, was also correlated to MVD (rτ = 0.6, P ≤ 0.05). CONCLUSION: DCE-US has been shown to detect vascular changes at a very early stage after radiotherapy, which is a great advantage since DCE-US is non-invasive, available at most hospitals, and is low in cost compared to other techniques used in clinical practice.

Radiation treatment monitoring using multimodal functional imaging: PET/CT ((18)F-Fluoromisonidazole & (18)F-Fluorocholine) and DCE-US.

BACKGROUND: This study aims to assess the effect of radiation treatment on the tumour vasculature and its downstream effects on hypoxia and choline metabolism using a multimodal approach in the murine prostate tumour model CWR22. Functional parameters derived from Positron Emission Tomography (PET)/Computer Tomography (CT) with (18)F-Fluoromisonidazole ((18)F-FMISO) and (18)F-Fluorocholine ((18)F-FCH) as well as Dynamic Contrast-Enhanced Ultrasound (DCE-US) were employed to determine the relationship between metabolic parameters and microvascular parameters that reflect the tumour microenvironment. Immunohistochemical analysis was employed for validation. METHODS: PET/CT and DCE-US were acquired pre- and post-treatment, at day 0 and day 3, respectively. At day 1, radiation treatment was delivered as a single fraction of 10 Gy. Two experimental groups were tested for treatment response with (18)F-FMISO and (18)F-FCH. RESULTS: The maximum Standardized Uptake Values (SUVmax) and the mean SUV (SUVmean) for the (18)F-FMISO group were decreased after treatment, and the SUVmean of the tumour-to-muscle ratio was correlated to microvessel density (MVD) at day 3. The kurtosis of the amplitude of the contrast uptake A was significantly decreased for the control tumours in the (18)F-FCH group. Furthermore, the eliminating rate constant of the contrast agent from the plasma k el derived from DCE-US was negatively correlated to the SUVmean of tumour-to-muscle ratio, necrosis and MVD. CONCLUSIONS: The present study suggests that the multimodal approach using (18)F-FMISO PET/CT and DCE-US seems reliable in the assessment of both microvasculature and necrosis as validated by histology. Thus, it has valuable diagnostic and prognostic potential for early non-invasive evaluation of radiotherapy.

State-of-the-Art Features for Early-Stage Detection of Diabetic Foot Ulcers Based on Thermograms.

Diabetic foot ulcers represent the most frequently recognized and highest risk factor among patients affected by diabetes mellitus. The associated recurrent rate is high, and amputation of the foot or lower limb is often required due to infection. Analysis of infrared thermograms covering the entire plantar aspect of both feet is considered an emerging area of research focused on identifying at an early stage the underlying conditions that sustain skin and tissue damage prior to the onset of superficial wounds. The identification of foot disorders at an early stage using thermography requires establishing a subset of relevant features to reduce decision variability and data misinterpretation and provide a better overall cost-performance for classification. The lack of standardization among thermograms as well as the unbalanced datasets towards diabetic cases hinder the establishment of this suitable subset of features. To date, most studies published are mainly based on the exploitation of the publicly available INAOE dataset, which is composed of thermogram images of healthy and diabetic subjects. However, a recently released dataset, STANDUP, provided data for extending the current state of the art. In this work, an extended and more generalized dataset was employed. A comparison was performed between the more relevant and robust features, previously extracted from the INAOE dataset, with the features extracted from the extended dataset. These features were obtained through state-of-the-art methodologies, including two classical approaches, lasso and random forest, and two variational deep learning-based methods. The extracted features were used as an input to a support vector machine classifier to distinguish between diabetic and healthy subjects. The performance metrics employed confirmed the effectiveness of both the methodology and the state-of-the-art features subsequently extracted. Most importantly, their performance was also demonstrated when considering the generalization achieved through the integration of input datasets. Notably, features associated with the MCA and LPA angiosomes seemed the most relevant.

Low-Cost Pseudo-Anthropomorphic PVA-C and Cellulose Lung Phantom for Ultrasound-Guided Interventions.

A low-cost custom-made pseudo-anthropomorphic lung phantom, offering a model for ultrasound-guided interventions, is presented. The phantom is a rectangular solidstructure fabricated with polyvinyl alcohol cryogel (PVA-C) and cellulose to mimic the healthy parenchyma. The pathologies of interest were embedded as inclusions containing gaseous, liquid, or solid materials. The ribs were 3D-printed using polyethylene terephthalate, and the pleura was made of a bidimensional reticle based on PVA-C. The healthy and pathological tissues were mimicked to display acoustic and echoic properties similar to that of soft tissues. Theflexible fabrication process facilitated the modification of the physical and acoustic properties of the phantom. The phantom's manufacture offers flexibility regarding the number, shape, location, and composition of the inclusions and the insertion of ribs and pleura. In-plane and out-of-plane needle insertions, fine needle aspiration, and core needle biopsy were performed under ultrasound image guidance. The mimicked tissues displayed a resistance and recoil effect typically encountered in a real scenario for a pneumothorax, abscesses, and neoplasms. The presented phantom accurately replicated thoracic tissues (lung, ribs, and pleura) and associated pathologies providing a useful tool for training ultrasound-guided procedures.

Multifrequency Microwave Radiometry for Characterizing the Internal Temperature of Biological Tissues.

The analysis of near-field radiometry is described for characterizing the internal temperature of biological tissues, for which a system based on multifrequency pseudo-correlation-type radiometers is proposed. The approach consists of a new topology with multiple output devices that enables real-time calibration and performance assessment, recalibrating the receiver through simultaneous measurable outputs. Experimental characterization of the prototypes includes a well-defined calibration procedure, which is described and demonstrated, as well as DC conversion from the microwave input power. Regarding performance, high sensitivity is provided in all the bands with noise temperatures around 100 K, reducing the impact of the receiver on the measurements and improving its sensitivity. Calibrated temperature retrievals exhibit outstanding results for several noise sources, for which temperature deviations are lower than 0.1% with regard to the expected temperature. Furthermore, a temperature recovery test for biological tissues, such as a human forearm, provides temperature values on the order of 310 K. In summary, the radiometers design, calibration method and temperature retrieval demonstrated significant results in all bands, validating their use for biomedical applications.

Bimodal microwave and ultrasound phantoms for non-invasive clinical imaging.

A precise and thorough methodology is presented for the design and fabrication of bimodal phantoms to be used in medical microwave and ultrasound applications. Dielectric and acoustic properties of human soft tissues were simultaneously mimicked. The phantoms were fabricated using polyvinyl alcohol cryogel (PVA-C) as gelling agent at a 10% concentration. Sucrose was employed to control the dielectric properties in the microwave spectrum, whereas cellulose was used as acoustic scatterer for ultrasound. For the dielectric properties at microwaves, a mathematical model was extracted to calculate the complex permittivity of the desired mimicked tissues in the frequency range from 500 MHz to 20 GHz. This model, dependent on frequency and sucrose concentration, was in good agreement with the reference Cole-Cole model. Regarding the acoustic properties, the speed of sound and attenuation coefficient were employed for validation. In both cases, the experimental data were consistent with the corresponding theoretical values for soft tissues. The characterization of these PVA-C phantoms demonstrated a significant performance for simultaneous microwave and ultrasound operation. In conclusion, PVA-C has been validated as gelling agent for the fabrication of complex multimodal phantoms that mimic soft tissues providing a unique tool to be used in a range of clinical applications.

Polyvinyl alcohol cryogel phantoms of biological tissues for wideband operation at microwave frequencies.

The aim of this work is to provide a methodology to model the dielectric properties of human tissues based on phantoms prepared with an aqueous solution, in a semi-solid form, by using off-the-shelf components. Polyvinyl alcohol cryogel (PVA-C) has been employed as a novel gelling agent in the fabrication of phantoms for microwave applications in a wide frequency range, from 500 MHz to 20 GHz. Agar-based and deionized water phantoms have also been manufactured for comparison purposes. Mathematical models dependent on frequency and sucrose concentration are proposed to obtain the complex permittivity of the desired mimicked tissues. These models have been validated in the referred bandwidth showing a good agreement to experimental data for different sucrose concentrations. The PVA-C model provides a great performance as compared to agar, increasing the shelf-life of the phantoms and improving their consistency for contact-required devices. In addition, the feasibility of fabricating a multilayer phantom has been demonstrated with a two-layer phantom that exhibits a clear interface between each layer and its properties. Thus, the use of PVA-C extends the option for producing complex multilayer and multimodal phantoms.