Browsing by Author "Filipovic, Nenad (35749660900)"

Heart failure is recognized as a modern epidemic and despite advances in therapy and research, heart failure still carries an ominous prognosis and a significant socioeconomic burden. The main aim of this paper is to demonstrate how novel Decision Support System (DSS) and computational platform like INTELHEART can transform the future of healthcare and early diagnosis of heart failure. The main idea is integration of patient-specific data (i.e. demographic and physical characteristics, medical history, symptoms and signs) and results obtained using existing and novel diagnostic technologies into the cloud environment. Data will be used by different tools for machine learning and computational modelling, developing virtual patient population. Moreover, voice as a biomarker will be collected among participating patients, in order to create a VoiceHeart mobile app. INTELHEART represents a transformative advancement in heart failure care, aiming to make treatment more personalized, and proactive. This initiative centers on precision medicine, using AI-driven analysis and a powerful DSS alongside the cloud-based platform and VoiceHeart mobile app to assist both clinicians and patients. Additionally, it incorporates assessments of psychological resilience and emotional well-being, addressing the oftenoverlooked mental health factors essential to comprehensive heart failure management. © 2024 IEEE.

Heart failure is recognized as a modern epidemic and despite advances in therapy and research, heart failure still carries an ominous prognosis and a significant socioeconomic burden. The main aim of this paper is to demonstrate how novel Decision Support System (DSS) and computational platform like INTELHEART can transform the future of healthcare and early diagnosis of heart failure. The main idea is integration of patient-specific data (i.e. demographic and physical characteristics, medical history, symptoms and signs) and results obtained using existing and novel diagnostic technologies into the cloud environment. Data will be used by different tools for machine learning and computational modelling, developing virtual patient population. Moreover, voice as a biomarker will be collected among participating patients, in order to create a VoiceHeart mobile app. INTELHEART represents a transformative advancement in heart failure care, aiming to make treatment more personalized, and proactive. This initiative centers on precision medicine, using AI-driven analysis and a powerful DSS alongside the cloud-based platform and VoiceHeart mobile app to assist both clinicians and patients. Additionally, it incorporates assessments of psychological resilience and emotional well-being, addressing the oftenoverlooked mental health factors essential to comprehensive heart failure management. © 2024 IEEE.

Objectives. Heart rate variability (HRV) and haemodynamic response to exercise (i.e. peak cardiac power output) are strong predictors of mortality in heart failure. The present study assessed the relationship between measures of HRV and peak cardiac power output. Design. In a prospective observational study of 33 patients (age 54 ± 16 years) with chronic heart failure with reduced left ventricular ejection fraction (29 ± 11%), measures of the HRV (i.e. R-R interval and standard deviation of normal R-R intervals, SDNN) were recorded in a supine position. All patients underwent maximal graded cardiopulmonary exercise testing with non-invasive (inert gas rebreathing) cardiac output assessment. Cardiac power output, expressed in watts, was calculated as the product of cardiac output and mean arterial blood pressure. Results. The mean RR and SDNN were 837 ± 166 and 96 ± 29 ms, peak exercise cardiac power output 2.28 ± 0.85 watts, cardiac output 10.34 ± 3.14 L/min, mean arterial blood pressure 98 ± 14 mmHg, stroke volume 91.43 ± 40.77 mL/beat, and oxygen consumption 19.0 ± 5.6 mL/kg/min. There was a significant but only moderate relationship between the RR interval and peak exercise cardiac power output (r = 0.43, p =.013), cardiac output (r = 0.35, p =.047), and mean arterial blood pressure (r = 0.45, p =.009). The SDNN correlated with peak cardiac power output (r = 0.42, p =.016), mean arterial blood arterial (r = 0.41, p =.019), and stroke volume (r = 0.35, p =.043). Conclusions. Moderate strength of the relationship between measures of HRV and cardiac response to exercise suggests that cardiac autonomic function is not good indicator of overall function and pumping capability of the heart in chronic heart failure. © 2019, © 2019 Informa UK Limited, trading as Taylor & Francis Group.

Salivary gland ultrasonography (SGUS) has proven to be a promising tool for diagnosing various diseases manifesting with abnormalities in salivary glands (SGs), including primary Sjögren's syndrome (pSS). At present, the major obstacle for establishing SUGS as a standardized tool for pSS diagnosis is its low inter/intra observer reliability. The aim of this study was to address this problem by proposing a robust deep learning-based solution for the automated segmentation of SGUS images. For these purposes, four architectures were considered: a fully convolutional neural network, fully convolutional “DenseNets” (FCN-DenseNet) network, U-Net, and LinkNet. During the course of the study, the growing HarmonicSS cohort included 1184 annotated SGUS images. Accordingly, the algorithms were trained using a transfer learning approach. With regard to the intersection-over-union (IoU), the top-performing FCN-DenseNet (IoU = 0.85) network showed a considerable margin above the inter-observer agreement (IoU = 0.76) and slightly above the intra-observer agreement (IoU = 0.84) between clinical experts. Considering its accuracy and speed (24.5 frames per second), it was concluded that the FCN-DenseNet could have wider applications in clinical practice. Further work on the topic will consider the integration of methods for pSS scoring, with the end goal of establishing SGUS as an effective noninvasive pSS diagnostic tool. To aid this progress, we created inference (frozen models) files for the developed models, and made them publicly available. © 2020 Elsevier Ltd

Salivary gland ultrasonography (SGUS) has proven to be a promising tool for diagnosing various diseases manifesting with abnormalities in salivary glands (SGs), including primary Sjögren's syndrome (pSS). At present, the major obstacle for establishing SUGS as a standardized tool for pSS diagnosis is its low inter/intra observer reliability. The aim of this study was to address this problem by proposing a robust deep learning-based solution for the automated segmentation of SGUS images. For these purposes, four architectures were considered: a fully convolutional neural network, fully convolutional “DenseNets” (FCN-DenseNet) network, U-Net, and LinkNet. During the course of the study, the growing HarmonicSS cohort included 1184 annotated SGUS images. Accordingly, the algorithms were trained using a transfer learning approach. With regard to the intersection-over-union (IoU), the top-performing FCN-DenseNet (IoU = 0.85) network showed a considerable margin above the inter-observer agreement (IoU = 0.76) and slightly above the intra-observer agreement (IoU = 0.84) between clinical experts. Considering its accuracy and speed (24.5 frames per second), it was concluded that the FCN-DenseNet could have wider applications in clinical practice. Further work on the topic will consider the integration of methods for pSS scoring, with the end goal of establishing SGUS as an effective noninvasive pSS diagnostic tool. To aid this progress, we created inference (frozen models) files for the developed models, and made them publicly available. © 2020 Elsevier Ltd

The aim of the present study was to investigate the influences of the presence and position of a lower third molar (M3) on the fragility of mandibular angle and condyle, using finite element analysis. From computed tomographic scans of a human mandible with normally erupted M3, two additional virtual models were generated: a mandibular model with partially impacted M3 and a model without M3. Two cases of impact were considered: a frontal and a lateral blow. The results are based on the chromatic analysis of the distributed von Mises and principal stresses, and calculation of their failure indices. In the frontal blow, the angle region showed the highest stress in the case with partially impacted M3, and the condylar region in the case without M3. Compressive stresses were dominant but caused no failure. Tensile stresses were recorded in the retromolar areas, but caused failure only in the case with partially impacted M3. In the lateral blow, the stress concentrated at the point of impact, in the ipsilateral and contralateral angle and condylar regions. The highest stresses were recorded in the case with partially impacted M3. Tensile stresses caused the failure on the ipsilateral side, whereas compressive stresses on the contralateral side. © 2015 European Association for Cranio-Maxillo-Facial Surgery.

The aim of the present study was to investigate the influences of the presence and position of a lower third molar (M3) on the fragility of mandibular angle and condyle, using finite element analysis. From computed tomographic scans of a human mandible with normally erupted M3, two additional virtual models were generated: a mandibular model with partially impacted M3 and a model without M3. Two cases of impact were considered: a frontal and a lateral blow. The results are based on the chromatic analysis of the distributed von Mises and principal stresses, and calculation of their failure indices. In the frontal blow, the angle region showed the highest stress in the case with partially impacted M3, and the condylar region in the case without M3. Compressive stresses were dominant but caused no failure. Tensile stresses were recorded in the retromolar areas, but caused failure only in the case with partially impacted M3. In the lateral blow, the stress concentrated at the point of impact, in the ipsilateral and contralateral angle and condylar regions. The highest stresses were recorded in the case with partially impacted M3. Tensile stresses caused the failure on the ipsilateral side, whereas compressive stresses on the contralateral side. © 2015 European Association for Cranio-Maxillo-Facial Surgery.

Although the concept of the occlusal load transfer through the facial skeleton along the buttresses has been extensively studied, there has been no study to link microarchitecture of the mid-facial bones to the occlusal load distribution. The aim of this study was to analyze micro-structural properties of the mid-facial bones in relation to occlusal stress. The study was performed by combining the three-dimensional finite element analysis (3D FEA) and micro-computed tomography analysis (micro-CT). Clenching was simulated on the computer model of the adult male human skull which was also used as a source of bone specimens. After the FEA was run, stress was measured at the specific sites in cortical shell and trabecular bone of the model along and between the buttresses. From the corresponding sites on the skull, twenty-five cortical and thirteen cancellous bone specimens were harvested. The specimens were classified into high stress or low stress group based on the stress levels measured via the FEA. Micro-architecture of each specimen was assessed by micro-CT. In the high stress group, cortical bone showed a tendency toward greater thickness and density, lower porosity, and greater pore separation. Stress-related differences in microstructure between the groups were more pronounced in trabecular bone, which showed significantly greater bone volume fraction (BV/TV) and trabecular thickness (Tb.Th) in the high stress group. Our results suggest that the mid-facial bones in the adult dentate male skull exhibit regional variations in cortical and trabecular bone micro-architecture that could be a consequence of different occlusal stress. © 2014 Elsevier Inc.

Although the concept of the occlusal load transfer through the facial skeleton along the buttresses has been extensively studied, there has been no study to link microarchitecture of the mid-facial bones to the occlusal load distribution. The aim of this study was to analyze micro-structural properties of the mid-facial bones in relation to occlusal stress. The study was performed by combining the three-dimensional finite element analysis (3D FEA) and micro-computed tomography analysis (micro-CT). Clenching was simulated on the computer model of the adult male human skull which was also used as a source of bone specimens. After the FEA was run, stress was measured at the specific sites in cortical shell and trabecular bone of the model along and between the buttresses. From the corresponding sites on the skull, twenty-five cortical and thirteen cancellous bone specimens were harvested. The specimens were classified into high stress or low stress group based on the stress levels measured via the FEA. Micro-architecture of each specimen was assessed by micro-CT. In the high stress group, cortical bone showed a tendency toward greater thickness and density, lower porosity, and greater pore separation. Stress-related differences in microstructure between the groups were more pronounced in trabecular bone, which showed significantly greater bone volume fraction (BV/TV) and trabecular thickness (Tb.Th) in the high stress group. Our results suggest that the mid-facial bones in the adult dentate male skull exhibit regional variations in cortical and trabecular bone micro-architecture that could be a consequence of different occlusal stress. © 2014 Elsevier Inc.

Colorectal carcinoma is acknowledged as the second leading cause of total cancer-related death in the European Region. The majority of deaths related to colorectal carcinoma are connected with liver metastatic disease. Approximately, in 25% of all patients, liver metastatic disease is diagnosed at the same time as the primary diagnosis, while up to a quarter of others would develop liver metastases in the course of the illness. In this study, we developed reaction-diffusion model and analyzed the effect of drug therapy on liver metastatic disease for a specific patient. Tumor volumes in specific time points were obtained using CT scan images. The nonlinear function for cell proliferation rate as well as data about clinically applied drug therapy was included in the model. Fitting procedure was used for parameter estimation. Good agreement of numerical and experimental results shows the feasibility and efficacy of the proposed system. © 2014 Elsevier Ireland Ltd.

Colorectal carcinoma is acknowledged as the second leading cause of total cancer-related death in the European Region. The majority of deaths related to colorectal carcinoma are connected with liver metastatic disease. Approximately, in 25% of all patients, liver metastatic disease is diagnosed at the same time as the primary diagnosis, while up to a quarter of others would develop liver metastases in the course of the illness. In this study, we developed reaction-diffusion model and analyzed the effect of drug therapy on liver metastatic disease for a specific patient. Tumor volumes in specific time points were obtained using CT scan images. The nonlinear function for cell proliferation rate as well as data about clinically applied drug therapy was included in the model. Fitting procedure was used for parameter estimation. Good agreement of numerical and experimental results shows the feasibility and efficacy of the proposed system. © 2014 Elsevier Ireland Ltd.

Background Both, open and endovascular, procedures are related to higher complication rate in abdominal aortic aneurysm (AAA) with shorter neck. Previous study showed that long-neck AAA might have lower risk of rupture. Estimation of biomechanical forces in AAA improves rupture risk assessment. The aim of this study was to compare morphological features and biomechanical forces in the short- and long-neck AAA with threshold of 15 mm. Methods Digital Imaging and Communication in Medicine images of 64 aneurysms were prospectively collected and analyzed in a case-control study. Using commercially available software, Peak wall Stress (PWS) and Rupture Risk Equivalent Diameter (RRED) were determined. Difference between the maximal aneurysm diameter (MAD) and RRED was calculated and expressed as an absolute and relative (percentage of the MAD) value. In addition, volume of intraluminal thrombus (ILT) was calculated and expressed relative to AAA volume. Results Study included 64 AAA divided in group with long (36, 56.25%), and short (28, 43.75%) neck. There was no correlation between neck length and MAD, PWS, and RRED (P = 0.646, P = 0.421, and P = 0.405, respectively). Relative ILT volume was greater in the short-neck aneurysms (P = 0.033). Relative difference between RRED and MAD was −4% and −14.8% in short- and long-neck aneurysms, respectively (P = 0.029). The difference between RRED and MAD was positive in 14/28 patients (50%) with short neck and in 6/35 patients (17.14%) with long neck (P = 0.011). Conclusions Based on our biomechanical analysis, in AAA with neck longer than 15 mm rupture risk might be lower than the risk estimated by its diameter. It might be explained with lower relative volume of ILT. © 2017 Elsevier Inc.

Understanding of the occlusal load distribution through the mid-facial skeleton in natural dentition is essential because alterations in magnitude and/or direction of occlusal forces may cause remarkable changes in cortical and trabecular bone structure. Previous analyses by strain gauge technique, photoelastic and, more recently, finite element (FE) methods provided no direct evidence for occlusal load distribution through the cortical and trabecular bone compartments individually. Therefore, we developed an improved three-dimensional FE model of the human skull in order to clarify the distribution of occlusal forces through the cortical and trabecular bone during habitual masticatory activities. Particular focus was placed on the load transfer through the anterior and posterior maxilla. The results were presented in von Mises stress (VMS) and the maximum principal stress, and compared to the reported FE and strain gauge data. Our qualitative stress analysis indicates that occlusal forces distribute through the mid-facial skeleton along five vertical and two horizontal buttresses. We demonstrated that cortical bone has a priority in the transfer of occlusal load in the anterior maxilla, whereas both cortical and trabecular bone in the posterior maxilla are equally involved in performing this task. Observed site dependence of the occlusal load distribution may help clinicians in creating strategies for implantology and orthodontic treatments. Additionally, the magnitude of VMS in our model was significantly lower in comparison to previous FE models composed only of cortical bone. This finding suggests that both cortical and trabecular bone should be modeled whenever stress will be quantitatively analyzed. © 2014 Elsevier GmbH.

Understanding of the occlusal load distribution through the mid-facial skeleton in natural dentition is essential because alterations in magnitude and/or direction of occlusal forces may cause remarkable changes in cortical and trabecular bone structure. Previous analyses by strain gauge technique, photoelastic and, more recently, finite element (FE) methods provided no direct evidence for occlusal load distribution through the cortical and trabecular bone compartments individually. Therefore, we developed an improved three-dimensional FE model of the human skull in order to clarify the distribution of occlusal forces through the cortical and trabecular bone during habitual masticatory activities. Particular focus was placed on the load transfer through the anterior and posterior maxilla. The results were presented in von Mises stress (VMS) and the maximum principal stress, and compared to the reported FE and strain gauge data. Our qualitative stress analysis indicates that occlusal forces distribute through the mid-facial skeleton along five vertical and two horizontal buttresses. We demonstrated that cortical bone has a priority in the transfer of occlusal load in the anterior maxilla, whereas both cortical and trabecular bone in the posterior maxilla are equally involved in performing this task. Observed site dependence of the occlusal load distribution may help clinicians in creating strategies for implantology and orthodontic treatments. Additionally, the magnitude of VMS in our model was significantly lower in comparison to previous FE models composed only of cortical bone. This finding suggests that both cortical and trabecular bone should be modeled whenever stress will be quantitatively analyzed. © 2014 Elsevier GmbH.

Salivary gland ultrasonography (SGUS) has shown good potential in the diagnosis of primary Sjögren's syndrome (pSS). However, a series of international studies have reported needs for improvements of the existing pSS scoring procedures in terms of inter/intra observer reliability before being established as standardized diagnostic tools. The present study aims to solve this problem by employing radiomics features and artificial intelligence (AI) algorithms to make the pSS scoring more objective and faster compared to human expert scoring. The assessment of AI algorithms was performed on a two-centric cohort, which included 600 SGUS images (150 patients) annotated using the original SGUS scoring system proposed in 1992 for pSS. For each image, we extracted 907 histogram-based and descriptive statistics features from segmented salivary glands. Optimal feature subsets were found using the genetic algorithm based wrapper approach. Among the considered algorithms (seven classifiers and five regressors), the best preforming was the multilayer perceptron (MLP) classifier (κ = 0.7). The MLP over-performed average score achieved by the clinicians (κ = 0.67) by the considerable margin, whereas its reliability was on the level of human intra-observer variability (κ = 0.71). The presented findings indicate that the continuously increasing HarmonicSS cohort will enable further advancements in AI-based pSS scoring methods by SGUS. In turn, this may establish SGUS as an effective noninvasive pSS diagnostic tool, with the final goal to supplement current diagnostic tests. © 2013 IEEE.

Salivary gland ultrasonography (SGUS) has shown good potential in the diagnosis of primary Sjögren's syndrome (pSS). However, a series of international studies have reported needs for improvements of the existing pSS scoring procedures in terms of inter/intra observer reliability before being established as standardized diagnostic tools. The present study aims to solve this problem by employing radiomics features and artificial intelligence (AI) algorithms to make the pSS scoring more objective and faster compared to human expert scoring. The assessment of AI algorithms was performed on a two-centric cohort, which included 600 SGUS images (150 patients) annotated using the original SGUS scoring system proposed in 1992 for pSS. For each image, we extracted 907 histogram-based and descriptive statistics features from segmented salivary glands. Optimal feature subsets were found using the genetic algorithm based wrapper approach. Among the considered algorithms (seven classifiers and five regressors), the best preforming was the multilayer perceptron (MLP) classifier (κ = 0.7). The MLP over-performed average score achieved by the clinicians (κ = 0.67) by the considerable margin, whereas its reliability was on the level of human intra-observer variability (κ = 0.71). The presented findings indicate that the continuously increasing HarmonicSS cohort will enable further advancements in AI-based pSS scoring methods by SGUS. In turn, this may establish SGUS as an effective noninvasive pSS diagnostic tool, with the final goal to supplement current diagnostic tests. © 2013 IEEE.

Detection of clinically relevant stenosis within coronary arteries as well as planning of treatment (stent implantation) are important topics in clinical cardiology. In this study a thorough methodology for virtual stenting assistance is proposed, that includes the 3D reconstruction of a patient-specific coronary artery from X-ray angiography images, hemodynamic simulations of blood flow, computation of a fractional flow reserve (FFR) equivalent, virtual stenting procedure and an optimization of the virtual stenting, by considering not only the value of computed FFR, but also the low and high WSS regions and the state of arterial wall after stenting. The evaluation of the proposed methodology is performed in two ways: the calculated values of FFR are compared with clinically measured values; and the results obtained for automated optimized virtual stenting are compared with virtual stenting performed manually by an expert clinician for the whole considered dataset. The agreement of the results in almost all cases demonstrates the accuracy of the proposed approach, and the small discrepancies only show the capabilities and benefits this approach can offer. The automated optimized virtual stenting technique can provide information about the most optimal stent position that ensures the maximum achievable FFR, while also considering the distribution of WSS and the state of arterial wall. The proposed methodology and developed software can therefore be used as a noninvasive method for planning of optimal patient-specific treatment strategies before invasive procedures and thus help to improve the clinical outcome of interventions and provide better treatment planning adapted to the particular patient. © 2024 Elsevier Ltd

Detection of clinically relevant stenosis within coronary arteries as well as planning of treatment (stent implantation) are important topics in clinical cardiology. In this study a thorough methodology for virtual stenting assistance is proposed, that includes the 3D reconstruction of a patient-specific coronary artery from X-ray angiography images, hemodynamic simulations of blood flow, computation of a fractional flow reserve (FFR) equivalent, virtual stenting procedure and an optimization of the virtual stenting, by considering not only the value of computed FFR, but also the low and high WSS regions and the state of arterial wall after stenting. The evaluation of the proposed methodology is performed in two ways: the calculated values of FFR are compared with clinically measured values; and the results obtained for automated optimized virtual stenting are compared with virtual stenting performed manually by an expert clinician for the whole considered dataset. The agreement of the results in almost all cases demonstrates the accuracy of the proposed approach, and the small discrepancies only show the capabilities and benefits this approach can offer. The automated optimized virtual stenting technique can provide information about the most optimal stent position that ensures the maximum achievable FFR, while also considering the distribution of WSS and the state of arterial wall. The proposed methodology and developed software can therefore be used as a noninvasive method for planning of optimal patient-specific treatment strategies before invasive procedures and thus help to improve the clinical outcome of interventions and provide better treatment planning adapted to the particular patient. © 2024 Elsevier Ltd

Aim: We sought the cardiopulmonary exercise testing (CPET) parameter that most accurately reflected therapeutic efficacy in patients with hypertrophic cardiomyopathy (HCM). Methods: Well-being questionnaire, N-terminal brain natriuretic peptide measurements, echocardiography, and CPET were performed in patients with symptomatic non-obstructive HCM during phase II, randomized, open-label multicentre study, before and after 16 weeks of traditional or sacubitril/valsartan treatment. Patients were followed 36 months after the initial CPET. Primary endpoints were changes in: 1) peak oxygen consumption (VO2); 2) VO2 at anaerobic threshold (AT); 3) oxygen pulse; 4) minute ventilation (VE)/carbon-dioxide (CO2) production slope; 5) VE/VCO2 at AT (VE/VCO2_AT); 6) VE/VCO2 nadir; 7) VE/VCO2 intercept; and 8) partial end-tidal pressure of carbon-dioxide (PETCO2) change during CPET. Results: Of 115 screened patients, 61 (52 ± 14 years, 43 % women) were included. Within subject therapy effects were detected only by the VE/VCO2 intercept and PETCO2 change, whereas the differences between medical regimens were detected by differences in VE/VCO2 nadir and VE/VCO2_AT changes after the treatment. The best predictors of the change in well-being were left ventricular outflow tract maximal gradient and VE/VCO2 intercept (B = 0.41,0.36; SE = 0.16,0.30; CI = 0.14–0.79, 0.15–1.14; p = 0.006,0.016, respectively). Adverse cardiac events were best predicted by the initial VE/VCO2 nadir. Conclusion: Ventilatory efficiency parameters outperform peak VO2 in gauging therapy effects in patients with HCM. © 2024 Elsevier Inc.

Anatomy of frontal sinuses varies individually, from differences in volume and shape to a rare case when the sinuses are absent. However, there are scarce data related to influence of these variations on impact generated fracture pattern. Therefore, the aim of this study was to analyse the influence of frontal sinus volume on the stress distribution and fracture pattern in the frontal region. The study included four representative Finite Element models of the skull. Reference model was built on the basis of computed tomography scans of a human head with normally developed frontal sinuses. By modifying the reference model, three additional models were generated: a model without sinuses, with hypoplasic, and with hyperplasic sinuses. A 7.7 kN force was applied perpendicularly to the forehead of each model, in order to simulate a frontal impact. The results demonstrated that the distribution of impact stress in frontal region depends on the frontal sinus volume. The anterior sinus wall showed the highest fragility in case with hyperplasic sinuses, whereas posterior wall/inner plate showed more fragility in cases with hypoplasic and undeveloped sinuses. Well-developed frontal sinuses might, through absorption of the impact energy by anterior wall, protect the posterior wall and intracranial contents. © 2017 Pajic, Antic, Vukicevic, Djordjevic, Jovicic, Savic, Saveljic, Janovic, Pesic, Djuric and Filipovic.