Research
A brief overview of research topics and experience gained.
I am interested in the application of computer vision and artificial intelligence algorithms for automation and digitalization of the decision-making processes in Industry and Healthcare.
Industrial engineering
I have developed the Computer vision solution for dimensional inspection of extruded rubber profiles, which is published in IEEE ACCESS (IF2018=4.098).
I have developed a JavaScript mobile app for the management of unsafe conditions and unsafe acts in SMEs, which is published in Safety Science (IF2018=3.69).
I have developed (Python 3, Qt 5, OpenCV 3.4) solution for tracking pallets by using QR codes and IP cameras.
Industry collaboration and consulting: I am working on the development of machine vision-based inspection of leather in the automotive industry.
(In progress) I am working on fusing Computer Vision (Pose estimation) with IoT (force sensors, wearables) for ensuring Workplace safety.
Biomedical engineering
I am developing tools for computer-aided diagnosis of Primary Sjogren's syndrome (pSS) from Salivary gland ultrasonography (SGUS) images. So far, we have developed the radiomics based approach - published in IEEE JBHI (IF2018=4.217). At the moment, we are working on the application of various Deep learning (DL) architectures for semantic segmentation (FCN, FC-ResNet, U-Net, LinkNet) and instance segmentation (Mask R-CNN) - work on DL is done in collaboration with Milos Radovic.
In collaboration with Danko Milasinovic (PI), I have contributed to the development of dfemtoolz (https://github.com/dmilashinovic/dfemtoolz); an open-source C++ library for efficient imposing of materials and boundary conditions in finite element biomedical simulations. The study is published in Computer Physics Communications journal (IF2018=3.309). https://doi.org/10.1016/j.cpc.2019.106996
Period 2014-2019
I have been focused on the topic of image-based modeling in biomedical engineering. It represents multidisciplinary research with aim to: a) make simulation-ready models (including patient-specific geometry, material properties and boundary conditions) of human physiology from its corresponding medical scans; b) analyze obtained simulation's results in order to find relevant conclusions.
Particularly, I have gained experience in the modeling of:
coronary arteries (CA) and coronary stents, led by Prof. Filipović and Prof. Jovičić.
I developed an in-house matlab/c++ code for modeling of CA by fusing X-ray angiography (XRA) and itravascular ultrasound (IVUS) images. Briefly, it performs 3D reconstruction of the catheter path from its two XRA images - and afterwards, it uses the obtained path for positioning of the CA wall contours segmented-extracted from sthe corresponding 2D IVUS images. Since the delivered CA geometry was in the NURBS format, it enables further parametric discretization of the geometry to structured hex8/quad4 mesh elements suitable for the Finite Element Analysis - FEA of blood flow and plaque progression. This study was released in collaboration with Prof. Svetlana Apostolovic.
By extending the procedure described above, I developed an in-house matlab/c++ code for modeling of complex CA trees from its two conventional XRA images-projections. In literature, it is known as a 3D Quantitative Coronary Analysis (3D QCA). Briefly, 3D QCA uses information from 2D CA projections (centerlines and borders are extracted by the traditional 2D QCA) in order to obtain its 3D geometry. Similarly, the geometry of CA lumen is delivered in the NURBS format enabling us further parametric discretization the mesh. The study was done in collaboration with Prof. Nikola Jagic and CISTIB Lab led by Prof. Alejandro Frangi.
I have been also involved in the studies focused on application of numerical techniques for fracture and fatigue analysis of coronary stents. In the initial study, we were focused on analyzing the durability of generic balloon-expandable L-605 Co-Cr stent assuming the hyperphysiological pulsatile pressure conditions for the projected work-life of 10 years (5 × 106 cycles-heartbeats).
I extended the code for extracting CA centerlines from the 2D XRA images for the purpose of extracting 3D CA centerlines from computerized tomography (CT) coronary angiography images. It remains to finish the extraction of CA lumen surface from the CT scans... in progress...
I am interested for incorporating the complex patient-specific conditions (see paragraphs above) in the coronary stents FEA.
We are also working towards numerical estimation of Fractional flow reverse (FFR) by applying the developed procedures (see paragraphs above).
human bones (dentistry and maxillofacial), led by Prof. Filipović and Prof. Jovičić and in collaboration with the Anthropology Lab led by Prof. Marija Djuric.
In our first study we considered biomechanical weakening of devitalized (treated) teeth. After we experimentally determined fracture forces, we used the structural FEA and a Failure Index (based on the maximum principal stress criterion - MPCS) to determine to which extent cavity preparation and each step of dentine removal in the process of root canal treatment (access cavity preparation and root canal enlargement) both individually and jointly contribute to the weakening of the tooth. To reach this, compressive and tensile stresses and corresponding Failure Indices were both calculated and analyzed.
The second study relates to the first one, with aim to estimate the influence of various mastication loads and different tooth treatments (composite restoration and endodontic treatment) on dentine fatigue. The analysis of fatigue behaviour of human dentine in intact and composite restored teeth with root-canal-treatment was performed using FEA and fatigue theory.
We also investigated the influences of the presence and position of a lower third molar (M3) on the fragility of mandibular angle and condyle, using the FEA. From CT 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.
In the next study we developed an improved three-dimensional FEA model of the human skull in order to clarify the distribution of occlusal forces through the cortical and trabecular bone during habitual masticatory activities.
The aim of the recent study was to analyze the influence of frontal sinus volume on the stress distribution and fracture pattern in the frontal region. The study included four representative FEA models of the skull: normally developed frontal sinuses, a model without sinuses, with hypoplasic, and with hyperplasic sinuses. Assuming a 7.7 kN frontal impact, the fracture risk for the developed models was calculated by using the MPCS criterion.
We assessed the biomechanical aspect of horizontal and vertical periodontal bone loss, and the impact of regenerative periodontal therapy by means of FEA. Three patient-specific 3D FE models were developed from the acquired CBCT scans, comprising the patient's upper left canine, first and second premolar, and adjacent alveolar bone. Model 1 represented horizontal bone loss, Model 2 included intrabony defect along distal aspect of tooth #24. Model 3 represented the situation six months after the regenerative periodontal surgery. Displacement, Von Mises and principal stresses were evaluated by means of FEA, following the vertical occlusal load of 150 and 200 N.
To release my tasks, besides using Finite Element Analysis - FEA (structural and fluid flow) i have gained skills to use various algorithms from the field of computational geometry, computer vision, image analysis, optimization, fitting and machine learning (primarily classification). Consequently, i was curious in free time and did some research by applying learned classification/optimization algorithms for developing various medical decision support systems for: predicting outcome of bladder cancer, predicting the outcome of choledocholithiasis surgery (both in collaboration with Prof. Miroslav Stojadinovic) and for efficient estimation of material properties of human cortical bones. Very interesting and useful topic tho.
Please click on links-subpages bellow to get more info (graphical abstracts and key-points) about studies related to specific topics.
