Marko Jocić

Towards Automatic Screening of Idiopathic Scoliosis Using Low-Cost Commodity Sensors—Validation Study

This paper presents some preliminary validation study results pertaining to our ongoing attempts to develop a noninvasive scoliosis and other spine disorder automated diagnostic solution implemented using commodity sensors only, thus limiting its overall cost, but still achieving adequate precision. The cost of many such commercial solutions is prohibitive to have them acquired and used by student healthcare institutes in countries such as Serbia, thus we are developing a low-cost one. If proven effective our solutions will be open-sourced.

Managing big data by directed graph node similarity

This paper shows a novel algorithm based on the theory of fuzzy sets for discovering similar nodes in very large directed graphs (millions of nodes with billions of connections), if provided with a sample of representative nodes that are highly affiliated with some feature. This approach is practically verified on Twitter social network case study to discover influential Twitter users in the field of science.

2D fuzzy spatial relations and their applications to DICOM medical images

In this paper we propose a model of fuzzy spatial relations, which can be used to describe imprecise spatial data on 2D images. This model defines relations very similar to those existing in natural language and reasoning, like “on the left side of the image”, “in the center of the image”, “right/left of”, “above/below of”, “(very) close to”, “(very) far from”, “at distance of”, etc. Composition operators and, or, not and subtract are defined in order to combine these basic fuzzy spatial relations into more complex ones. In addition, formula for calculating membership value of imprecise region to fuzzy spatial relation is defined. Proposed model is verified through the example of practical application on DICOM medical images.