2019 № 1 Automated system of bacterioscopic diagnosis of tuberculosis
The article describes the scheme of work and requirements for software and hardware complex for automated bacterioscopic diagnosis of tuberculosis. The basic functionality of the hardware of such an automated system and the required capabilities provided by its software are listed. The stages of automated analysis of digital microscopic images of sputum stained by the method of Ziehl-Nielsen are presented. Own algorithms and mathematical models which can be included in such hardware software complex are presented.
2018 № itm Means of intellectual data analysis and support of decision-making in diagnostics and treatment of drug-dependent
Early detection of the drug used by the patient is essential In the diagnosis and treatment of drug addicts. There are specific symptoms of drug use, according to which it is determined that the patient used before the laboratory tests. The use of methods of data mining allows you to identify the characteristic signs of using several drugs and establish previously unexplored symptoms for new drugs, identify typical and atypical patients. In the work, the patterns between the narcotic drugs used and the symptoms are mathematically described using associative rules. Algorithms Apriori, Close and the MClose algorithm proposed by the authors are used to find these rules. The MClose algorithm finds the most significant strict associative rules (rules with reliability 1). The article presents a proposal on expert pre-pro- cessing of melon, which allows to significantly reduce the number of generated associative rules and improve the quality of their interpretation. The developed methods and means is aimed at diagnosing and supporting decision-making in the treatment of drug addicts.
2018 № 2 Reduction of the features space when processing multiple drug resistance of mycobacteria in patients with pulmonary tuberculosis
Early detection of the presence of multiple drug resistance of mycobacteria to essential antituberculosis drugs is relevant in the diagnosis and treatment of pulmonary tuberculosis. Mathematical methods and information technologies can help solving this medical problem by excluding those not informative features from the set of features (indicators of the patient’s health status). The Kulbak method is used for assessment of informative features of the multiple drug resistance. The selection of features is made by the sorted (by informativeness) list of features through evaluating the quality of classification performed by ROC analysis. The performed researches showed that 6 features selected from the suggested method (out of 26 considered) allow to select patients with high probability of not having multiple drug resistance, which creates conditions for their adequate treatment.