Інститут комп'ютерних наук та інформаційних технологій

Permanent URI for this communityhttps://ena.lpnu.ua/handle/ntb/61741

Browse

Now showing 1 - 2 of 2

Enhancing X-ray Diagnosis through AI: Lung Disease Detection
(Lviv Polytechnic National University, 2023) Khoroshchuk , Dariia; Lviv Polytechnic National University
Master’s degree work of the student of the group CSAI-22 Dariia Khoroshchuk. The topic is “Enhancing X-ray Diagnosis through AI: Lung Disease Detection”. The work aims to obtain a master’s degree in 122 “Computer Science”. The object of the research is the development and optimization of the pneumothorax detection system, considering the bias of the models towards the presence of chest tubes. The subject of the research involves methods for enhancing the processing of chest X-ray scans to achieve precise and rapid pneumothorax detection. The goal of the research is achieved by improving existing methods for detecting pneumothorax in X-ray scans without the presence of chest tubes. These cases are regarded as potentially positive and demand careful consideration by healthcare professionals. Additionally, the research encompasses methods to identify chest tubes on X-rays, allowing the categorization of such scans as pneumothorax negative. The presence of chest tubes on X-ray scans indicates that the patient has already received the necessary medical assistance. As a result of the master’s qualification work, a system was developed to determine the presence of a chest tube, followed by the identification of pneumothorax in cases where a tube is absent. The total volume of work: 72 pages, 34 figures, 39 references.
Program synthesis for genome compression
(Lviv Polytechnic National University, 2023) Maletskyi, Denys; Lviv Polytechnic National University
In the era where genomic data is being generated at an unprecedented pace, the imperative to develop efficient methods for its compression cannot be overstated. This is particularly critical for large-scale genomic projects where the sheer volume of data presents substantial challenges in terms of storage and analysis. This thesis delves into the realm of genome data compression, addressing its significance and exploring innovative solutions to overcome the associated challenges. Central to this discourse is the introduction and exploration of program synthesis as a formidable tool for data compression. Program synthesis, in this context, is leveraged to create sophisticated programs capable of efficiently representing and reproducing genomic data. This technique emerges as a promising approach for genome compression, particularly due to its ability to process large datasets effectively while maintaining data integrity. Throughout this thesis, readers will gain a comprehensive understanding of program synthesis – its mechanics, applications, and how it can be specifically tailored for genome data compression. A significant focus is placed on elucidating how program synthesis can enhance the compression of genome sequences, offering not only a more efficient alternative to traditional methods but also ensuring faster processing times and reduced computational demands. Moreover, the insights and methodologies discussed extend beyond the confines of genomic data. The principles and techniques expounded upon in this thesis have broader applications and can be adapted for compressing various data types. This universality provides a fresh and expansive perspective to the ongoing conversation around data compression strategies, making the findings of this research relevant to a wider audience. In conclusion, this thesis presents a novel approach to genome data compression using program synthesis, specifically the equality saturation approach. The proposed method stands as a testament to the potential of cross-disciplinary 4 innovation in tackling the challenges posed by the ever-growing expanse of genomic data. As the field of genomics continues to evolve, so too must the strategies for managing its data, and this research contributes a pivotal piece to that evolving puzzle. The total volume of work: 61 pages, containing 5 figures and 17 references.