Mühendislik Fakültesi / Faculty of Engineeringhttp://hdl.handle.net/11727/14012024-03-29T06:55:52Z2024-03-29T06:55:52ZAutomated Fracture Detection in the Ulna and Radius Using Deep Learning on Upper Extremity RadiographsErdas, Cagatay Berkehttp://hdl.handle.net/11727/119542024-03-27T07:55:43Z2023-01-01T00:00:00ZAutomated Fracture Detection in the Ulna and Radius Using Deep Learning on Upper Extremity Radiographs
Erdas, Cagatay Berke
Objectives: This study aimed to detect single or multiple fractures in the ulna or radius using deep learning techniques fed on upper-extremity radiographs.
Materials and methods: The data set used in the retrospective study consisted of different types of upper extremity radiographs obtained from an open-source dataset, with 4,480 images with fractures and 4,383 images without fractures. All fractures involved the ulna or radius. The proposed method comprises two distinct stages. The initial phase, referred to as preprocessing, involved the removal of radiographic backgrounds, followed by the elimination of nonbone tissue. In the second phase, images consisting only of bone tissue were processed using deep learning models, such as RegNetX006, EfficientNet B0, and InceptionResNetV2. Thus, whether one or more fractures of the ulna or the radius are present was determined. To measure the performance of the proposed method, raw images, images generated by background deletion, and bone tissue removal were classified separately using RegNetX006, EfficientNet B0, and InceptionResNetV2 models. Performance was assessed by accuracy, F1 score, Matthew's correlation coefficient, receiver operating characteristic area under the curve, sensitivity, specificity, and precision using 10-fold cross-validation, which is a widely accepted technique in statistical analysis.
Results: The best classification performance was obtained with the proposed preprocessing and RegNetX006 architecture. The values obtained for various metrics were as follows: accuracy (0.9921), F1 score (0.9918), Matthew's correlation coefficient (0.9842), area under the curve (0.9918), sensitivity (0.9974), specificity (0.9863), and precision (0.9923).
Conclusion: The proposed preprocessing method is able to detect fractures of the ulna and radius by artificial intelligence.
2023-01-01T00:00:00ZSix Sigma Project Prioritization and Selection Using AHP-CODAS Integration: A Case Study in Healthcare IndustryCan, Gulin FeryalToktas, PelinPakdil, Fatmahttp://hdl.handle.net/11727/119202024-03-21T12:45:19Z2023-01-01T00:00:00ZSix Sigma Project Prioritization and Selection Using AHP-CODAS Integration: A Case Study in Healthcare Industry
Can, Gulin Feryal; Toktas, Pelin; Pakdil, Fatma
Given the complex nature of Six Sigma project (SSP) prioritization and selection processes, multicriteria decision-making (MCDM) methods may help organizations identify the most effective projects. Considering potential limitations of subjective methods and advantages of MCDM methods, this article proposes a model that integrates analytical hierarchy process (AHP) and combinative distance-based assessment (CODAS) in SSP prioritization and selection process. In the proposed approach, AHP is employed to assign criteria weights, and CODAS is performed to determine priorities of SSPs. CODAS was advanced in term of its threshold function. Differences between Euclidean distances of two alternatives were compared, based on the standard deviation of Euclidean distances of all alternatives to overcome the subjectivity. This is the first study that combines AHP and CODAS methods for SSP selection, and CODAS is used with objective threshold value computation, and developed for the healthcare industry. In this article, ten SSPs were evaluated for four key criteria groups as financial, operational, patient centric, and organizational main criteria groups. In total, 18 subcriteria were considered under these four main criteria groups. This article provides a support for executives who make implementation plans for the potential SSPs.
2023-01-01T00:00:00ZGoodness-of-fit and Randomness Tests for the Sun's Emissions True Random Number GeneratorTanyer, Suleyman GokhunAtalay, Kumru DidemInam, Sitki Cagdashttp://hdl.handle.net/11727/119022024-03-20T11:47:24Z2014-01-01T00:00:00ZGoodness-of-fit and Randomness Tests for the Sun's Emissions True Random Number Generator
Tanyer, Suleyman Gokhun; Atalay, Kumru Didem; Inam, Sitki Cagdas
Random number generators ( RNGs) are one of the key tools necessary for statistical analysis and optimization methods such as Monte Carlo, particle swarm optimization ( PSO) and the genetic algorithm. Various pseudo and true RNGs are available today, and they provide sufficient randomness. Unfortunately, they generate data that do not always represent the required distribution accurately, especially when the data length is small. This could possibly threaten the 'repeatability' of an academic study. A novel true RNG ( TRNG) using the method of uniform sampling ( MUS) is recently proposed. In this work, the Sun's RF emissions MUS-TRNG is comparatively tested with well known pseudo and true RNGs. It is observed that both randomness and very high goodness-of-fit qualities are possible.
2014-01-01T00:00:00ZTrue Random Number Generation of Very High Goodness-of-fit and Randomness QualitiesTanyer, Suleyman Gokhunhttp://hdl.handle.net/11727/119012024-03-20T11:45:40Z2014-01-01T00:00:00ZTrue Random Number Generation of Very High Goodness-of-fit and Randomness Qualities
Tanyer, Suleyman Gokhun
The statistical nature of numerous problems in mathematics, physics and engineering have led to the development of methods for generating random data for a given distribution. Ancient methods include; dice, coin flipping and shuffling of cards. Today, various pseudo, quasi and true random generators ( RNGs) are being proposed for their improved properties. In this work, test metrics for goodness-of-fit and randomness are reviewed. The method of uniform sampling ( MUS) is modified for improving the randomness without harming the goodness-of-fit qualities. The test results illustrate that very high goodness-of-fit can be obtained even when the number of observed samples is as small as 10.
2014-01-01T00:00:00Z