Browsing by Author "Fareed, Syeda Uneeza"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Diagnosing cancer using the fractal analysis method
    (Bursa Uludağ Üniversitesi, 2023-10-30) Fareed, Syeda Uneeza; Kırcı, Pınar; Bursa Uludağ Üniversitesi/Fen Bilimleri Enstitüsü/Bilgisayar Mühendisliği Anabilim Dalı.
    Cancer is a prevalent and potentially life-threatening disease that necessitates accurate and timely diagnosis. Recently, there has been increasing interest in leveraging ultrasound diagnostic technology to identify and classify benign and malignant nodules. This advancement is particularly significant as it can help spare patients with benign nodules from undergoing unnecessary and invasive needle biopsy procedures, reducing patient discomfort and healthcare costs. In the light of these considerations, this study aimed to investigate the effectiveness of fractal analysis in differentiating between benign and malignant thyroid nodules during ultrasonography. Fractal analysis holds promise as a potential tool to enhance diagnostic accuracy by assessing the complex structural patterns present within the thyroid nodules. The fractal analysis box-counting technique was performed on normal, benign and malignant ultrasound images of the Thyroid Digital Image Database and Breast Ultrasound Image Dataset. The Mann-Whitney U test was performed to check if the fractal dimensions belonging to malignant and benign categories differed significantly. When comparing the malignant and benign datasets using the Mann-Whitney U test, a result of "significantly different" (p<0.0032) was found. In addition, features including fractal dimension and logarithm (box counts) were used to train machine learning models like KNN, ANN and DT to extend the results. For better comparison, Fourier transform was applied to the ultrasound images and features were used to train machine learning models. The accuracy rates of the models trained on fractal features and fourier features were comparable. The findings indicate that the fractal dimension serves as a valuable characteristic for distinguishing between different types of thyroid and breast nodules. Moreover, studying the fractal properties of the B-mode ultrasound images can offer a dependable reference for identifying and categorizing cancerous tissues during the process of ultrasound diagnosis.