همایش ملی بیوانفورماتیک ایران

صفحه اصلی / 4th international edition and 13th Iranian Conference on Bioinformatics

Machine Learning for Enhanced Diagnosis of Endometriosis: Challenges and Opportunities

نویسندگان :

Maedeh Darodi¹ Toktam Dehghani²

1- Ferdowsi University of Mashhad 2- Department of Medical Biotechnology and Nanotechnology, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran

کلمات کلیدی :

Endometriosis،Machine Learning،Deep Learning،Artificial Intelligence

چکیده :

Endometriosis is a chronic and complex condition that significantly affects the quality of life for over 190 million women globally. Delayed diagnosis can lead to serious complications such as infertility and chronic pain. This study explores the challenges and opportunities of employing machine learning and deep learning models to enhance the diagnosis of endometriosis and improve healthcare outcomes. Key challenges include the variability and complexity of clinical symptoms, a lack of high-quality data, the necessity for specialized knowledge in algorithm implementation, and the absence of standardized evaluation metrics for comparing models (Ellis et al., 2022). Artificial intelligence can potentially reveal hidden patterns in clinical and imaging data. At the same time, machine learning algorithms can facilitate the development of non-invasive screening tools and generate more accurate predictions of treatment outcomes. These advancements are likely to improve diagnostic accuracy and reduce healthcare costs. The study examines various input data, including clinical information, imaging data (MRI and laparoscopic images), and laboratory results (biochemical markers such as CA125 and VEGF1) (Goldstein & Cohen, 2023). It evaluates various models, including deep learning models like ResNet50 and classical models, including decision trees, random forests, logistic regression, and AdaBoost (Zhang et al., 2023). The findings demonstrate that the AdaBoost model performs best in diagnosing endometriosis, achieving an accuracy of 94% and a sensitivity of 93% (Balica et al., 2023). In comparison, the ResNet50 model achieves an accuracy of 91% and a sensitivity of 82% (Visalaxi & Muthu, 2021) To further enhance research in this field, it is recommended that datasets be expanded to incorporate more diverse patient populations and that models be compared across various conditions and similar contexts. Furthermore, clear guidelines for applying artificial intelligence in diagnosing and treating endometriosis are essential. Despite existing challenges, machine learning and deep learning use in analyzing and predicting endometriosis presents significant potential, necessitating ongoing research to refine model performance and increase confidence in their clinical applications.