An integrated adaptive bilateral filter-based framework and attention residual U-net for detecting polycystic ovary syndrome

Introduction: PCOS is a global hormonal disorder in women, increasing the risk of certain cancers. Early diagnosis and treatment can manage it. While metabolic markers are typical for detection, MRI and ultrasound physiological features are gaining interest. Ultrasound is especially promising for early-stage detection. AI and ML are enhancing this detection process. Methodology: The methodology focuses on developing a hybrid model for detecting PCOS, an early indicator of gynecological cancer in women. This model incorporates adaptive bilateral filter-based image pre-processing and a novel attention residual u-net (AResUNet). The adaptive bilateral filtering allows for efficient noise reduction, while AResUNet's architecture ensures adaptability across both 2D and multi-modal images. Findings: The findings of the study demonstrate a significant improvement in PCOS detection through the proposed hybrid model. It has been observed that the AResUNet model performs exceptionally well on both 2D and multi-modal images. Conclusion: The 2% improvement over current models reflects the innovation and potential applicability of this approach, providing a promising direction for further research and clinical practice in managing PCOS and other related health concerns.