Noise cancellation of polycystic ovarian syndrome ultrasound images using robust two-dimensional fractional fourier transform filter and VGG-16 model

The occurrence of Polycystic Ovarian Syndrome (PCOS) among women of reproductive age is increasing worldwide. PCOS is recognized as a critical health issue that can harm physical and mental health of women. This motivates clinicians and radiologists to identify PCOS at initial stages avoiding further complications. Conditions such as hirsutism and gynecomastia are analysed to detect PCOS through ultrasound imaging to obtain position, size and number of follicles. However, noise in ultrasound imaging is a serious problem since it can lead to inaccurate disease diagnosis. It is critical to restore image quality by lowering noise levels in order to extract information from medical images. The proposed work develops a denoising filter by using Two-Dimensional Fractional Fourier Transform (2D-FrFT) to remove artifacts from PCOS ultrasound images in time–frequency plane. The optimal value of a critical parameter of 2D-FrFT—fractional operator (a), used for efficient denoising at different angles, is evaluated by a VGG-16 deep learning model after synthetically corrupting the images with speckle noise of various levels. Experimental results uphold the pre-eminence of suggested method over available denoising methods by attaining Root-Mean-Square Error (RMSE) of 0.2309, Peak Signal-to-Noise Ratio (PSNR) of 72.9689 and Structural Similarity (SSIM) of 0.99. The proposed method filters the high frequency noises from the PCOS Ultrasound dataset to yield better quality images for enhanced segmentation and classification models.