Fig 1.
Schematic overview of the outlined research in PCOS detection.
Fig 2.
GAN for data augmentation of PCOS and Non PCOS images.
Fig 3.
F-Net architecture for the detection of PCOS.
Fig 4.
YOLOv8 output (a) follicles detected, (b) no follicles detected.
Fig 5.
Follicle Segmentation process in ultrasound images of PCOS a) original Ultrasound image of PCOS, b) noise removed using CLAHE operation, c) cluster 1 represents edges in the cyst, d) cluster 2 shows segmented fluids with some follicles edges, e) cluster 3 indicates segmented follicles, f) cluster 4 depicts the background fluids, g) displays the final segmented follicles.
Fig 6.
Image segmented without PCOS a) original Ultrasound image without follicles, b) noise removed using CLAHE operation, c) cluster 1 represents edges of the image, d) cluster 2 shows region of interest, e) cluster 3 indicates the surrounding region of the image, f) cluster 4 represents the fluid region of the image, g) displays final segmented image.
Fig 7.
Correlation plot between AI based segmentation and ground truth segmentation.
Table 1.
GLCM feature extraction from segmented PCOS and normal images.
Table 2.
Confusion matrix for the three different ML classifiers for PCOS detection for dataset1 (N = 100) and dataset 2 (N = 200).
Table 3.
Performance metrics of various pre-trained and custom model.
Table 4.
Confusion matrix for custom F-Net classifier.
Fig 8.
a. ROC curve for Machine learning and Deep learning classifier for Dataset1. b. ROC curve for Machine learning and Deep learning classifier for Dataset2.
Table 5.
Performance comparison of existing literature related to machine learning and deep learning techniques in PCOS detection.