Table 1 Key outcomes from the related works.

Segmentation-focused CNN: AResUNet¹, CR-UNet²², Ocys-Net¹⁷, CNN with Segmentation⁹, Threshold and Watershed^21,39, Hybrid³⁷, GrabCut⁴¹

Filtering, Otsu, contour, watershed

Deep U-Net variants, reverse bottleneck, watershed and contour-based segmentation

Improved follicle, cyst localization and tracking

Deep CNN classifiers: VGG16¹⁰, SqueezeNet²⁶, CNN + FC²⁷, ITL-CNN¹⁹, InceptionV3^7,28, EfficientNetB6³⁸

Ultrasound preprocessing, TL, normalization

CNN-based classification with batch norm, TL, activation optimization

Higher PCOS detection and classification accuracy

Hybrid DL models: VGG16 + Ensemble^8,20, CNN with BiLSTM¹², CNN with KNN¹⁵, CNN with SVM, DT, RF^29,40, Sequential 2D CNN with FS³¹

Feature selection, ensemble integration

CNN fused with ensembles, clustering, sequential learning, wrapper FS

Robust diagnosis across diverse datasets

Wavelet and spectral approaches: PCOS-WaveConvNet⁶, Elman NN with Wavelet³², BPA with Gabor³⁵

Spectral domain, Gabor wavelets

Wavelet-based ConvNets, Elman NNs, Gabor filters

Strong spectral, spatial feature extraction

GAN and augmentation methods: GAN with CNN⁴

Data augmentation

Generative adversarial networks with CNN classifier

Overcame overfitting, improved generalization

Feature-driven and hybrid Methods (ESDPCOS¹³, AMCNN¹⁴, MLOD¹⁶, GIST-MDR⁴², Probabilistic PCOS³³

Ultrasound texture, GLCM, fuzzy logic

Hybrid models combining CNN with GLCM, fuzzy logic, texture scoring

Enhanced robustness in classification

Attention and context learning (ASPPNet + ResNet¹¹, AMCNN¹⁴, EfficientNetB6 with Attention UNet³⁸

Ultrasound

Attention mechanisms, dilated conv, spatial pyramid pooling

Better multiscale learning and context capture

Other imaging modalities: Ovarian quantification², Eye scleral biomarkers³, MRI follicle count³⁶

Ultrasound, scleral, MRI

Quantification of follicles, scleral biomarkers, MRI follicle mapping

Multimodal biomarkers for improved PCOS detection