Segmentation of ovarian cyst in ultrasound images using AdaResU-net with optimization algorithm and deep learning model.

This section outlines the results achieved from segmenting, preprocessing, and classifying ovarian cysts using ultrasound images in Python. The dataset was sourced from https://dev.cancerimagingarchive.net/ . Innovative techniques and algorithms were implemented to achieve high accuracy in both segmentation and classification tasks. Performance metrics such as sensitivity, specificity, accuracy, precision, and F1 scores were calculated to evaluate the outcomes. The segmentation accuracy was assessed using DLC and WCE methods. The dataset, consisting of 700 images, was split into 80% for training and 20% for testing purposes. The proposed technique and algorithm demonstrated superior accuracy compared to existing methods. Guided trilateral filter (GTF): The GTF is designed to reduce noise while maintaining important structural information within the image. It operates by considering the spatial closeness, intensity similarity, and edge information. Standard deviation parameter: When pre-processing an image using different filtering algorithms, the standard deviation (σ) parameter plays a crucial role in determining the extent of smoothing. This parameter controls the weight assigned to the intensity differences, effectively influencing the degree of noise reduction and edge preservation. Example: Gaussian filter: For instance, when using a Gaussian filter, the standard deviation parameter σ determines the width of the Gaussian kernel. A larger σ results in more blurring, whereas a smaller σ maintains more detail. Here is an example of how the standard deviation parameter can be specified: Gaussian Filter with σ = 1.5: This value provides moderate smoothing, suitable for reducing noise while preserving fine details. Figures  4 and 5 illustrate the ovarian cyst image and elucidate the pre-processing and segmentation methods utilized. The image has undergone pre-processing to eliminate noise and enhance visualization using GTF. The image becomes clearer after undergoing preprocessing, in contrast to the original image. Subsequently, segmentation is carried out to accurately identify the cyst within the pre-processed image. This is accomplished by segmenting the desired cyst based on pixel values in the image. Figure 4 Pre-processing. Figure 5 Segmentation. Pre-processing. Segmentation. Accuracy: the level of the number of records grouped accurately versus the all-out records displayed in the situation beneath: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Accuracy = \frac{TP + TN}{{TP + FP + TN + FN}}$$\end{document} A c c u r a c y = T P + T N T P + F P + T N + F N Sensitivity otherwise recall: It estimates the number of positive cases that a model can foresee accurately among all sure cases in the dataset. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Sensitivity\frac{TP}{{TP + FN}}$$\end{document} S e n s i t i v i t y TP T P + F N Precision: It estimates the quantity of positive expectations that are recognized accurately. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Pr ecision = \frac{TP}{{TP + FP}}$$\end{document} Pr e c i s i o n = TP T P + F P F1 score: The F1 score lets us know the consonant mean on accuracy and review. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F1\,score = \frac{2xprecisionxrecall}{{\Pr ecision + recall}}$$\end{document} F 1 s c o r e = 2 x p r e c i s i o n x r e c a l l Pr e c i s i o n + r e c a l l Table 1 presents the current results achieved in segmenting cysts based on their size, which typically ranges from 5 to 10 mm. The study collected cysts of various sizes, each categorized by type. Many existing techniques struggle to accurately segment cysts based on size due to complex network structures. The proposed model utilizes segmentation methods to effectively identify standard cyst sizes and evaluates its performance by comparing it with other established techniques. Table 1 Segmentation results based on size. Size of cyst (mm) SegNet PSPNet (pyramid scene parsing network) DLNN U-Net Proposed 4 67.87 55.6 72.34 80.32 Benign 5 68.7 58.67 72.36 82.45 Benign 2.5 72.45 60.78 73.56 84.65 Benign 6 70.6 65.7 74.78 85.67 Malignant 8 70.8 68.9 76.65 86.56 Malignant 7.5 78.6 68.91 76.66 87.67 Malignant 9 67.56 63.34 79.87 88.76 Malignant Segmentation results based on size. Table 2 offers a detailed analysis of the ground truth data, focusing specifically on segmenting cyst regions based on pixel values. Using ultrasound images of the ovary as input, the study segmented these images into various regions: the overall ovary region, cyst region, and the specific side of the ovary where the cyst is located. The cyst region is further subdivided based on specific sectors, considering whether the cyst is situated on the right or left side of the ovary. Table 2 Ground truth table for cyst area segmentation. Input image Segmented image Ovary area in pixel Cyst area Ratio (%) Cyst area region wise (pixel) Left Right 512 500 1.024 342 0 2586 498 5.192 1711 0 968 470 2.059 658 0 898 528 1.700 0 762 1067 654 1.63 0 973 845 824 1.025 0 721 2341 959 2.802 0 1453 1054 652 1.616 0 934 890 750 1.186 756 653 2564 850 3.016 1856 1845 Ground truth table for cyst area segmentation. Table 3 presents the dataset utilized for both training and testing purposes. Any deep learning model that wants to function effectively must have access to an appropriate dataset. The model is trained by a dataset consisting of 700 ultrasound images, encompassing both benign and malignant cases. This dataset has two categories separated: the training set400 images and, the testing set 300 images. Table 4 compares the segmentation accuracy (%) of five different deep-learning models across varying iterations. U-Net starts with 75.23% accuracy at 15 iterations, while DeepLabv3 + shows consistent improvement, reaching 95.27% at 25 iterations. Mask R-CNN peaks at 98.47% accuracy by 35 iterations. FCN achieves high accuracy early and stabilizes, ending at 99.08% by 60 iterations. The proposed method, AdaResU-net, consistently improves and achieves the highest accuracy of 99.98% after 80 iterations, demonstrating its effectiveness in ovarian cyst segmentation. Table 3 Ovarian cyst detection using images present in the dataset. Dataset Category Image count Training Benign 200 Malignant 200 Testing Benign 150 Malignant 150 Overall dataset – 700 Table 4 Segmentation accuracy analysis based on iterations. Iterations U-Net Deep Labv3 +  Mask R-CNN FCN (fully convolutional network) Proposed 15 75.23 78.29 65.20 53.20 94.20 20 70.43 89.86 87.62 88.62 95.82 25 62.63 67.70 95.27 85.27 96.27 30 65.67 53.66 69.83 80.83 97.83 35 72.93 36.01 84.47 65.47 98.47 40 51.20 60.96 65.71 65.81 87.81 45 59.62 66.87 56.61 78.61 98.65 60 71.74 68.89 72.08 83.08 99.08 80 87.51 82.28 56.98 77.98 99.98 100 74.18 79.69 53.47 85.47 99.47 Ovarian cyst detection using images present in the dataset. Segmentation accuracy analysis based on iterations. The proposed network achieves 99% accuracy in cyst segmentation and accurately determines cyst sizes, outperforming other methods. It demonstrates superior detection accuracy through iterative improvements. Each convolutional layer acts as a filter during training, identifying specific image features before passing them to the next layer. Table 5 compares cyst segmentation results between the proposed and existing techniques, showing better performance by the proposed network. Table 6 details the hyperparameters used to fine-tune AdaResU-Net with the WHO optimizer. Batch size indicates the number of training instances processed in each network update, while the learning rate controls the magnitude of weight adjustments during training. Table 5 Segmented cyst image with proposed and existing technique. Input image Watershed IAKmeans-RA Adaptive thresholding Technique Proposed Table 6 Hyperparameter tuning for AdaResU-net model. Hyperparameter Values Batch size 128 Learning rate 0.001 Epoch 100 Optimizer WHO Activation function ReLu Segmented cyst image with proposed and existing technique. Hyperparameter tuning for AdaResU-net model. Figure  6 illustrates the objective function DLC. When epochs ranging from 0 to 100 are represented by the x-axis, and DLC values ranging from 0 to 0.75 are represented by the y-axis. In this case, they have utilized DLC to segment the cyst image from the ultrasound ovarian image to easily diagnose the problem. The techniques compared include CNN, DLNN, SVM, and the proposed techniques. The proposed technique achieved the highest DLC among the existing techniques, reaching its maximum value at epoch 100. The WCEL function is utilized to isolate the picture of a cyst from the surroundings of the overall image. The x-axis represents the epoch’s value, which spans from 0 to 100, while the y-axis represents the WCEL function values, ranging from 0 to 100. Figure 6 DLC and WCE. DLC and WCE. Figure  7 explains the prediction analysis conducted on the testing dataset. The confusion matrix in this analysis includes both correct and incorrect classifications. There are four possibilities: True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN). The rows represent the predicted values, while the columns represent the actual values. The cells on the diagonal represent correctly classified cases, while the remaining cells represent incorrectly classified cases. The different types of cysts are classified namely Endometrioid, corpus leuteum, Dermoid, Haemorrhagic, and Mucinous Cystadenoma. The accuracy of detecting cysts is determined by the results of this model. Figure 7 Confusion matrix. Confusion matrix. Figure  8 illustrates the loss curve for both the training and testing data. In this case, they have conducted 8 iterations (epochs) until achieving the minimum loss. The loss values range from 0.1 to 0.5. By repeatedly iterating with the ovarian image dataset, we were able to reduce the training loss. Using a segmented ovarian cyst image, the proposed network calculated an accuracy and loss curve. The curve gradually decreases from top to bottom indicates during training data the loss is reduced. Figure  9 illustrates the accuracy graph for both the training and testing data. The proposed algorithm significantly enhanced the training accuracy by repeating the iterations in the hidden layer network. The maximum accuracy was attained at the 8th iteration. However, the accuracy of the test data was comparatively lower. It contains the ovarian cyst image either cancerous or non-cancerous. From the above two graphs, they have observed that the accuracy is increased gradually by training the data, and loss is reduced. Figure 8 Loss curve. Figure 9 Accuracy curve. Loss curve. Accuracy curve. The effectiveness of the suggested model has been observed in recent ovarian cyst detection systems, including U-Net, DeepLabV3 +, mask R-CNN, and FCN classifier when compared to the proposed PDC network. The results shown in Fig.  10 , demonstrate that the proposed model outperforms other methods in terms of accuracy. The proposed model achieves an enhanced accuracy rate of 98.1% compared to ML (97.2%), CNN (96.6%), DLNN (95.7%), and SVM (95.2%). Therefore, in comparison to current ovarian cyst detection techniques, the proposed PDC network exhibits superior performance in cyst detection. Figure 10 Cyst detection compared with existing and proposed approach. Cyst detection compared with existing and proposed approach. Figure  11 illustrates the convergence curves of the proposed WHO algorithm alongside existing firefly and butterfly optimization methods. The WHO algorithm demonstrates superior convergence efficiency, achieving a faster rate of convergence and more stable performance compared to both firefly and butterfly algorithms. This is evidenced by its consistently lower convergence time and smoother curve trajectory throughout the optimization process. These results underscore the effectiveness of WHO in optimizing [specific application or problem], offering significant improvements in efficiency and reliability over established optimization techniques. Figure 11 Convergence curve. Convergence curve. Figure  12 depicts the fitness improvement achieved by the WHO algorithm over iterations. The plot demonstrates a steady decrease in fitness function values, indicating effective optimization progress. WHO consistently outperforms existing firefly and butterfly algorithms, showcasing its robust capability in enhancing solution quality. These findings highlight WHO's efficacy in achieving superior fitness outcomes across optimization iterations. Figure 12 Fitness improvement for WHO. Fitness improvement for WHO. In Table 7 , the proposed adaptive deep learning-based segmentation technique achieves a segmentation accuracy of 98.87% when applied to ovarian ultrasound cyst images. This outperforms existing datasets such as OASIS (95.02%), OVARI (92%), OC400 (96.89%), and SCIROCCO (95.67%). The higher accuracy of 98.87% underscores the effectiveness of the proposed approach in improving diagnostic precision and patient care outcomes compared to established datasets, highlighting its potential for enhancing medical imaging applications in ovarian cyst diagnosis. Table 7 Comparison with the existing dataset. Dataset Values (%) Proposed 98.87 OASIS dataset 95.02 OVARI dataset 92 OC400 (ovarian cyst 400 dataset) 96.89 SCIROCCO ( screening cysts in robotic colonoscopy dataset) 95.67 Comparison with the existing dataset. Table 8 compares the experimental findings of different approaches for ovarian cyst segmentation and classification. The proposed method utilizes a Guided Trilateral Filter (GTF) for noise reduction in pre-processing, an AdaResU-net architecture for segmentation, and a Pyramidal Dilated Convolutional (PDC) network for classification, achieving a segmentation accuracy of 98.87%. In contrast, existing approaches such as those employing standard noise reduction methods and the U-net architecture with SVM or traditional methods achieve accuracies of 95.02% and 96.89%, respectively. The Gaussian smoothing approach with ResNet and Decision Trees achieves a segmentation accuracy of 96.89%. This comparison highlights the superior performance of the proposed method in accurately segmenting ovarian cysts and classifying benign/malignant types, demonstrating its potential for enhancing diagnostic accuracy and patient care outcomes compared to traditional and other advanced techniques. Table 8 Experimental comparison. Pre-processing method Segmentation technique Classification method Segmentation accuracy (%) GTF (proposed) AdaResU-net Pyramidal dilated convolution 98.87 Standard noise reduction methods U-net SVM or traditional methods 95.02 Gaussian smoothing ResNet Decision trees 96.89 Experimental comparison.

The female ovaries can develop cysts, which are sacs that contain fluid. These cysts can be identified at an early stage through the use of ultrasound imaging. This technique involves employing adaptive deep-learning methods and an optimization algorithm to classify ovarian cysts. The initial step involves pre-processing the images by applying a guided trilateral filter (GTF) to eliminate any noise present in the input image. Subsequently, the cysts are segmented based on their size. By utilizing an Adaptive Convolutional Neural Network (AdaResU-Net), they can predict whether the cysts are benign or malignant. To achieve the optimal accuracy of AdaResU-Net, the Wild Horse Optimizer (WHO) is employed to fine-tune hyperparameters such as the learning rate, batch size, and epoch count. The optimization algorithm addresses two metrics, namely Dice Loss Coefficient (DLC) and weighted Cross-Entropy (WCE), to evaluate the segmentation output without any loss. This approach has successfully classified different types of cysts with an impressive accuracy rate of 98.87%. Figure  1 explains the overall proposed diagram of ovarian cyst segmentation. Figure 1 Proposed methodology. Proposed methodology. Advanced AI Techniques: The use of an Adaptive Convolutional Neural Network (AdaResU-net) represents a sophisticated application of deep learning specifically tailored for medical image segmentation. This network architecture adapts to the complexity and variability of ovarian cyst images, enhancing segmentation accuracy. Optimization with WHO Algorithm: The integration of the Wild Horse Optimization (WHO) algorithm for hyperparameter tuning is novel in the context of medical image analysis. This optimization technique helps to fine-tune the segmentation model, potentially improving its robustness and performance. Clinical Relevance: By focusing on ovarian cyst detection and classification (benign vs. malignant), the method addresses a critical need in women’s health. Early detection through accurate segmentation can significantly impact patient outcomes by enabling timely medical interventions. The novelty of this work lies in its integration of advanced artificial intelligence techniques, specifically tailored for early disease detection through deep learning-based segmentation algorithms. As of the current stage (2024), the field has evolved significantly with the introduction of adaptive deep learning architectures like AdaResU-net, which can dynamically adjust to the complexities and variability of medical images such as ultrasound scans of ovarian cysts. This adaptability enhances accuracy in detecting and classifying diseases at early stages, surpassing traditional methods that may struggle with image noise and variability. The use of innovative optimization techniques like the Wild Horse Optimization (WHO) algorithm further enhances the precision of these algorithms, marking a significant advancement in medical imaging and diagnostic capabilities. Background and detailed explanation Ovarian cysts, fluid-filled sacs within the ovaries, often develop asymptomatically but can lead to serious health complications such as ovarian torsion, infertility, and ovarian cancer. Early detection and accurate characterization are crucial for timely treatment and preventing adverse outcomes. Ultrasonography is the primary imaging modality due to its non-invasiveness, real-time capability, and lack of ionizing radiation. However, interpreting ultrasound images of ovarian cysts presents challenges like weak contrast, speckle noise, and hazy boundaries. To address these, this study proposes an advanced deep learning-based segmentation technique. It employs a Guided Trilateral Filter (GTF) in pre-processing to reduce noise while preserving edge information for clearer images. The Adaptive Convolutional Neural Network (AdaResU-net) adapts to the variability in cyst images, accurately segmenting and classifying cysts as benign or malignant using learned features. The Wild Horse Optimization (WHO) Algorithm optimizes hyperparameters like Dice Loss Coefficient and Weighted Cross-Entropy to maximize segmentation accuracy across diverse cyst types. Furthermore, a Pyramidal Dilated Convolutional (PDC) Network enhances diagnostic utility by classifying ovarian cyst types, thus improving clinical decision-making beyond segmentation alone. They picked input pictures to apply a pre-processing technique that enhances their quality by removing noise through GBF. This technique prepares them for further processing. Let \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Psi$$\end{document} Ψ measured as the entire picture space \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M_{p} (x)$$\end{document} M p ( x ) is the confined spatial region centered at pixel region \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x(x \in \psi )$$\end{document} x ( x ∈ ψ ) with sizes \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$[ - p,p, - p,p]$$\end{document} [ - p , p , - p , p ] . \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$jand\,\,\,\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{j}$$\end{document} j a n d j ⌢ D describes a noisy picture also it’s a guessed separated picture 30 . Then, at that point, the likelihood of the leftover \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$a_{t} = \overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{J} \,(y) - J(y + t)$$\end{document} a t = J ⌢ ( y ) - J ( y + t ) , among \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{J} \,(x)\,\,and\,\,J(X + t)$$\end{document} J ⌢ ( x ) a n d J ( X + t ) every local pixel \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t(t \in M_{p} (x)$$\end{document} t ( t ∈ M p ( x ) . 1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(a_{t} ) \prec \left( {\frac{1}{s}e^{{\theta \left[ {\frac{{a_{t} }}{s}} \right]^{2} }} } \right)^{{q_{t} }}$$\end{document} P ( a t ) ≺ 1 s e θ a t s 2 q t where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S$$\end{document} S denotes scale boundary, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varphi (.)$$\end{document} φ ( . ) is a differentiable kernel capability with monotonically expanding assets, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b_{t}$$\end{document} b t is a sign of the closeness of the noiseless powers in pixels area \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x\,\,\,\,and\,x + t$$\end{document} x a n d x + t . 2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{J}_{GTF}^{L + 1} = \frac{{\sum\limits_{{t \in \in M_{r} (y)}} {h_{t} \delta_{t} \varphi } \left( {\left[ {\frac{{\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{J}^{L} (y) - J(Y + 1)}}{s}} \right]^{2} } \right)J(y + t)}}{{\sum\limits_{{t \in \in M_{r} (y)}} {h_{t} \delta_{t} \varphi } \left( {\left[ {\frac{{\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{J}^{L} (y) - J(Y + 1)}}{s}} \right]^{2} } \right)}}$$\end{document} J ⌢ GTF L + 1 = ∑ t ∈ ∈ M r ( y ) h t δ t φ J ⌢ L ( y ) - J ( Y + 1 ) s 2 J ( y + t ) ∑ t ∈ ∈ M r ( y ) h t δ t φ J ⌢ L ( y ) - J ( Y + 1 ) s 2 where weight \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta_{t}$$\end{document} δ t represents yields high qualities as well as advances smoothing. It has a low rate when the shape is available, therefore, protecting the boundaries. c is the iteration integer, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$J^{(0)} (y) = J(y)$$\end{document} J ( 0 ) ( y ) = J ( y ) at the initialization \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L = 0$$\end{document} L = 0 . \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$h_{t}$$\end{document} h t Stand for the spatial space weight. The primary goal of the segmentation process is to precisely separate the cyst from the background image. The proposed method categorizes cysts based on their sizes and classifies them as benign or malignant using AdaResU-Net. The network's hyperparameters, such as batch size, learning rate, and epoch count, were optimized by WHO through iterative algorithm enhancements. AdaResU-Net's expected design is shown in Fig.  2 , which is dependent on the ResU-Net design. Like U-net, AdaResU-Net comprises a downsampling pass to the left and an upsampling pass to the right. Yet, the essential components of AdaResU-Net are the remaining knowledge systems, each consisting of three padded convolutional layers. The leftover building blocks within the reduction pass were utilized by the max-pooling activity of step 2, which logically reduces the size of the component map. This comprehensive approach employs upsampling, convolutional layers toward gradually expanding the dimensions of the element map up until it gets to the initial information dimension. By employing upsampling and downsampling techniques, the quality and level of detail in the shared image can be adjusted, fostering a strong residual association between blocks of similar dimensions. The final convolutional layer in the tissue possesses a suitable 1 × 1 dimensional channel and is activated using a sigmoid function 31 . Figure 2 AdaResU-net architecture. AdaResU-net architecture. The encoder, also known as the contracting path, plays a crucial role in extracting relevant spatial and feature information from the input data. Convolutional Layers: Convolutional layers are often used as the initial component in the encoder. These layers utilize filters to identify features such as edges, textures, and patterns in the input data. Each convolutional layer builds upon the information obtained from the previous layer, allowing the model to learn progressively more complex attributes. Activation Function (ReLU): After applying each convolutional layer, an activation function, typically ReLU (Rectified Linear Unit), is used. ReLU introduces non-linearity into the network, enabling it to understand complex data relationships 32 . Pooling Layers: Pooling layers, often max-pooling layers, typically reduce the spatial dimensions of the feature maps while preserving essential information. These layers help to reduce computational requirements and decrease the risk of overfitting. Feature compression: As encoders evolve, the number of channels (depth) increases as the characteristic map's spatial dimension shrinks. This feature compression allows the network to collect hierarchical and abstract input data representation. The decoder, also known as the expansion path, works with the encoder to reconstruct the segmented output. The decoder gradually increases the spatial decision of the characteristic map. The process operates through the following steps: Transposed Convolutional Layers (Up-convolution): The decoder uses transposed convolution layers known as up-convolution or deconvolution to increase its spatial resolution and upscale feature map. These layers recover the original input image. Skip Connections: The incorporation of skip connections plays a pivotal role in U-Net and several other CNN architectures. Skip connections establish connections between layers in the encoder and decoder. By doing so, the decoder gains access to the high-resolution feature maps of the encoder, which serve as a crucial spatial data source. Skip connections are indispensable in preserving intricate details and ensuring accurate segmentation. Output Layer: The final step involves generating the segmentation mask, typically performed by the decoder. Depending on the task, this layer may use either a SoftMax or sigmoid activation function to produce a probability for each pixel, indicating its association with different classes or segments. 3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Soft\max (b_{i} ) = \frac{{e^{bi} }}{{\sum\limits_{j = 1}^{n} {e^{bi} } }}$$\end{document} S o f t max ( b i ) = e bi ∑ j = 1 n e bi where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b_{i}$$\end{document} b i received neuron output, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document} i output level \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n$$\end{document} n output number layer. Horses in social organizations are separated into two categories: non-protective and protective. The public grouping, connection and grazing, mating performance, headship hierarchy, and control of these two sorts of organizations are different. However, the focus of the WHO approach is on non-territorial groups, which consist of group leaders (stallions), various mares, and their offspring, including young horses. Stallions are positioned near mares for contact, and mating can transpire whenever. Foals typically begin grazing within the initial week of existence and increase their grazing while reducing relaxes while they grow grown-up. Before reaching puberty 33 , Male horses join solitary groups, while foals go from their parent groups to reach maturity for mating. In non-territorial horses that roam freely, the dominant mare usually assumes the role of the family group's leader, with the remainder of the team following in a descending hierarchy of power, and typically, the stallion is positioned a little distance behind the assembly. In the study, they optimize many problems and lay out a WHO via group behaviors, grazing, mating, dominance, and leadership. Figure  3 explains the proposed flowchart of AdaResU-net with WHO. Figure 3 AdaResU-net-WHO. AdaResU-net-WHO. Population Initialization: First, they divide this initial population into several groups. It \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M$$\end{document} M is the number of members of the population, the number of groups is 4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g = \left[ {M \times QR} \right]$$\end{document} g = M × Q R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$QR$$\end{document} QR is the proportion of stallions in all the people. Behavior of grazing: To incorporate grazing behavior, the approach designates the stallion as the focal point of the pasture, while the group members explore the surroundings in search of food. To simulate this behavior, they introduced Eq. ( 1 ). Using Eq. ( 1 ), the group members are encouraged to explore the vicinity of the leader, each within a distinct radius. 5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{Y}^{i}_{j,g} = 2Q\cos (2\prod RQ) \times (stallion^{i} - Y^{j}_{j,g} ) + stallion^{j}$$\end{document} Y ⌢ j , g i = 2 Q cos ( 2 ∏ R Q ) × ( s t a l l i o n i - Y j , g j ) + s t a l l i o n j \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{Y}^{i}_{j,g}$$\end{document} Y ⌢ j , g i is the group member's present position (foal if mare otherwise). \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$stallion^{j}$$\end{document} s t a l l i o n j is the stallion's (group head's) position. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q$$\end{document} Q is a computed adaptive mechanism using Eq. ( 3 ). \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R$$\end{document} R is an even random number within the range \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$[ - 2,2]$$\end{document} [ - 2 , 2 ] . The pi value of 3.14 is equivalent to the horses grazing at various angles (360°) relative to the group leader, 6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{gathered} V = \vec{R}_{1} < TDR \hfill \\ IDX = (p = = 0) \hfill \\ Q = R_{2} \otimes IDX + R_{3} \otimes ( \approx IDX) \hfill \\ \end{gathered}$$\end{document} V = R → 1 < T D R I D X = ( p = = 0 ) Q = R 2 ⊗ I D X + R 3 ⊗ ( ≈ I D X ) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V$$\end{document} V is a vector with the problem's lengths between 0 and 1., \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\vec{R}_{1} and\,\,\mathop{R}\limits^{\rightharpoonup} _{3}$$\end{document} R → 1 a n d R 3 ⇀ are uniformly distributed random vectors in the range \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$[0,1].$$\end{document} [ 0 , 1 ] . \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R_{2}$$\end{document} R 2 is a uniformly distributed random digit in the range [0,1]. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$IDX$$\end{document} IDX random vector index vector \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\vec{R}_{1}$$\end{document} R → 1 returns that assure the condition \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p = = 0$$\end{document} p = = 0 . \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TDR$$\end{document} TDR is an adaptive parameter that has a starting value of and decreases following Eq. ( 4 ) as the algorithm is executed, concluding the algorithm's execution 0. 7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TDR = 1 - Current\,iter \times (\frac{1}{\max iteration})$$\end{document} T D R = 1 - C u r r e n t i t e r × ( 1 max i t e r a t i o n ) Horse mating behavior: Horses exhibit a unique behavior compared to other animals, involving the separation of young horses from the herd to facilitate companionship. Before reaching puberty, foals leave the main group; male foals join a separate solitary horse group, while female foals join a second family group to mature and find a mate. This division is essential to prevent mating between fathers and siblings or offspring. To implement this behavior, a specific procedure is followed: a foal from one group leaves and joins a temporary group, and a foal from another group also joins this temporary group. When these two foals reach adolescence, they can mate if they are unrelated and of different genders. The resulting offspring then leave the temporary group and join another group. All different horse groups undergo this cycle of leaving, mating, and reproducing again. To simulate this departure and mating behavior, Eq. ( 4 ) has been proposed, which is equivalent to the Crossover operator of the mean type. 8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y^{{p_{{}} }}_{g,k} = Crossover(Y^{q}_{g,i} ,Y^{Q}_{g,j} )\,\,\,\,\,\,\,\,\,i \ne j \ne k,p = q = end,\,$$\end{document} Y g , k p = C r o s s o v e r ( Y g , i q , Y g , j Q ) i ≠ j ≠ k , p = q = e n d , 9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Crossover = mean$$\end{document} C r o s s o v e r = m e a n \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y^{{p_{{}} }}_{g,k}$$\end{document} Y g , k p is the location of the horse \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p$$\end{document} p within the group, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k$$\end{document} k when it leaves the group, it is replaced by a horse whose breeder(s) also left the group and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$j\,\,and\,i$$\end{document} j a n d i has reached maturity. These horses lack any familial ties, yet they have successfully mated and given birth to offspring. The foal's position \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$q$$\end{document} q is determined by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document} i its departure from the group and its subsequent mating with the horse \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document} i in position \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y^{z}_{g,j}$$\end{document} Y g , j z within the group, which then leaves the group \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$j$$\end{document} j . The group leader is responsible for guiding the group to a suitable location, defined as the water hole. The group must move toward this water hole, and other groups also head in the same direction. The leaders compete to control this water hole, with only the dominant group allowed to use it while other groups are prohibited until the dominant group vacates. The group leaders must guide their groups to the water hole and use it if they dominate. However, if another group gains dominance, they must retreat. To achieve this, it is suggested to use Eq. ( 5 ) to estimate the gap and contact. 10 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$stallion_{g1} = \left\{ \begin{gathered} 2Q\cos (2\prod RQ) \times (WH - Stallion_{g1} ) + WH\,\,\,if\,R_{3} > 0.5 \hfill \\ 2Q\cos (2\prod RQ) \times (WH - Stallion_{g1} ) + WH\,\,\,if\,R_{3} \le 0.5 \hfill \\ \end{gathered} \right\}$$\end{document} s t a l l i o n g 1 = 2 Q cos ( 2 ∏ R Q ) × ( W H - S t a l l i o n g 1 ) + W H i f R 3 > 0.5 2 Q cos ( 2 ∏ R Q ) × ( W H - S t a l l i o n g 1 ) + W H i f R 3 ≤ 0.5 where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$stallion_{g1}$$\end{document} s t a l l i o n g 1 the leader of the next position \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document} i , \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$WH$$\end{document} WH represents the water hole position, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$stallion_{g1}$$\end{document} s t a l l i o n g 1 denotes the present place of the leader of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document} i the group, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q$$\end{document} Q and represents an adaptive mechanism by Eq. ( 6 ). \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R$$\end{document} R is a uniform arbitrary integer in the range \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$- 2,2$$\end{document} - 2 , 2 . Exchange and selection leaders: Initially, leaders are chosen randomly to maintain the algorithm's inherent randomness. With the algorithm's advancement, the selection of leaders is based on their fitness. If a group member exhibits superior fitness compared to the present leader, both the role of the leader and the corresponding member's position will be altered following Eq. ( 11 ). 11 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Stallion_{g1} = \left\{ \begin{gathered} Y_{gj} \,\,\,if\,\cos t(Y_{gj} ) \cos \,t\,(stallion_{g1} ) \hfill \\ \hfill \\ \end{gathered} \right\}$$\end{document} S t a l l i o n g 1 = Y gj i f cos t ( Y gj ) cos t ( s t a l l i o n g 1 ) The cyst in the ovarian images has been successfully fragmented through the implementation of the AdaResU-net design. The organization is finely tuned by the WHO algorithm through the acquisition of the optimal configuration. Additionally, the hyperparameters of AdaResU-net are optimized by solving the objective function DLC and WCE. The hyperparameters of AdaResU-net are improved by addressing the beneath objective function 12 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$DLC = \frac{2TP}{{2TP + FN + FP}}$$\end{document} D L C = 2 T P 2 T P + F N + F P 13 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\log G(r,s) = \sum\limits_{i = 1}^{n} {\omega_{i} p(y_{i} )\log_{e} (q(y_{i} )}$$\end{document} log G ( r , s ) = ∑ i = 1 n ω i p ( y i ) log e ( q ( y i ) where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p(y_{i} )$$\end{document} p ( y i ) addresses the genuine conveyance of the sample, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$q(y_{i} )$$\end{document} q ( y i ) addresses the circulation anticipated by the sample, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\omega_{i}$$\end{document} ω i and denotes the weight loss function. The classification procedure employs the Pyramidal Dilated Convolutional (PDC) network to classify cysts into types such as Endometrioid cyst, mucinous cystadenoma, follicular, dermoid, corpus luteum, and hemorrhagic cyst. This network uses a reduced feature set to enhance the accuracy of input images and generate improved images with optimal features. Dilated convolution is a technique that modifies the standard convolution kernel by introducing gaps to increase its receptive field. The dilation rate, which determines the size of these gaps, must be manually specified. For a 3 × 3 convolution kernel, different dilation rates affect the receptive field size. When the dilation rate is set to 1, the dilated convolution kernel is identical to the original basic convolutional layer. However, increasing the dilation rate to 2 expands the 3 × 3 convolution kernel to resemble a 5 × 5 convolution kernel 34 . During this dilation process, the gaps are filled with zeros, except at the central position as depicted in Fig.  6 (b). This results in a receptive field of 5 × 5 for the convolution operation. By using dilated convolution, the receptive field is increased without changing the integer of parameters or using pooling. This guarantees that the amount of the output feature map remains the same, while also incorporating multi-scale information in the convolutional output. In Fig.  7 the symbols Dilate1 − Dilate3 represent the dilated convolution kernels, while Conv1 − Conv4 represent the common convolution kernels. The symbol {∙} represents concatenated algorithms, and F1 − F3 shows the various speeds at which the outputs from dilated convolution are produced. Based on these representations, the output Y can be expressed using the following formula, assuming X as the input: 14 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F1 = X \otimes Conv1 \otimes Dilate1$$\end{document} F 1 = X ⊗ C o n v 1 ⊗ D i l a t e 1 15 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F2 = X \otimes Conv2 \otimes Dilate2$$\end{document} F 2 = X ⊗ C o n v 2 ⊗ D i l a t e 2 16 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F3 = X \otimes Conv3 \otimes Dilate3$$\end{document} F 3 = X ⊗ C o n v 3 ⊗ D i l a t e 3 17 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y = (F1,F2,F3) \otimes Conv4 + X$$\end{document} Y = ( F 1 , F 2 , F 3 ) ⊗ C o n v 4 + X The PDC structure utilizes dilated convolution by varying dilation rates to expand the receiving area devoid of the need for pooling. Moreover, the pyramid arrangement effectively combines information from diverse receptive fields, thereby enhancing the network's performance. This dilated pyramid module emulates the functioning of the human eye, which amalgamates features at different scales when observing an object. Similarly, the component for pyramid dilated convolution merges the output from distinct dilated convolutional blocks with different degrees of dilation, mimicking the human eye's process to some extent.

Study introduces an innovative approach to ovarian cyst segmentation and classification using deep learning techniques. By applying a Guided Trilateral Filter for noise reduction and leveraging AdaResU-net for segmentation along with a Pyramidal Dilated Convolutional network for classification, they achieved a segmentation accuracy of 98.87%. This surpasses existing methods and promises enhanced diagnostic accuracy for ovarian cysts, addressing challenges such as weak contrast and speckle noise in ultrasound images. optimized approach, guided by the Wild Horse Optimization algorithm and objective functions like Dice Loss Coefficient and Weighted Cross-Entropy, demonstrates significant potential to improve patient care outcomes through more precise diagnosis and treatment planning. Segmentation is a crucial task whereas cyst detection according to their size is a major role. For resolving this research analysis proposed DL model is discussed for accurate detection. The proposed model used the traditional AdaResU-net deep neural learning form a 128-layer neural network trained on an ovarian cyst dataset image. This model is used for segmenting the cyst and predicted as benign or malignant. To enhance the network's performance, the WHO algorithm is used to fine-tune the hyperparameters with their training procedure. Fine-tuning is done to attain great precision when compared with existing techniques. The proposed classification technique classifies the cyst types as Endometrioid, corpus leuteum, Dermoid, Haemorrhagic, and Mucinous Cystadenoma. To assess the proposed model performance measures such as Sensitivity, Specificity, Precision, F1 score, and Accuracy statistical measures are used, given averages of 98.5%, 99.6%, 98.4%, 98.8%, and 99.9%. The proposed model obtained various statistical values for various classified cysts. In the future analysis, the effective DL will be used with a more significant number of datasets, due to the limited size of the dataset utilized in this instance. For effective segmentation, DL with hybrid optimization will be used for training a greater number of images will get more accuracy. AdaResU-net's adaptive architecture and parameter optimization increase computational demands during training and inference. GTF's noise reduction in preprocessing involves computationally intensive operations, impacting processing time and resource allocation. AdaResU-net's adaptive architecture and parameter optimization increase computational demands during training and inference. GTF's noise reduction in preprocessing involves computationally intensive operations, impacting processing time and resource allocation. The solution's dependency on extensive and diverse training data limits generalization to varied ultrasound image qualities and clinical scenarios. High computational requirements for WHO algorithm optimization and PDC network classification may restrict real-time deployment in resource-constrained healthcare settings. The solution's dependency on extensive and diverse training data limits generalization to varied ultrasound image qualities and clinical scenarios. High computational requirements for WHO algorithm optimization and PDC network classification may restrict real-time deployment in resource-constrained healthcare settings.

A new Deep Learning (DL) model is presented in this research study that incorporates hyperparameter tuning to segment ovarian cyst images. Through simulation analysis, they have demonstrated that the proposed DL learning framework, known as AdaResU-Net, effectively adapts to ovarian datasets. AdaResU-Net achieves a remarkable level of segmentation accuracy and spatial definition on ovarian image sets, surpassing the performance of both comparing U-Net and ResU-Net based on the average dice coefficient. On the other hand, U-Net and ResU-Net exhibit more complex operations and yield significantly lower mean Dice coefficients when applied to the ovarian dataset. These varying results highlight the insufficiency of solely tuning the learning rate and dropout for adapting architecture to specific datasets. However, by carefully selecting a set of hyperparameters for learning framework, they have successfully achieved optimal results. To accomplish this, they introduce the WHO algorithm, which tunes the network's hyperparameters to obtain the best possible segmentation accuracy. Furthermore, presented AdaResU-Net demonstrates superior adaptability and performance compared to U-Net in the segmentation of both benign and malignant cases. Considering the successful application of U-Net in natural image segmentation, they believe that AdaResU-Net can also be utilized in non-medical segmentation tasks while offering more compact architectures. Following segmentation, they perform classification to categorize the different types of ovaries. proposed technique significantly improves segmentation accuracy by 5–10% compared to existing studies.

The research paper titled "Utilizing Watershed Division and Shape Examination for Ovarian Cysts on Ultrasound Pictures" was proposed by Nabilah et al. 20 . Upon receiving an ultrasound picture at the medical clinic, it underwent a preprocessing process as part of the system to eliminate noise in the image. Subsequently, the segmentation process was carried out using the watershed strategy. The results of the segmentation were then utilized for highlight extraction, where the presence of tumors and papillary structures, as well as their sizes, were identified by the bounding box technique in contour analysis. The extracted information was used for cyst classification. This framework demonstrates a precision rate of 97.8%. In the year 2023, Begam et al. 21 presented a novel approach to automatically classify thecyst category in digital ultrasonography pictures. These approaches employ preprocessing and segmentation techniques to acquire essential Regions of Interest (ROI) as well as Feature Extraction to take out the required feature vectors. The Convolutional Neural Networks (CNN) classification method is utilized to detect abnormalities and identify various ovarian cyst types, including Dermoid cysts, Hemorrhagic cysts, and Endometrioma cysts. The diagnostic tool that is automated aims to minimize costs and shorten the diagnosis period, enabling prompt and accurate treatment. In the year 2023, Fan J et al. 22 implemented a different approach to bottleneck planning and employed worldwide data to improve the capability of extracting components. Additionally, they incorporated the Efficient Channel Attention (ECA) component toward detecting local cross-channel communication, thereby giving ample attention to important data features and compensating for any drawbacks resulting from channel aspect reduction. Being a lightweight organization approach considers the model's effective learning performance, assessed using the dataset for ovarian blisters, achieving a high level of accuracy. The classification accuracy of this approach is 95.93%, showcasing its significant potential in the field of medical research and application. In 2023, a method was proposed by Sheikdavood et al. 23 to identify Polycystic Ovary Syndrome (PCOS) using a series of steps including pre-processing, segmentation, feature selection, and classification. The initial step involved removing any noise spots from the images and enhancing them for further processing. The researchers utilized an improved version of the K-means algorithm called IAKmeans-RSA, which incorporated the Reptile Search Algorithm, for growth division and follicle recognition. To extract features from fragmented images, the Convolutional Neural Network (CNN), a deep learning algorithm, was working. The data was then classified using the Deep Neural Network (DNN) approach. In 2023, a model was developed by Suganya et al. 24 to determine the location and arrangement of ovarian blisters utilizing a Deep Learning Neural Network (DLNN). Initially, the image quality was enhanced through pre-processing techniques such as Hu moments, Haralick features, and various histograms. The proposed DLNN method employed the Inception model for feature extraction to evaluate different types of masses. Ultimately, the detection of ovarian cancer was carried out using the Extreme Gradient Boosting (XGBoost) classifier. The purpose of the experiments was to evaluate the DLNN model's presentation. According to the results, the DLNN form and the XGBoost classifier were able to attain the highest finding of 98%. In 2023, Sheela et al., 25 developed a diagnostic model aimed at identifying the early stage of ovarian cysts and minimizing unnecessary biopsy procedures and patient discomfort. The model consisted of three main steps: enhancing the quality of transvaginal 2D B-mode ultrasound pictures through pre-processing, segmenting Region of Interest (ROI) to isolate the ovarian cyst, and extracting textural features using Local Binary Pattern (LBP) analysis. To accurately diagnose and classify ovarian cysts as either benign or malignant, a Support Vector Machine (SVM) was working after feature extraction. The SVM performance exhibited significant improvement, achieving an average accuracy of 92%. Consequently, the classification of ovarian cysts in the transvaginal 2D B-mode ultrasound pictures was accomplished based on the extracted textural characteristics obtained using the initial dark value-based LBP. In the year 2022, Priya et al., 26 introduced a technique for identifying follicles and establishing the order of ovaries. This method relies on estimating dimensions and performing calculations for each follicle. The research paper encompasses four distinct stages: preprocessing, object detection utilizing a specialized division algorithm, partitioning employing an enhanced watershed algorithm, and feature extraction through the utilization of diverse mathematical and measurable components like size, count, mean, standard deviation, and more. To classify follicles as either PCOS or non-PCOS, the SVM classifier was employed. In 2023, Narmatha et al., 27 introduced an innovative approach for identifying ovarian cysts. They utilized ultrasound images from a continuous dataset and followed a systematic process involving pre-processing, feature extraction, and classification. To accomplish this, a novel deep reinforcement learning technique combined with a Harris Hawks Optimization (HHO) classifier was employed. A Convolutional Neural Network (CNN) method was utilized for automatic feature extraction, where the extracted image features served as inputs in the learning algorithm. The researchers developed a deep Q-organization (DQN) to train the model and detect the disease. The HHO technique was employed to optimize the DQN hyperparameters model, referred to as HHO-DQN. The suggested HHO-DQN method surpassed current dynamic learning techniques for the categorization of ovarian cysts, according to investigational tests carried out using datasets. In 2023 Athithan et al. 28 proposed using ultrasound-based discovery of ovarian growth with improved AI algorithms and staging methods utilizing advanced classifiers. The study focused on using power-based gathering and textural information for follicle discovery and blisters in the ovary, which depends on AI (ML). They investigated the standard AI procedures to classify ovarian groups. Upon examining the results of the various classifiers, SVM had the highest precision of 98.5%. The study conducted by Poorani et al. 29 2003 proposed a method that involves preprocessing information images by applying edge-preserving filters to maintain edges, such as a bilateral filter, a median filter, and a Gaussian filter. The research aims to improve the detection of cysts of various ovary sizes. According to the rules of Rotterdam, the size and number of cysts require extensive analysis. The approach uses shape segmentation to delineate the ovary borderline and divide every ovary region. The paper examines the exactness of specific edge-preserving filters utilized for preprocessing images. After the preprocessing stage, they draw shapes to indicate the regions with follicles. To accurately identify these regions, we established a threshold of greater than 0.2. The results show that a Gaussian filter accurately segmented the shapes, achieving a similarity coefficient of 0.94. Research on cyst segmentation and classification has revealed several shortcomings. A primary challenge is achieving precise segmentation of cysts in postmenopausal women due to their small size. Current methods, including Adaptive Thresholding, Adaptive K-means, and the Watershed algorithm, struggle with accurate diagnosis. Additionally, existing optimization algorithms like HHO and RSA are insufficient for precise cyst description and require extensive training time. Segmentation of the cyst image's edges is difficult, leading to potential overfitting and incorrect size calculation due to improper weight updates. Classification techniques such as SVM, AI, and DLNN suffer from low accuracy, negatively impacting ultrasound image analysis. In contrast, the proposed algorithmic technique addresses these issues effectively, offering the highest accuracy for cyst detection in ultrasound images.

In recent years, infertility has emerged as a significant concern among individuals of reproductive age. A study conducted by the World Health Organization on 8500 infertile couples revealed that male infertility accounted for 8% of cases, while female infertility and a combination of both contributed to 37% and 35% respectively. Additionally, 5% of the couples experienced unexplained infertility, and 15% were able to conceive during the study. Various factors were identified as contributors to female infertility, including ovulatory dysfunction (25%), unexplained infertility (20%), endometriosis (15%), fallopian tube issues (22%), pelvic disorders (12%), and hyperprolactinemia (7%). Notably, ovarian cysts were found to be a common cause of female infertility, affecting a majority of infertile women 1 , 2 . These cysts resemble pimples and are located on both sides of the uterus in the lower abdomen. They play a crucial role in the production of eggs, estrogen, and progesterone hormones 3 , 4 . It is important to note that cysts, which are fluid-filled sacs, can significantly impact the health of female ovaries 5 . Cysts are commonly painless. In the early stages of an ovarian cyst's development, women typically experience irregular menstruation 6 . Within the sac, an egg called a follicle forms during the menstrual cycle. If the sac does not open during fertilization, 7 the fluid inside condenses into a cyst known as a cystic ovary. There are different sorts of cysts, such as endometriomas, corpus luteum cysts, dermoid cysts, 8 and cystadenomas. These cysts can also cause symptoms such as dizziness and breast fever. The cyst can grow up to 9 approximately 20 mm in size. It may cause back pain or other symptoms. The pain may persist in the affected area, especially during exercise 10 , 11 . In some cases, these cysts can lead to cancer, so surgical removal is necessary 12 . They can also cause pre-menstrual confusion, pelvic pain, and abdominal cramps, which can sometimes, be associated with uterine fibroids 13 . As the tumor grows, it can block and twist blood flow, resulting in worsening pain 14 . Despeckle filtering algorithms are an integral part of existing segmentation methodologies. These algorithms play a crucial role in refining segmentation outputs by reducing noise and artifacts present in image data. By effectively removing unwanted speckles and small anomalies, despeckle filters contribute significantly to improving the quality and reliability of segmented regions. They employ a variety of filtering techniques tailored to different types of noise and imaging conditions, thereby enhancing the overall effectiveness of segmentation processes in diverse applications such as medical imaging, satellite imagery analysis, and industrial quality control 15 , 16 . Ovarian ultrasounds pictures assist in notice the presence otherwise lack of cyst/follicle 9 . Ultrasound tests are employed to detect ovarian cysts by utilizing high-frequency sound waves to generate visual representations of the internal organs 17 , 18 . The elliptical shape of the cyst, along with certain darker regions within the image, can be utilized to identify and track its boundaries 19 . Nevertheless, distinguishing between benign and malignant cysts poses a challenge due to the diverse range of cyst types and associated symptoms. Previous research endeavors have employed different artificial intelligence and machine learning techniques to segment the cyst; however, achieving accuracy and reducing processing time still present obstacles. The contribution of the work is to segment cysts from ultrasound ovarian images. Initially, pre-processes an input ultrasound image to remove noise in the image using GTF. The proposed segmentation network portions of the cyst as benign or malignant based on their size, were segmented utilizing the proposed Adaptive Convolutional Neural Network upgraded by Wild Horse Optimizer (AdaResU-net with WHO). An objective function of Dice Loss Coefficient (DLC) combined with Weighted Cross Entropy (WCE) yields the segmented output. The cyst in the ovary is classified by its sort utilizing Pyramidal Dilated Convolution (PDC). Initially, pre-processes an input ultrasound image to remove noise in the image using GTF. The proposed segmentation network portions of the cyst as benign or malignant based on their size, were segmented utilizing the proposed Adaptive Convolutional Neural Network upgraded by Wild Horse Optimizer (AdaResU-net with WHO). An objective function of Dice Loss Coefficient (DLC) combined with Weighted Cross Entropy (WCE) yields the segmented output. The cyst in the ovary is classified by its sort utilizing Pyramidal Dilated Convolution (PDC). The structure of the paper is as follows: Section " Introduction " explains the female reproductive system with how ovarian cysts form. Section " Literature review " covers the segmentation and classification of ovarian cysts based on existing techniques. In Section " Proposed methodology " research gap was consulted. Section " Result and discussion " presents a proposed methodology that utilizes new algorithms and techniques, and the results and discussion are shown in “Discussion”. The paper is finally concluded in “ Conclusion and future works ”.