Table 1 Summary of included contributions (n=20)

Lehtinen et al. [21]

Finland

*

Case study

Comparing the performance of SOM and TPFFN in anticipating the possibility of PCOS

Patients: 54

Control group: 29

27 +

33 +

TPFFN accuracy was better than SOM.

Small sample data volume

*

SOM, TPFFN, MLP

*

Zhang et al. [22]

USA

*

RCT

Construction of classification models for the anticipation of the occurrence of ovulation treatment in women with PCOS

418

Clomiphene citrate: 27.9 + 4.0

The combination of clomiphene citrate and metformin: 28.3 + 4.0

Clomiphene citrate alone is better and superior to the other two methods for treating PCOS.

*

Decision trees

*

Mehrotra et al. [2]

India

*

Original

described a method that

Enables automatic diagnosis of PCOS based on features

Normal: 150

abnormal: 50

Normal: 32.24 ± 2.02

Abnormal: 31.24 ± 2.48

Bayesian classifier gives higher accuracy than logistic regression.

Using the probabilistic model helps doctors to screen early patients who are more likely to develop the disease.

Need to improve accuracy by using other classifiers

*

Bayesian Classifier,

Multivariate LR

*

Rethinavalli et al. [23]

India

*

Original

Proposing a new combinatorial structure to discover the severity of the disease in people with the disease

31

SQL

MATLAB R 2016 a

Dataset: Polycystic Ovarian Syndrome

Proliferative Phase Endometrial Cell Types

The structure based on fuzzy logic can be used in risk anticipation

The severity of the disease was improved. The proposed model performed better than the other created models with an accuracy of 93.64%

*

NFRS, ANN

*

Cahyono1 et al. [24]

Indonesia

*

Original

Designing and creating a system based on convolutional neural network to classify ultrasound images into two categories, sick and healthy

Patient: 40

Healthy: 14

3D matrix

Softmax

Loss function

Dropout

SGD method

F1-Measure

Micro-average F1-Measure

Automatic classification of images into two categories, sick and healthy, by the designed system

It was done well and was very accurate

*

CNN

*

Dewi et al. [25]

Indonesia

*

Original

System design based on machine learning and AI to help

Doctors can diagnose the disease more easily through ultrasound images

Gabor Wavelet method

The use of competitive neural network can increase the accuracy of diagnosis in this article

The highest accuracy is estimated at 80.84%. According to the results, the number of adopted features has a direct relationship with accuracy

*

Competitive Neural Network

*

Thufailah et al. [26]

Indonesia

*

Original

System design based on the Gibber-Violet method to extract features and

Helping to diagnose and classify disease

16–32 features

Gabor Wavelet method

The best accuracy of using the elemental neural network was 78.1%, which was achieved with 32 features.

A higher number of data for training the network can increase the accuracy of the network

More data for training

affects the time of diagnosis

*

Elman Neural Network,

Polynomial SVM,

Radial Basis Function SVM,

Linear SVM

*

Vikas et al. [27]

India

*

Original

Identify recurring patterns among the symptoms of PCOS patients using a set of frequently used items

119

18–22

PCOS Dataset source:

https://github.com/PCOS-Survey/PCOSData

Frequent Itemset Mining (FIM)

Spss

Using the mentioned algorithm to extract the main widgets

Here, the main signs have performed well for anticipation as well as determining relationships between features

The data set used is not enough.

In addition, Patients’ concerns about information disclosure

*

*

Apriori algorithm

*

Denny et al. [28]

India

*

Original

Designing and creating a system based on AI for assistance

To diagnose and anticipate PCOS disease

patients:

177

Healthy:

364

18–40

SPSS V 22.0

Principal Component Analysis (PCA)

Spyder Python IDE

HTML with SQL for designing a proper user

interface

Among the algorithms used, Algorithm

RF performed best with 89% accuracy. The system designed according to experts can be useful in early disease diagnosis and save time.

*

NB, LR, KNN, CART, RF, SVM

*

*

Thakre et al. [18]

India

*

Original

Design and build system based

On AI for help

to diagnose and anticipate PCOS disease

30 features

Jupyter Notebook

Python

This system helps in the early diagnosis and prediction of PCOS, and the RF algorithm is the most accurate and reliable algorithm with an accuracy of 90.9.

*

*

RF, LR

Linear SVM,

Radial SVM,

KNN,

Gaussian Naive Bayes

*

*

Abu Adla et al. [29]

Lebanon

*

Original

Designing a proposed model

for automatic diagnosis of PCOS

Patients:

177

Healthy:

364

18–40

“Polycystic Ovary Syndrome”dataset,

ML application

The best performance was related to the linear support vector machine, which was 90% accurate with 24.

Despite high accuracy in automatic model recognition

Suggestions did not show good performance in recall

*

SFFS, LR, DT, NB,

Linear SVM,

Polynomial SVM,

Radial Basis Function SVM,

Linear Discriminant Classifier,

Quadratic Discriminant, RF

*

Hassan et al. [30]

India

*

Original

Design and build system based

on AI for help

to diagnose PCOS and compare the performance of different algorithms

42 variables

R-language

R libraries: e1071, CARET, naivebayes, rpart, randomForest, klaR, ggplot2

Among the 5 algorithms used, RF algorithm and support vector machine respectively

Accuracy of 96% and 95% performed better.

*

LR, SVM

NB, CART, RF

*

Kodipalli et al. [31]

India

*

Original

Designing a model for disease anticipation and related mental disorders based

624

Patients under 25

Questionnaire,

K10 tool,

matplotlib,

Fuzzy TOPSIS

The use of the system is cost-effective. The performance of SVM and fuzzy algorithms was 94.01% and 98.2%, respectively.

*

D-Tree, KNN,

SVM, Fuzzy

*

Song et al. [32]

China

*

Original

This study proposed a model based on Artificial intelligence algorithm, which is a non-invasive method with the help of captured images

It was from the eyes to help diagnose PCOS.

721

U-Net network,

convolutional block attention module (CBAM),

multi-instance (MIL), MLP,

Resnet18

A non-invasive method,

The accuracy of this method was estimated at 0.978%.

Ambiguities in the images,

There is a need to conduct more studies to generalize the results

*

CNNs: V3, Vgg16, and Vgg19

*

Mandal et al. [16]

India

*

Original

Providing an automated diagnostic approach for

Detection of follicles in the ovary using ultrasound (US) images during infertility treatment.

19

histogram equalization

This method can automatically detect the follicles

Ultrasound images are effective in reducing the workload of doctors.

To determine the exact shape and size of the follicles

There are more features that need to be considered.

*

K-means clustering

*

Nilofer et al. [33]

India

*

Original

Presenting a proposed method for

automatic division of areas in ultrasound images into areas with follicles and without follicles.

Wiener filter,

Takagi–Sugeno–Kang (TSK),

fuzzy inference

method,

Maximum Likelihood (ML),

Extreme Learning Adaptive Neuro-inference System (ELANFIS)

The proposed combined model had 99% accuracy in detecting follicles.

Further research is needed

to be done by institutions and stakeholders to confirm the model.

*

Fuzzy logicis,

Hybrid, Intelligent Water Drop (IWD),

KNN,

SVM

*

Zhang et al. [34]

China

*

Original

Designing a system based on deep learning for the anticipation of diseases related to genetics including PCOS

Thousans of genetic variants

DisGeNET,

GWAS Catalog,

GTEx Portal

The current algorithm in the field of predicting the relationship of disease with genetics compared with algorithms

Classics such as RF and Support Vector Machine performed better.

*

CNN, GCN

*

Hosain et al. [35]

Bangladesh

*

Observational study

Development of a system called PCONet

To help diagnose pcos through convolutional neural network-based ultrasound images

Dataset 1: 1730 images

Dataset 2: 339 images

Image Data Generator,

Keras

The present system not only performed well in diagnosing the disease through images, but also performed better with an accuracy of 98.12.

*

CNN,

InceptionV3

*

Zigarelli et al. [36]

United States of

America

*

Retrospective study

developing self-diagnostic prediction models for PCOS in potential patients and clinical providers

541

20–48

Rotterdam criteria

PCA Method

The prediction accuracy was estimated to be 87.5 to 90.1%

The sample was drawn from a specific population in India

from several hospitals.

*

K-Means Clustering,

CatBoost model

*

Nsugbe et al. [37]

England

*

Original

Designing and creating a decision

support system based on AI to diagnose PCOS and determine the stage of the disease

Patients:

177

Healthy:

364

Kaggle website

SVM performed better than

other used algorithms.

More samples with more diverse data for

presenting the model in the clinical environment is needed

*

DT, LDA, LR, KNN, SVM

*