A system for the early detection of polycystic ovary syndrome

Abstract

A system (100) for the early detection of polycystic ovary syndrome (PCOS), wherein the system (100) comprises the following: an input unit (102) configured to capture PCOS records of a patient from a data set; a preprocessing unit (104) connected to the input unit (102), which is configured to identify instances with missing values ​​in the data set by eliminating the PCOS data sets with missing data; a classification unit (106) that is operationally connected to the preprocessing unit (104) and is configured to classify instances with missing values, wherein the classification unit (106) comprises a plurality of classifiers that include the following: a variety of linear classifiers (106a), including Gaussian Naive Bayes, Logistic Regression, Support Vector Machine and Linear Discriminant Analysis, configured to separate a variety of classes by constructing linear decision boundaries in the feature space; a variety of nonlinear classifiers (106b), including K-Nearest Neighbors, Support Vector Machine with a polynomial kernel, quadratic discriminant analysis and decision tree, configured to capture complex relationships between features and classes; where output predictions from the multitude of classifiers are chained together to improve accuracy and generalization; a cross-validation unit (108) that is operationally integrated with the classification unit (106) and is configured to perform a 10-fold cross-validation of the classifiers to evaluate the performance of the classifiers in detecting PCOS; a performance evaluation unit (110) connected to the cross-validation unit (108) and configured to calculate evaluation measures including accuracy and F1 score for each classifier and to identify the best-performing classifier based on the evaluation measures; and an output interface (112) which is integrated into a display (112a) and is configured to display the diagnostic results including the classifier with the highest accuracy and the highest F1 score.

Description

AREA OF INVENTION

[0001] The present invention relates to the field of medical systems. In particular, the present invention relates to the early detection of polycystic ovary syndrome (PCOS). More precisely, the present invention relates to a system for diagnosing polycystic ovary syndrome (PCOS). BACKGROUND OF THE INVENTION

[0002] Polycystic ovary syndrome (PCOS) is a widespread hormonal disorder affecting millions of women worldwide, particularly during puberty and the reproductive years. It is one of the leading causes of female infertility and is associated with several long-term health complications, including type 2 diabetes, cardiovascular disease, and high blood pressure. An estimated 6 to 18% of women during puberty and 8 to 13% of women of reproductive age are affected by PCOS. Although PCOS is incurable, early diagnosis and appropriate treatment can significantly alleviate symptoms and reduce the risk of these long-term health complications. However, diagnosing PCOS is challenging because the symptoms are highly varied and differ from woman to woman, making it difficult to establish a single, definitive diagnostic pattern.

[0003] Conventional diagnostic methods for PCOS are typically based on a combination of the patient's medical history, physical examinations, blood test results, and ultrasound findings. The presence of multiple ovarian cysts, identified by ultrasound, is a common feature of PCOS, but not all women with PCOS have these cysts. Therefore, ultrasound diagnoses are not entirely reliable and frequently lead to misdiagnoses. This inconsistency has led to the development of alternative diagnostic methods based solely on physical and clinical parameters.

[0004] In recent years, machine learning techniques have proven highly promising in the healthcare sector, particularly in diagnosing complex conditions such as PCOS. Machine learning models are capable of processing large datasets and recognizing complex patterns within the data, making them well-suited for medical diagnosis. These models are generally divided into two categories: single-classifier models, which rely on individual interfaces, and ensemble models, which combine multiple classifiers to generate more robust and accurate predictions.

[0005] Although various machine learning models have been proposed for PCOS detection, there is no comprehensive study report comparing the performance of single classifier-based and ensemble models for PCOS diagnosis.

[0006] To overcome the aforementioned limitations, there is therefore a need for early detection of PCOS symptoms using physical and clinical parameters for the diagnosis of polycystic ovary syndrome (PCOS), based on a known Kaggle PCOS dataset and evaluated using 10-fold cross-validation and key performance indicators such as accuracy, precision, recall and F1 score.

[0007] The technical advances disclosed by the present invention overcome the limitations and disadvantages of existing and conventional systems and methods. SUMMARY OF THE INVENTION

[0008] The present invention relates to a system for diagnosing polycystic ovary syndrome (PCOS). PCOS is a complex disorder that affects women during their reproductive years and frequently leads to infertility and long-term health complications such as diabetes, cardiovascular disease, and hypertension. The present system addresses the challenges posed by variable symptoms in PCOS patients. Using the Kaggle PCOS dataset, a comparative analysis is performed on models based on both single classifiers (e.g., Gaussian Naive Bayes, logistic regression, support vector machine, and decision tree) and ensemble models (e.g., random forest, gradient boosting, XGBoost). After 10-fold cross-validation, the random forest classifier demonstrated the best performance, achieving an accuracy of 90.1% and an F1 score of 83.6%.The present system offers a powerful diagnostic tool that helps medical professionals to make timely and accurate PCOS diagnoses, thus enabling better treatment planning and better patient outcomes.

[0009] One objective of the present invention is to provide a system for the early detection of polycystic ovary syndrome (PCOS).

[0010] Another objective of the present invention is the development and optimization of a system that improves the diagnostic accuracy of polycystic ovary syndrome (PCOS).

[0011] Another objective of the present invention is to identify the most effective classifier for PCOS diagnosis based on key performance metrics such as accuracy, F1 score and robustness, thereby supporting healthcare providers in the timely and reliable diagnosis of PCOS.

[0012] Another objective of the present invention is to improve the overall efficiency of PCOS detection, reduce diagnostic delays and contribute to better patient outcomes through early intervention.

[0013] A further objective of the present invention is to provide a system for diagnosing polycystic ovary syndrome (PCOS), wherein the system comprises: an input unit configured to extract PCOS patient records from a dataset; a processing unit configured to identify instances with missing values ​​in the dataset by eliminating the records with missing data; a single classification unit comprising for combining predictions from multiple classifiers, including: multiple linear classifiers, including Gaussian Naive Bayes, Logistic Regression, Support Vector Machine, and Linear Discriminant Analysis, configured to separate different classes by drawing linear decision boundaries in the feature space;Several nonlinear classifiers, including K-Nearest Neighbors, Support Vector Machine with polynomial kernel, Quadratic Discriminant Analysis, and Decision Tree, configured to capture complex relationships between features and classes that cannot be represented by linear decision boundaries; an ensemble classification unit configured to combine predictions from multiple classifiers to improve accuracy and generalization; a cross-validation unit configured to perform 10-fold cross-validation of the classifiers to evaluate their performance in diagnosing PCOS; and a performance evaluation unit configured to calculate evaluation metrics, including accuracy and F1 score, for each classifier and identify the best-performing classifier based on these metrics.and an output interface integrated into a display, configured to show diagnostic results, including the highest accuracy classifier and the highest F1 score.

[0014] To further clarify the advantages and features of the present disclosure, a more detailed description of the invention is given with reference to specific embodiments thereof, which are illustrated in the accompanying drawing. It should be noted that this drawing only represents typical embodiments of the invention and is therefore not to be considered as limiting its scope. The invention is described and explained with additional details and information with reference to the accompanying drawing. BRIEF DESCRIPTION OF THE FIGURE

[0015] These and other features, aspects, and benefits of this disclosure will be better understood when the following detailed description is read with reference to the accompanying drawing, in which the same symbols consistently represent the same parts. The following applies: Fig. Figure 1 shows a block diagram of a system for diagnosing polycystic ovary syndrome according to an embodiment of the present disclosure.

[0016] Furthermore, those skilled in the art will recognize that elements in the drawing are shown for the sake of simplicity and are not necessarily drawn to scale. For example, the flowcharts illustrate the method by outlining the main steps to enhance understanding of aspects of this disclosure. Additionally, with regard to the design of the device, one or more components of the device may be represented in the drawing by conventional symbols, and the drawing may show only the specific details relevant to understanding the embodiments of this disclosure, so as not to clutter the drawing with details that are readily apparent to those skilled in the art who benefit from the description herein. DETAILED DESCRIPTION:

[0017] To facilitate understanding of the principles of the invention, reference is now made to the embodiment illustrated in the drawing, which is described in specific language. However, it is clear that this is not intended to limit the scope of the invention, since such changes and further modifications of the illustrated system and such further applications of the principles of the invention as they are shown therein are considered as would normally occur to a person skilled in the art in the field to which the invention relates.

[0018] It is clear to the person skilled in the art that the preceding general description and the following detailed description are exemplary and explanatory of the invention and are not intended as a limitation of it.

[0019] References in this specification to “an aspect”, “another aspect”, or similar formulations mean that a particular feature, structure, or property described in connection with the embodiment is included in at least one embodiment of the present disclosure. Therefore, occurrences of the phrase “in one embodiment”, “in another embodiment”, and similar formulations in this specification may all refer to the same embodiment, but need not.

[0020] The terms "includes," "comprehensive," or other variations thereof are intended to cover non-exclusive inclusion, so that a process or method that includes a list of steps does not only include those steps but may also include other steps not expressly listed or inherent in such process or method. Likewise, one or more devices, subsystems, elements, structures, or components preceded by "includes..." do not, without further limitations, preclude the existence of other devices, subsystems, elements, structures, components, or additional devices, subsystems, elements, structures, or components.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as they are generally understood by a person skilled in the art in the field to which this invention belongs. The system, methods, and examples provided herein are for illustrative purposes only and are not to be understood as limiting.

[0022] Embodiments of the present disclosure are described in detail below with reference to the attached drawing.

[0023] Fig.Figure 1 shows a block diagram of a system (100) for diagnosing polycystic ovary syndrome (PCOS), wherein the system (100) comprises: an input unit (102) configured to extract PCOS patient records from a dataset; a preprocessing unit (104) connected to the input unit (102) configured to identify instances with missing values ​​in the dataset by eliminating the records with missing data;a classification unit (106) that is operationally connected to and configured with the preprocessing unit (104) to classify the instances with missing values, wherein the classification unit (106) comprises a variety of classifiers, including: a variety of linear classifiers (106a), including Gaussian Naive Bayes, Logistic Regression, Support Vector Machine and Linear Discriminant Analysis, configured to separate different classes by drawing linear decision boundaries in the feature space;a variety of nonlinear classifiers (106b), including K-Nearest Neighbours, Support Vector Machine with polynomial kernel, quadratic discriminant analysis, and decision tree, configured to capture complex relationships between features and classes that cannot be represented by linear decision boundaries, with output predictions from the variety of classifiers chained together to improve accuracy and generalization; a cross-validation unit (108) configured to perform 10-fold cross-validation of the classifiers to evaluate their performance in detecting PCOS; a performance evaluation unit (110) connected to the cross-validation unit (108) configured to calculate evaluation measures, including accuracy and F1 score, for each classifier and to identify the best-performing classifier based on the evaluation measures;and an output interface (112) integrated into a display (112a) configured to display the diagnostic results, including the classifier with the highest accuracy and the highest F1 score.;

[0024] In one embodiment, the processing unit (104) is also configured to exclude, in order to maintain data integrity, the data records that make up less than 1% of the data record.

[0025] In one embodiment, the classification unit (106) uses Random Forest as the primary ensemble classifier, which achieves an accuracy of 90.1% and an F1 score of 83.6% in the diagnosis of PCOS.

[0026] In one embodiment, the classification unit (106) comprises Extra Trees, Random Forest, Gradient Boosting, AdaBoost, and XGBoost, wherein Extra Trees reduces overfitting by introducing random thresholds for feature partitions, Random Forest is an ensemble of decision trees in which each tree is trained on a bootstrap sample and random features are selected for each partition, Gradient Boosting creates weak learners sequentially, with each learner attempting to correct the errors of the previous one, AdaBoost focuses on misclassified instances in each iteration and creates a strong classifier from several weak learners, and XGBoost incorporates regularization and parallel processing to improve performance.

[0027] In one embodiment, the cross-validation unit (108) applies k-Nearest Neighbors (KNN) with an adjustable parameter for selecting the number of neighbors (k) in the non-linear classification.

[0028] In one embodiment, the performance evaluation unit (110) implements regularization techniques in ensemble classifiers, including L2 regularization in XGBoost for improved model performance.

[0029] In one embodiment, the PCOS dataset used for the present system is collected from the Kaggle Dataset Repository and has been used in several research papers in recent years. The dataset contains medical records of 541 women from 10 different hospitals in Kerala, India. Besides PCOS, which is the target, the dataset contains 41 attributes. Of these 41 attributes, 9 are categorical. Therefore, the dataset contains mixed attribute types as follows: 1. PCOS (Y / N) 2. Age 3. Weight 4. Height 5. Body Mass Index 6. Blood type 7. Pulse rate 8. RR (breaths / min) 9. Hemoglobin 10th menstrual cycle 11. Cycle length 12. Marital status 13. Pregnant (Y / N) 14. Number of abortions 15. I beta-HCG 16. II Beta-HCG 17. FSH hormone 18. LH hormone 19. FSH / LH 20. Hip circumference 21. Waist circumference 22. Waist:hip ratio 23. TSH hormone 24. AMH hormone 25. Prolactin 26. Vitamin D3 27. Progesterone 28. Blood sugar 29. Weight gain 30. Hair growth 31. Skin darkening 32. Hair loss 33. Pimples 34. Fast Food 35. Regular exercise 36. Systolic blood pressure 37. Diastolic blood pressure 38. Follicle No. (L) 39. Follicle No. (R) 40. Average F-size (L) 41. Average F-size (R) 42. Endometrium

[0030] The key performance indicators (KPIs) used to evaluate the performance of machine learning models in diagnosing patients with and without PCOS are accuracy, precision, recall, and F1 score. The mathematical equations used to calculate these KPIs are listed below: Accuracy = TP + TNTP + FP + TN + FN Precision = TPTP + FP Recall = TPTP + FN F1−Score=2∗Precision∗RecallPrecision+Recall where TP, TN, FP and FN stand for "True Positive", "True Negative", "False Positive" and "False Negative", respectively.

[0031] Two categories of machine learning models—single classifier-based models (including both linear and nonlinear classifiers) and ensemble models—are applied to the Kaggle PCOS dataset to compare the performance of different classifiers in diagnosing patients with and without PCOS. Tenfold cross-validation is used to robustly and reliably assess model performance. Among the linear classifiers, linear discriminant analysis (LDA) achieves the highest performance with an accuracy of 89.3% and an F1 score of 82.4%, closely followed by support vector machine (SVM) and logistic regression (LR), which exhibit accuracies of 88.8% and 88.4%, respectively, and F1 scores of 82.1% and 81.1%. Gaussian Naive Bayes (GNB) performs worst among the linear classifiers, with an accuracy of 83.6% and an F1 score of 76.1%.This is probably due to the assumptions about feature independence and normal distribution, which may not apply to the data set.

[0032] Among nonlinear classifiers, Decision Tree delivers the best results with an accuracy of 83% and an F1 score of 74%, followed by Quadratic Discriminant Analysis (QDA), SVM with a nonlinear kernel, and K-Nearest Neighbors (KNN). The performance of nonlinear classifiers lags behind their linear counterparts, possibly due to the linear nature of the dataset, data noise, data size, or insufficient hyperparameter tuning. Among ensemble models, Random Forest outperforms Extra Trees with an accuracy of 90.1% and an F1 score of 83.6%, likely because Random Forest requires less hyperparameter tuning. Among boosting models, Gradient Boost achieves the best performance with an accuracy of 89% and an F1 score of 81.9%, followed by XGBoost and AdaBoost.Overall, linear classifiers, especially LDA, and ensemble models such as Random Forest and Gradient Boost show better performance in diagnosing PCOS. Models based on a single classifier Evaluation metrics Linear Accuracy it precision Retrievals n F1_score GNB 83.6 71.4 83.8 76.1 LR 88.4 82.6 80 81.1 SVM (Kernel: linear) 88.8 83.1 81.5 82.1 LDA 89.3 87.4 78.3 82.4 Nonlinear KNN 68.3 52 33.8 40.3 SVM (Kernel: Poly) 82.1 72.3 70 71 QDA 83 76.6 68.6 71.1 Decision tree 83 73.9 72.4 74 Ensemble models sagging Additional trees 87.7 85.7 74.6 79.1 Random forest 90.1 87.4 76.1 83.6 Reinforcement AdaBoost 88 84.8 77.3 80.3 XGBoost 88.6 84 79.6 81.5 Gradient enhancement 89 84 79.7 81.9

[0033] From the results above, it can be observed that Random Forest outperforms all single-classifier-based and ensemble models. The accuracy and F1 score achieved by the Random Forest bagging ensemble model are 90.1% and 83.6%, respectively, representing the best values ​​of all models considered in this study. The experimental results demonstrate the superiority of ensemble learning models over single-classifier-based models. Furthermore, bagging shows better performance than the boosting technique in this study, likely due to the following characteristics of the dataset: i) the dataset is not very large, and ii) the dataset contains mixed attribute types. In such cases, bagging tends to perform better than boosting because it involves training multiple models independently on different subsets of the data.Each model therefore sees a slightly different version of the data set, which is why bagging helps to reduce variance, avoid overfitting in small data sets, and handle different attribute types more effectively.

[0034] The drawing and the preceding description provide examples of embodiments. Those skilled in the art will recognize that one or more of the described elements can be combined to form a single functional element. Alternatively, certain elements can be divided into several functional elements. Elements from one embodiment can be added to another embodiment. For example, the sequences of the processes described herein can be changed and are not limited to the manner described herein. Furthermore, the actions of any flowchart need not be implemented in the order shown; nor do all actions necessarily need to be performed. Actions that are not dependent on other actions can also be performed in parallel with the other actions. The scope of embodiments is in no way limited by these specific examples.Numerous variations are possible, whether explicitly stated in the specification or not, such as differences in structure, dimensions, and material use. The range of embodiments is at least as broad as specified in the following claims.

[0035] Advantages, other merits, and problem solutions have been described above in relation to specific embodiments. However, the advantages, merits, problem solutions, and all components that may lead to an advantage or solution occurring or becoming more apparent are not to be construed as a critical, necessary, or essential feature or component of any or all claims. REFERENCES 100 A System for Diagnosing Polycystic Ovary Syndrome (PCOS). 102 Input unit 104 Pre-processing unit 106 Classification Unit 106a Variety of Linear Classifiers 106b Variety of Nonlinear Classifiers 108 Cross-validation unit 110 performance evaluation units 112 Output interface 112a Ad

Claims

[1] A system (100) for the early detection of polycystic ovary syndrome (PCOS), the system (100) comprising: an input unit (102) configured to capture PCOS records of a patient from a data set; a preprocessing unit (104) connected to the input unit (102) and configured to identify instances with missing values ​​in the data set by eliminating the PCOS records with missing data; a classification unit (106) operatively connected to the preprocessing unit (104) and configured to classify the instances with missing values, the classification unit (106) comprising a plurality of classifiers comprising: a variety of linear classifiers (106a), including Gaussian Naive Bayes, Logistic Regression, Support Vector Machine, and Linear Discriminant Analysis, configured to separate a plurality of classes by constructing linear decision boundaries in the feature space; a variety of nonlinear classifiers (106b), including K-Nearest Neighbors, Support Vector Machine with a polynomial kernel, Quadratic Discriminant Analysis, and Decision Tree, configured to capture complex relationships between features and classes; wherein output predictions of the plurality of classifiers are concatenated to improve accuracy and generalization; a cross-validation unit (108) operatively integrated with the classification unit (106) and configured to perform 10-fold cross-validation of the classifiers to evaluate the performance of the classifiers in detecting PCOS; a performance evaluation unit (110) connected to the cross-validation unit (108) and configured to calculate evaluation measures including accuracy and F1 score for each classifier and to identify the best-performing classifier based on the evaluation measures; and an output interface (112) integrated with a display (112a) and configured to display the diagnostic results including the classifier with the highest accuracy and the highest F1 score. [2] The system (100) of claim 1, wherein the preprocessing unit (104) is further configured to exclude the records that constitute less than 1% of the data set to maintain data integrity. [3] The system (100) of claim 1, wherein the classification unit (106) uses Random Forest as the primary ensemble classifier for early and accurate detection of PCOS. [4] The system (100) of claim 1, wherein the classification unit (106) comprises Extra Trees, Random Forest, Gradient Boosting, AdaBoost, and XGBoost, wherein Extra Trees reduces overfitting by introducing random thresholds for feature splits, wherein Random Forest is an ensemble of decision trees where each tree is trained on a bootstrap sample and random features are selected for each split, wherein Gradient Boosting builds weak learners one after the other, each learner trying to correct the errors of the previous one, wherein AdaBoost focuses on misclassified instances in each iteration and builds a strong classifier from multiple weak learners, and wherein XGBoost incorporates regularization and parallel processing to improve performance. [5] The system (100) of claim 1, wherein the cross-validation unit (108) applies k-nearest neighbors (KNN) with an adjustable parameter for selecting the number of neighbors (k) in the non-linear classification. [6] The system (100) of claim 1, wherein the performance evaluation unit (110) implements regularization techniques in ensemble classifiers, including L2 regularization in XGBoost to improve model performance.

Images