chronic kidney disease using ai modelss

medical data ppt

1. DATA-DRIVEN EARLY DIAGNOSIS OF CHRONIC KIDNEY DISEASE USING AI MODEL PRESENTATION by

2. Agenda Abstract Introduction Existing Models Proposed system Proposed models Results Conclusion

3. Abstract Chronic Kidney Disease (CKD) is currently experiencing a growing worldwide incidence and can lead to premature mortality if diagnosed late, resulting in rising costs to healthcare systems. Artificial Intelligence (AI) and Machine Learning (ML) offer the possibility of an early diagnosis of CKD that could revert further kidney damage. However, clinicians may be hesitant to adopt AI models if the reasoning behind the predictions is not understandable. Since explainable AI (XAI) addresses the clinicians’ requirement of understanding AI models’ output, this work presents the development and evaluation of an explainable CKD prediction model that provides information about how different patient’s clinical features contribute to CKD early diagnosis. 3

4. This notebook focuses on predicting chronic kidney disease using various machine learning models. The dataset contains features like age, blood pressure, specific gravity, albumin, sugar levels, blood cell counts, and other medical indicators. After loading and preprocessing the data, the notebook performs exploratory data analysis using visualizations like histograms, violin plots, scatter plots, and bar charts to understand the distribution and relationships between features. Several classification models are then built and evaluated, including Decision Tree, AdaBoost, Gradient Boosting, Stochastic Gradient Boosting, XGBoost, CatBoost , Extra Trees, and LightGBM. Hyperparameter tuning is performed for some models using techniques like GridSearchCV . Model performance metrics like accuracy, confusion matrix, and classification report are computed on test data. 4

5. The results show that ensemble methods like AdaBoost, Extra Trees, XGBoost, and CatBoost achieve the highest accuracy around 96-97% in predicting chronic kidney disease. Decision Trees have the lowest accuracy of around 90%. The notebook concludes by visualizing the comparison of all the model scores using a bar chart. 5

6. Introduction  Chronic kidney disease (CKD) is a major public health concern worldwide, affecting millions of people and posing a significant burden on healthcare systems. Early detection and accurate prediction of CKD progression are crucial for timely intervention and improving patient outcomes. Kidneys are critical organs within the human frame that perform a ramification of important features. Humans have two fist-sized kidneys. Their most important reason is to filter the blood. It removes waste products and extra water and turns them into urine. 6

7. It additionally facilitates preserve the body's chemical stability, controls blood pressure, and produces hormones. Kidney disorder affects greater than 750 million human beings worldwide. Kidney disorder is a condition that affects human beings worldwide, however disorder prevalence, detection, and treatment vary broadly. Kidney failure is the main reason of loss of life in modern-day society. This scenario is exacerbated by smoking, binge drinking, high cholesterol, and numerous other risk factors. 7

8. Situations that injure your kidneys and lessen their ability to keep you healthy by filtering waste from your blood constitute chronic kidney disease. Wastes might accumulate to excessive levels for your blood and make you feel sick if renal disease develops. Additionally, kidney disease increases your risk of developing heart and blood vessel problems. Those problems could develop slowly over a long period of time. Chronic renal disease can frequently be prevented from going worse with early detection and treatment. As a kidney condition worsens, renal failure could develop, necessitating dialysis or a kidney transplant to maintain life. 8

9. 9 Normal and abnormal kidney images

10. Literature Review Prevalence and Impact of CKD. Challenges in Early Diagnosis. Role of Data-Driven Approaches. Explainable AI Models. Development of AI Models for CKD Diagnosis. Evaluation and Validation. Clinical Implementation and Impact. Ethical and Social Implications. 10

11. 3. E-Voting Security: Review literature on e-voting security protocols, including encryption techniques, authentication methods, and tamper detection mechanisms. Examine research papers and articles addressing vulnerabilities in e-voting systems and proposed solutions to mitigate risks. 4. Usability and Accessibility: Study literature on usability and accessibility considerations in e-voting systems, particularly for users with disabilities or limited technological proficiency. Investigate user interface design principles and human-computer interaction (HCI) guidelines relevant to voting applications . 11

12. Prevalence and Impact of CKD : Numerous studies underscore the global prevalence and impact of CKD, emphasizing its status as a major public health issue. CKD affects millions worldwide, leading to significant morbidity, mortality, and economic burden. Challenges in Early Diagnosis : Early detection of CKD is crucial for effective management and prevention of complications. However, traditional diagnostic methods may be limited in their ability to detect CKD in its early stages, leading to delayed interventions and poorer outcomes. Role of Data-Driven Approaches : Data-driven approaches, particularly those leveraging artificial intelligence (AI) and machine learning (ML), hold promise for improving early CKD diagnosis. These methods can analyze large datasets encompassing patient demographics, clinical records, laboratory results, and imaging findings to identify patterns indicative of CKD onset or progression. 12

13. Explainable AI Models : Explainable AI models are gaining traction in healthcare applications due to their ability to provide transparent insights into decision-making processes. In the context of CKD diagnosis, explainable AI models offer clinicians and patients explanations or justifications for the predictions they generate, enhancing trust and facilitating clinical decision-making. Development of AI Models for CKD Diagnosis : Researchers have developed various AI models for CKD diagnosis, ranging from traditional machine learning algorithms to more advanced deep learning architectures. These models are trained on diverse datasets and may incorporate features such as laboratory values, imaging results, comorbidities , and lifestyle factors. 13

14. Evaluation and Validation : Robust evaluation and validation of AI models are essential to ensure their reliability and generalizability in clinical practice. Studies often employ rigorous methodologies, including cross-validation, external validation, and comparison with existing diagnostic tools, to assess the performance of AI models in detecting CKD. Clinical Implementation and Impact : Successful implementation of AI models in clinical practice requires considerations such as integration with electronic health record systems, regulatory compliance, clinician training, and patient engagement. Studies exploring the real-world impact of AI-based CKD diagnosis on patient outcomes, healthcare utilization, and cost-effectiveness are emerging. 14

15. Ethical and Social Implications : The adoption of AI in healthcare raises ethical and social implications related to data privacy, algorithm bias, equity in access to care, and the role of healthcare providers in decision-making. Addressing these concerns is paramount to ensuring the responsible and equitable use of AI in CKD diagnosis. 15

16. EXISISTING MODELS Random Forest Extra Tress AdaBoost XGBoost 16

17. Random forest Random forest is a commonly-used machine learning algorithm, trademarked by Leo Breiman and Adele Cutler, that combines the output of multiple decision trees to reach a single result. Its ease of use and flexibility have fueled its adoption, as it handles both classification and regression problems.

18. Extra trees Extra Trees is an extension of Random Forests. In Extra Trees, randomness is further increased by using random thresholds for each feature rather than searching for the best possible thresholds. This extra randomness makes Extra Trees faster to train than Random Forests because it doesn't need to search for the optimal split at each node.

19. A daBoost Ada -boost, known as Adaptive Boosting, is one ensemble boosting classifier. It combines multiple classifiers to increase the accuracy of classifiers. It is an iterative ensemble method. AdaBoost classifier builds a robust classifier by combining multiple poorly performing classifiers to get high accuracy robust classifier. The basic concept behind Adaptive Boosting eka AdaBoost is to set the weights of classifiers and to train the data sample in each iteration such that it ensures accurate predictions of unusual or unseen observations.

20. XG Boost XGBoost classifier incorporates a sparsity-aware split finding algorithm to handle different types of sparsity patterns in the data. Weighted quantile sketch: Most existing tree based algorithms can find the split points when the data points are of equal weights (using quantile sketch algorithm).

21. Proposed system Data collection Data Loading and Exploration Data Preprocessing Model Building Model Comparison 21

22. Data collection: In this project I'm collecting the data in kaggle datasets 22

23. Data Loading and Exploration: Loading the kidney disease dataset Checking for missing values and dealing with them through random value imputation and mode imputation Exploratory data analysis (EDA) through visualizations like histograms, violin plots, scatter plots to understand data distributions 23

24. Data Preprocessing: Encoding categorical features using label encoding Splitting data into training and test sets 24

25. Model Building: Implementing various machine learning classification algorithms like Decision Tree, AdaBoost, Gradient Boosting, Stochastic Gradient Boosting, XGBoost, CatBoost , Extra Trees, LightGBM Tuning hyperparameters of some models using techniques like GridSearchCV Evaluating models on test data using metrics like accuracy, confusion matrix, classification report 25

26. Model Comparison: Comparing the performance scores (accuracy) of the different models Identifying the best performing models like AdaBoost, Extra Trees, XGBoost, CatBoost for chronic kidney disease prediction 26

27. Proposed block 

28. Proposed MODELS Ada Boost Extra Trees classifier Random Forest Classifier XgBoost Cat Boost Stochastic Gradient Boosting Gradient Boosting Classifier Decision Tree Classifier KNN

29. COMPARISION TABLE SNO MODEL SCORE 1 Extra Trees Classifier 0.991667 2 Gradient Boosting Classifier 0.983333 3 Stochastic Gradient Boosting 0.983333 4 XgBoost 0.983333 5 Cat Boost 0.983333 6 Decision Tree Classifier 0.975000 7 Random Forest Classifier 0.975000 8 Ada Boost Classifier 0.975000 9 KNN 0.716667

30. MODELS COMPARISON 30

31. Conclusion 31 In conclusion, ensemble boosting techniques like AdaBoost , Gradient Boosting and tree-based models like Extra Trees and XGBoost performed exceptionally well, achieving around 98% accuracy in predicting chronic kidney disease from the given dataset. The notebook provides a comprehensive analysis for tackling this binary classification problem effectively.

32. Feature scope 32 Future research in this area should focus on Expanding the scope of the model to detect other kidney abnormalities and on validating the model's performance on larger datasets. Expanding the model's capabilities to identify other kidney problems.

33. Any Quires? 33

34. THANK YOU

35. REfrences [1] Adeola Ogunleye et al., XGBoost Model for Chronic Kidney Disease Diagnosis. [2] IEEE/ACM Transactions on Computational Biology and Bioinformatics, Volume: 17,2019, Issue: 6, pp.2131 – 2140. [3] Adriano Luiz Ammirati , Chronic Kidney Disease, Rev assoc med bras,2020 66(SUPPL 1): S3-S9. [4] Angela C Webster et al. (Chronic Kidney Disease), The Lancet, Volume389, Issue 10075, 25–31 2017, Pages 1238-1252. [5] Dervla M. Connaughton et al. Monogenic causes of Chronic Kidney Disease in adults, Clinical investigation, 2019, volume 95, issue 4, p914-928.