A novel machine learning approach for diagnosing diabetes with a self-explainable interface

Healthcare Analytics, Journal Year: 2024, Volume and Issue: 5, P. 100301 - 100301

Published: Jan. 21, 2024

This study introduces the first-ever self-explanatory interface for diagnosing diabetes patients using machine learning. We propose four classification models (Decision Tree (DT), K-nearest Neighbor (KNN), Support Vector Classification (SVC), and Extreme Gradient Boosting (XGB)) based on publicly available dataset. To elucidate inner workings of these models, we employed learning interpretation method known as Shapley Additive Explanations (SHAP). All exhibited commendable accuracy in with diabetes, XGB model showing a slight edge over others. Utilising SHAP, delved into model, providing in-depth insights reasoning behind its predictions at granular level. Subsequently, integrated SHAP's local explanations an to predict patients. serves critical role it diagnoses offers transparent decisions made, users heightened awareness their current health conditions. Given high-stakes nature medical field, this developed can be further enhanced by including more extensive clinical data, ultimately aiding professionals decision-making processes.

Language: Английский

---

Identifying top ten predictors of type 2 diabetes through machine learning analysis of UK Biobank data

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Jan. 24, 2024

Abstract The study aimed to identify the most predictive factors for development of type 2 diabetes. Using an XGboost classification model, we projected diabetes incidence over a 10-year horizon. We deliberately minimized selection baseline fully exploit rich dataset from UK Biobank. value features was assessed using shap values, with model performance evaluated via Receiver Operating Characteristic Area Under Curve, sensitivity, and specificity. Data Biobank, encompassing vast population comprehensive demographic health data, employed. enrolled 450,000 participants aged 40–69, excluding those pre-existing Among 448,277 participants, 12,148 developed within decade. HbA1c emerged as foremost predictor, followed by BMI, waist circumference, blood glucose, family history diabetes, gamma-glutamyl transferase, waist-hip ratio, HDL cholesterol, age, urate. Our achieved Curve 0.9 prediction, reduced 10-feature achieving 0.88. Easily measurable biological surpassed traditional risk like diet, physical activity, socioeconomic status in predicting Furthermore, high prediction accuracy could be maintained just top 10 factors, additional ones offering marginal improvements. These findings underscore significance markers prediction.

Language: Английский

---

Recent applications of machine learning and deep learning models in the prediction, diagnosis, and management of diabetes: a comprehensive review

Diabetology & Metabolic Syndrome, Journal Year: 2022, Volume and Issue: 14(1)

Published: Dec. 27, 2022

Abstract Diabetes as a metabolic illness can be characterized by increased amounts of blood glucose. This abnormal increase lead to critical detriment the other organs such kidneys, eyes, heart, nerves, and vessels. Therefore, its prediction, prognosis, management are essential prevent harmful effects also recommend more useful treatments. For these goals, machine learning algorithms have found considerable attention been developed successfully. review surveys recently proposed (ML) deep (DL) models for objectives mentioned earlier. The reported results disclose that ML DL promising approaches controlling glucose diabetes. However, they should improved employed in large datasets affirm their applicability.

Language: Английский

---

Machine learning for diabetes clinical decision support: a review

Advances in Computational Intelligence, Journal Year: 2022, Volume and Issue: 2(2)

Published: April 1, 2022

Language: Английский

---

A robust predictive diagnosis model for diabetes mellitus using Shapley-incorporated machine learning algorithms

Healthcare Analytics, Journal Year: 2023, Volume and Issue: 3, P. 100166 - 100166

Published: March 28, 2023

Language: Английский

---

Development and validation of an insulin resistance model for a population without diabetes mellitus and its clinical implication: a prospective cohort study

EClinicalMedicine, Journal Year: 2023, Volume and Issue: 58, P. 101934 - 101934

Published: April 1, 2023

Insulin resistance (IR) is associated with diabetes mellitus, cardiovascular disease (CV), and mortality. Few studies have used machine learning to predict IR in the non-diabetic population.In this prospective cohort study, we trained a predictive model for populations using US National Health Nutrition Examination Survey (NHANES, from JAN 01, 1999 DEC 31, 2012) database Taiwan MAJOR (from 2008 2017) database. We analysed participants NHANES were excluded if they aged <18 years old, had incomplete laboratory data, or DM. To investigate clinical implications (CV all-cause mortality) of model, tested it biobank (TWB) 10, NOV 30, 2018. then SHapley Additive exPlanation (SHAP) values explain differences across models.Of all (combined MJ databases), randomly selected 14,705 training group, 4018 validation group. In their areas under curve (AUC) >0.8 (highest being XGboost, 0.87). test AUC also >0.80 0.88). Among 9 features (age, gender, race, body mass index, fasting plasma glucose (FPG), glycohemoglobin, triglyceride, total cholesterol high-density cholesterol), BMI highest value feature importance on (0.43 XGboost 0.47 RF algorithms). All TWB separated into group non-IR according algorithm. The Kaplan-Meier survival showed significant difference between groups (p < 0.0001 CV mortality, p = 0.0006 mortality). Therefore, has clear predicting IR, aside mortality.To patients high accuracy, only easily obtained are needed prediction accuracy our model. Similarly, predicts significantly higher can be applied both Asian Caucasian practice.Taichung Veterans General Hospital, Japan Society Promotion Science KAKENHI Grant Number JP21KK0293.

Language: Английский

---

Machine Learning as a Support for the Diagnosis of Type 2 Diabetes

International Journal of Molecular Sciences, Journal Year: 2023, Volume and Issue: 24(7), P. 6775 - 6775

Published: April 5, 2023

Diabetes is a chronic, metabolic disease characterized by high blood sugar levels. Among the main types of diabetes, type 2 most common. Early diagnosis and treatment can prevent or delay onset complications. Previous studies examined application machine learning techniques for prediction pathology, here an artificial neural network shows very promising results as possible valuable aid in management prevention diabetes. Additionally, its superior ability long-term predictions makes it ideal choice this field study. We utilized methods to uncover previously undiscovered associations between individual's health status development with goal accurately predicting determining risk level. Our study employed binary classifier, trained on scratch, identify potential nonlinear relationships diabetes set parameters obtained from patient measurements. Three datasets were utilized, i.e., National Center Health Statistics' (NHANES) biennial survey, MIMIC-III MIMIC-IV. These then combined create single dataset same number individuals without Since was balanced, primary evaluation metric model accuracy. The outcomes encouraging, achieving accuracy levels up 86% ROC AUC value 0.934. Further investigation needed improve reliability considering multiple measurements over time.

Language: Английский

---

A Novel Proposal for Deep Learning-Based Diabetes Prediction: Converting Clinical Data to Image Data

Diagnostics, Journal Year: 2023, Volume and Issue: 13(4), P. 796 - 796

Published: Feb. 20, 2023

Diabetes, one of the most common diseases worldwide, has become an increasingly global threat to humans in recent years. However, early detection diabetes greatly inhibits progression disease. This study proposes a new method based on deep learning for diabetes. Like many other medical data, PIMA dataset used contains only numerical values. In this sense, application popular convolutional neural network (CNN) models such data are limited. converts into images feature importance use robust representation CNN diagnosis. Three different classification strategies then applied resulting image data. first, fed ResNet18 and ResNet50 models. second, features ResNet fused classified with support vector machines (SVM). last approach, selected fusion by SVM. The results demonstrate robustness diagnosis

Language: Английский

---

Directive Explanations for Monitoring the Risk of Diabetes Onset: Introducing Directive Data-Centric Explanations and Combinations to Support What-If Explorations

Published: March 27, 2023

Explainable artificial intelligence is increasingly used in machine learning (ML) based decision-making systems healthcare. However, little research has compared the utility of different explanation methods guiding healthcare experts for patient care. Moreover, it unclear how useful, understandable, actionable and trustworthy these are experts, as they often require technical ML knowledge. This paper presents an dashboard that predicts risk diabetes onset explains those predictions with data-centric, feature-importance, example-based explanations. We designed interactive to assist such nurses physicians, monitoring recommending measures minimize risk. conducted a qualitative study 11 mixed-methods 45 51 diabetic patients compare our terms understandability, usefulness, actionability, trust. Results indicate participants preferred representation data-centric explanations provide local global overview over other methods. Therefore, this highlights importance visually directive method assisting gain insights from health records. Furthermore, we share design implications tailoring visual experts.

Language: Английский

---

Predictive value of machine learning for the progression of gestational diabetes mellitus to type 2 diabetes: a systematic review and meta-analysis

BMC Medical Informatics and Decision Making, Journal Year: 2025, Volume and Issue: 25(1)

Published: Jan. 13, 2025

This systematic review aims to explore the early predictive value of machine learning (ML) models for progression gestational diabetes mellitus (GDM) type 2 (T2DM). A comprehensive and search was conducted in Pubmed, Cochrane, Embase, Web Science up July 02, 2024. The quality studies included assessed. risk bias assessed through prediction model assessment tool a graph drawn accordingly. meta-analysis performed using Stata15.0. total 13 were present review, involving 11,320 GDM patients 22 ML models. showed pooled C-statistic 0.82 (95% CI: 0.79 ~ 0.86), sensitivity 0.76 (0.72 0.80), specificity 0.57 (0.50 0.65). has favorable diagnostic accuracy T2DM. provides evidence development tools with broader applicability.

Language: Английский

---