Prediction of diabetes disease using an ensemble of machine learning multi-classifier models

BMC Bioinformatics, Journal Year: 2023, Volume and Issue: 24(1)

Published: Sept. 12, 2023

Diabetes is a life-threatening chronic disease with growing global prevalence, necessitating early diagnosis and treatment to prevent severe complications. Machine learning has emerged as promising approach for diabetes diagnosis, but challenges such limited labeled data, frequent missing values, dataset imbalance hinder the development of accurate prediction models. Therefore, novel framework required address these improve performance.In this study, we propose an innovative pipeline-based multi-classification predict in three classes: diabetic, non-diabetic, prediabetes, using imbalanced Iraqi Patient Dataset Diabetes. Our incorporates various pre-processing techniques, including duplicate sample removal, attribute conversion, value imputation, data normalization standardization, feature selection, k-fold cross-validation. Furthermore, implement multiple machine models, k-NN, SVM, DT, RF, AdaBoost, GNB, introduce weighted ensemble based on Area Under Receiver Operating Characteristic Curve (AUC) imbalance. Performance optimization achieved through grid search Bayesian hyper-parameter tuning.Our proposed model outperforms other predicting diabetes. The achieves high average accuracy, precision, recall, F1-score, AUC values 0.9887, 0.9861, 0.9792, 0.9851, 0.999, respectively.Our demonstrates results accurately diabetic patients. addresses associated imbalance, leading improved performance. This study highlights potential techniques management, can serve valuable tool patient care. Further research build upon our work refine optimize explore its applicability diverse datasets populations.

Language: Английский

---

Applied Machine Learning Techniques to Diagnose Voice-Affecting Conditions and Disorders: Systematic Literature Review

Journal of Medical Internet Research, Journal Year: 2023, Volume and Issue: 25, P. e46105 - e46105

Published: May 23, 2023

Background Normal voice production depends on the synchronized cooperation of multiple physiological systems, which makes sensitive to changes. Any systematic, neurological, and aerodigestive distortion is prone affect through reduced cognitive, pulmonary, muscular functionality. This sensitivity inspired using as a biomarker examine disorders that voice. Technological improvements emerging machine learning (ML) technologies have enabled possibilities extracting digital vocal features from for automated diagnosis monitoring systems. Objective study aims summarize comprehensive view research voice-affecting uses ML techniques samples where systematic conditions, nonlaryngeal disorders, neurological are specifically interest. Methods literature review (SLR) investigated state art voice-based diagnostic systems with technologies, targeting without direct relation box point applied health technology. Through search string, studies published 2012 2022 databases Scopus, PubMed, Web Science were scanned collected assessment. To minimize bias, retrieval relevant references in other field was ensured, 2 authors assessed studies. Low-quality removed quality assessment data extracted summary tables analysis. The articles checked similarities between author groups prevent cumulative redundancy bias during screening process, only 1 article included same group. Results In analysis 145 studies, support vector machines most utilized technique (51/145, 35.2%), studied disease being Parkinson (PD; reported 87/145, 60%, studies). After 2017, 16 additional examined, contrast 3 previously. Furthermore, an upsurge use artificial neural network–based architectures observed after 2017. Almost half last years (2021 2022). A broad interest many countries observed. Notably, nearly one-half (n=75) relied 10 distinct sets, 11/145 (7.6%) used demographic input models. Conclusions SLR revealed considerable across diagnosing PD disorder. However, identified several gaps, including limited unbalanced set usage focus test rather than disorder-specific monitoring. Despite limitations constrained by peer-reviewed publications written English, provides valuable insights into current ML-based disorder highlighting areas address future research.

Language: Английский

---

Utilizing CNN-LSTM techniques for the enhancement of medical systems

Alexandria Engineering Journal, Journal Year: 2023, Volume and Issue: 72, P. 323 - 338

Published: April 15, 2023

Language: Английский

---

Machine learning approach for detecting Covid-19 from speech signal using Mel frequency magnitude coefficient

Signal Image and Video Processing, Journal Year: 2023, Volume and Issue: 17(6), P. 3155 - 3162

Published: March 25, 2023

Language: Английский

---

Extracting cancer concepts from clinical notes using natural language processing: a systematic review

BMC Bioinformatics, Journal Year: 2023, Volume and Issue: 24(1)

Published: Oct. 29, 2023

Abstract Background Extracting information from free texts using natural language processing (NLP) can save time and reduce the hassle of manually extracting large quantities data incredibly complex clinical notes cancer patients. This study aimed to systematically review studies that used NLP methods identify concepts automatically. Methods PubMed, Scopus, Web Science, Embase were searched for English papers a combination terms concerning “Cancer”, “NLP”, “Coding”, “Registries” until June 29, 2021. Two reviewers independently assessed eligibility inclusion in review. Results Most software programs concept extraction reported developed by researchers ( n = 7). Rule-based algorithms most frequently developing these programs. In articles, criteria accuracy 14) sensitivity 12) evaluate algorithms. addition, Systematized Nomenclature Medicine-Clinical Terms (SNOMED-CT) Unified Medical Language System (UMLS) commonly terminologies concepts. focused on breast 4, 19%) lung 19%). Conclusion The use symptoms has increased recent years. rule-based are well-liked developers. Due algorithms' high identifying concepts, we suggested future extract other diseases as well.

Language: Английский

---

Quantum-Inspired Heuristic Algorithm for Secure Healthcare Prediction Using Blockchain Technology

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2023, Volume and Issue: 28(6), P. 3371 - 3378

Published: Aug. 11, 2023

People's health is adversely affected by environmental changes and poor nutritional habits, emphasizing the importance of awareness. The healthcare system encounters significant challenges, including data insufficiency, threats, errors, delays. To address these issues advance medical care, we propose a secure prediction method, prioritizing patient privacy transmission efficiency. Quantum-inspired heuristic algorithm combined with Kril Herd Optimization (QKHO) introduced for prediction, along comparison to Deep Forward Neural Network (DFNN) optimized using Krill (KHO) Optimization. proposed QKHO model outperforms conventional models exhibits higher accuracy, precision, recall, F1-score. Blockchain technology ensures server, surpassing security level existing RSA Diffie-Hellman algorithms.

Language: Английский

---

Identifying metabolic dysfunction-associated steatotic liver disease in patients with hypertension and pre-hypertension: An interpretable machine learning approach

Digital Health, Journal Year: 2024, Volume and Issue: 10

Published: Jan. 1, 2024

Metabolic dysfunction-associated steatotic liver disease (MASLD) is one of the most prevalent diseases and associated with pre-hypertension hypertension. Our research aims to develop interpretable machine learning (ML) models accurately identify MASLD in hypertensive pre-hypertensive populations.

Language: Английский

---

Automated machine learning for predicting liver metastasis in patients with gastrointestinal stromal tumor: a SEER-based analysis

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

Published: May 30, 2024

Gastrointestinal stromal tumors (GISTs) are a rare type of tumor that can develop liver metastasis (LIM), significantly impacting the patient's prognosis. This study aimed to predict LIM in GIST patients by constructing machine learning (ML) algorithms assist clinicians decision-making process for treatment. Retrospective analysis was performed using Surveillance, Epidemiology, and End Results (SEER) database, cases from 2010 2015 were assigned developing sets, while 2016 2017 testing set. Missing values addressed multiple imputation technique. Four utilized construct models, comprising traditional logistic regression (LR) automated (AutoML) such as gradient boost (GBM), deep neural net (DL), generalized linear model (GLM). We evaluated models' performance LR-based metrics, including area under receiver operating characteristic curve (AUC), calibration curve, decision (DCA), well AutoML-based feature importance, SHapley Additive exPlanation (SHAP) Plots, Local Interpretable Model Agnostic Explanation (LIME). A total 6207 included this study, with 2683, 1780, 1744 allocated training, validation, test respectively. Among different models evaluated, GBM demonstrated highest cohorts, respective AUC 0.805, 0.780, 0.795. Furthermore, outperformed other AutoML terms accuracy, achieving 0.747, 0.700, 0.706 Additionally, revealed size location most significant predictors influencing model's ability accurately LIM. The utilizing algorithm effectively risk provide reference individualized treatment plans.

Language: Английский

---

Integration of machine learning and experimental validation to identify the prognostic signature related to diverse programmed cell deaths in breast cancer

Frontiers in Oncology, Journal Year: 2025, Volume and Issue: 14

Published: Jan. 6, 2025

Programmed cell death (PCD) is closely related to the occurrence, development, and treatment of breast cancer. The aim this study was investigate association between various programmed patterns prognosis cancer (BRCA) patients. levels 19 different deaths in were assessed by ssGSEA analysis, these PCD scores summed obtain PCDS for each sample. relationship with immune as well metabolism-related pathways explored. PCD-associated subtypes obtained unsupervised consensus clustering differentially expressed genes analyzed. prognostic signature (PCDRS) constructed best combination 101 machine learning algorithm combinations, C-index PCDRS compared 30 published signatures. In addition, we analyzed relation therapeutic responses. distribution cells explored single-cell analysis spatial transcriptome analysis. Potential drugs targeting key Cmap. Finally, expression clinical tissues verified RT-PCR. showed higher normal. Different groups significant differences pathways. PCDRS, consisting seven genes, robust predictive ability over other signatures datasets. high group had a poorer strongly associated cancer-promoting tumor microenvironment. low exhibited anti-cancer immunity responded better checkpoint inhibitors chemotherapy-related drugs. Clofibrate imatinib could serve potential small-molecule complexes SLC7A5 BCL2A1, respectively. mRNA upregulated tissues. can be used biomarker assess response BRCA patients, which offers novel insights monitoring personalization

Language: Английский

---

Peritoneal cytology predicting distant metastasis in uterine carcinosarcoma: machine learning model development and validation

World Journal of Surgical Oncology, Journal Year: 2025, Volume and Issue: 23(1)

Published: April 26, 2025

This study develops and validates a machine learning model using peritoneal cytology to predict distant metastasis in uterine carcinosarcoma, aiding clinical decision-making. utilized detailed data findings from carcinosarcoma patients the SEER database. Eight algorithms-Logistic Regression, SVM, GBM, Neural Network, RandomForest, KNN, AdaBoost, LightGBM-were applied metastasis. Model performance was assessed AUC, calibration curves, DCA, confusion matrices, sensitivity, specificity. The Logistic Regression visualized with nomogram, its results were analyzed. SHAP values used interpret best-performing model. Peritoneal cytology, T stage, age, tumor size key factors influencing patients. had significant weight prediction models. logistic regression demonstrated excellent predictive an AUC of 0.882 training set 0.881 internal test set. interpreted nomogram. In comprehensive evaluations, GBM identified as explained values. Additionally, DCA curves indicated that both models have potential utility. introduces first effective tool for predicting by integrating features into construction. It aids early identification high-risk patients, enhancing follow-up monitoring during development, supports optimization personalized treatment strategies.

Language: Английский

---