PD_EBM: An Integrated Boosting Approach Based on Selective Features for Unveiling Parkinson's Disease Diagnosis With Global and Local Explanations

Engineering Reports, Journal Year: 2025, Volume and Issue: 7(1)

Published: Jan. 1, 2025

ABSTRACT Early detection and characterization are crucial for treating managing Parkinson's disease (PD). The increasing prevalence of PD its significant impact on the motor neurons brain impose a substantial burden healthcare system. Early‐stage is vital improving patient outcomes reducing costs. This study introduces an ensemble boosting machine, termed PD_EBM, PD. PD_EBM leverages machine learning (ML) algorithms hybrid feature selection approach to enhance diagnostic accuracy. While ML has shown promise in medical applications detection, interpretability these models remains challenge. Explainable (XML) addresses this by providing transparency clarity model predictions. Techniques such as Local Interpretable Model‐agnostic Explanations (LIME) SHapley Additive exPlanations (SHAP) have become popular interpreting models. Our experiment used dataset 195 clinical records patients from University California Irvine (UCI) Machine Learning repository. Comprehensive data preparation included encoding categorical features, imputing missing values, removing outliers, addressing imbalance, scaling data, selecting relevant so on. We propose framework that focuses most important features prediction. employs Decision Tree (DT) classifier with AdaBoost, followed linear discriminant analysis (LDA) optimizer, achieving impressive accuracy 99.44%, outperforming other

Language: Английский

---

Decoding the black box: Explainable AI (XAI) for cancer diagnosis, prognosis, and treatment planning-A state-of-the art systematic review

International Journal of Medical Informatics, Journal Year: 2024, Volume and Issue: 193, P. 105689 - 105689

Published: Nov. 4, 2024

Language: Английский

---

Classifying Schizophrenia Using Functional MRI and Investigating Underlying Functional Phenomena

Brain Research Bulletin, Journal Year: 2025, Volume and Issue: 223, P. 111279 - 111279

Published: March 7, 2025

BACKGROUND：: Existing studies have revealed functional abnormalities in certain brain regions of patients with schizophrenia (SZ), but the relationships between these and their impact on disease progression remain unclear. Fifty-six SZ 56 healthy controls were included. Based resting-state magnetic resonance imaging, we analyzed fractional amplitude low-frequency fluctuations (fALFF), regional homogeneity (ReHo), degree centrality (DC). Statistically significant metrics selected as features, machine learning models used to distinguish controls. Analyze importance features optimal model. The Louvain community detection algorithm structural equation modeling investigate potential causal effects. average prediction accuracy various ML classifiers reached 0.9241 by fALFF, ReHo, DC values. SVM model highest performance an 0.9464. Abnormal ReHo right middle frontal gyrus contributed most this classifier participated direct SZ. All ultimately constituted two clusters (FClus), which exhibit internal influences. FClus1 had a positive influence SZ, cascade starting from abnormal fALFF inferior temporal gyrus. FClus2 negative left fusiform gyrus.Abnormal caudate nucleus, angular gyrus, lentiform nucleus do not disease. We identified interactions among within FClus that potentially onset schizophrenia, including epicenter phenomenon FClus, for inhibiting function without impact. Additionally, believe contribution classification may indicate size disease, necessarily process.

Language: Английский

---

AI-Derived Blood Biomarkers for Ovarian Cancer Diagnosis: Systematic Review and Meta-Analysis

Journal of Medical Internet Research, Journal Year: 2025, Volume and Issue: 27, P. e67922 - e67922

Published: March 24, 2025

Background Emerging evidence underscores the potential application of artificial intelligence (AI) in discovering noninvasive blood biomarkers. However, diagnostic value AI-derived biomarkers for ovarian cancer (OC) remains inconsistent. Objective We aimed to evaluate research quality and validity AI-based OC diagnosis. Methods A systematic search was performed MEDLINE, Embase, IEEE Xplore, PubMed, Web Science, Cochrane Library databases. Studies examining accuracy AI were identified. The risk bias assessed using Quality Assessment Diagnostic Accuracy Studies–AI tool. Pooled sensitivity, specificity, area under curve (AUC) estimated a bivariate model meta-analysis. Results total 40 studies ultimately included. Most (n=31, 78%) included evaluated as low bias. Overall, pooled AUC 85% (95% CI 83%-87%), 91% 90%-92%), 0.95 0.92-0.96), respectively. For contingency tables with highest accuracy, 95% 90%-97%), 97% 95%-98%), 0.99 0.98-1.00), Stratification by algorithms revealed higher sensitivity specificity machine learning (sensitivity=85% specificity=92%) compared those deep (sensitivity=77% specificity=85%). In addition, serum reported substantially (94%) (96%) than plasma (sensitivity=83% specificity=91%). external validation demonstrated significantly (specificity=94%) without (specificity=89%), while reverse observed (74% vs 90%). No publication detected this Conclusions demonstrate satisfactory performance diagnosis are anticipated become an effective modality future, potentially avoiding unnecessary surgeries. Future is warranted incorporate into models, well prioritize adoption methodologies. Trial Registration PROSPERO CRD42023481232; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023481232

Language: Английский

---

Explainable AI-based feature importance analysis for ovarian cancer classification with ensemble methods

Frontiers in Public Health, Journal Year: 2025, Volume and Issue: 13

Published: March 26, 2025

Introduction Ovarian Cancer (OC) is one of the leading causes cancer deaths among women. Despite recent advances in medical field, such as surgery, chemotherapy, and radiotherapy interventions, there are only marginal improvements diagnosis OC using clinical parameters, symptoms very non-specific at early stage. Owing to computational algorithms, ensemble machine learning, it now possible identify complex patterns parameters. However, these do not provide deeper insights into prediction diagnosis. Explainable artificial intelligence (XAI) models, LIME SHAP Kernels, can decision-making process thus increasing their applicability. Methods The main aim this study design a computer-aided diagnostic system that accurately classifies detects ovarian cancer. To achieve objective, three-stage model game-theoretic approach based on values were built evaluate visualize results, analyzing important features responsible for prediction. Results Discussion results demonstrate efficacy proposed with an accuracy 98.66%. model’s consistency advantages compared single classifiers. validated conventional statistical methods p -test Cohen’s d highlight method. further validate ranking features, we -values top five bottom features. AI-based method detection, diagnosis, prognosis multi-modal real-life data, which mimics move clinician demonstration high performance. strategy lead reliable, accurate, consistent AI solutions detection management higher patient experience outcomes low cost, morbidity, mortality. This be beneficial millions women living resource-constrained challenging economies.

Language: Английский

---

A CECT-Based Radiomics Nomogram Predicts the Overall Survival of Patients with Hepatocellular Carcinoma After Surgical Resection

Biomedicines, Journal Year: 2025, Volume and Issue: 13(5), P. 1237 - 1237

Published: May 19, 2025

Objective: The primary objective of this study was to develop and validate a predictive nomogram that integrates radiomic features derived from contrast-enhanced computed tomography (CECT) images with clinical variables predict overall survival (OS) in patients hepatocellular carcinoma (HCC) after surgical resection. Methods: This retrospective analyzed the preoperative enhanced CT data 202 HCC who underwent resection at Affiliated Hospital North Sichuan Medical College (Institution 1) June 2017 2021 Nanchong Central 2) 2020 2022. Among these patients, 162 Institution 1 were randomly divided into training cohort (112 patients) an internal validation (50 7:3 ratio, whereas 40 2 assigned as independent external cohort. Univariate multivariate Cox proportional hazards regression analyses performed identify risk factors associated OS Using 3D-Slicer software, tumor lesions manually delineated slice by on non-contrast-enhanced (NCE) CT, arterial phase (AP), portal venous (PVP) generate volumetric regions interest (VOIs). Radiomic subsequently extracted VOIs. LASSO analysis employed for dimensionality reduction feature selection, culminating construction signature (Radscore). models, including model, radiomic–clinical developed prediction. performance models assessed via concordance index (C-index) time–ROC curves. optimal model further visualized nomogram, its accuracy evaluated calibration curves decision curve (DCA). Finally, interpreted Shapley additive explanations (SHAP). Results: revealed BCLC stage, albumin–bilirubin (ALBI), NLR–PLR score predictors three exhibited highest performance, C-indices 0.789, 0.726, 0.764 training, cohorts, respectively. Furthermore, showed 1-year 3-year AUCs 0.837 0.845 cohort, 0.801 0.880 0.773 0.840 Calibration DCA demonstrated model’s excellent applicability. Conclusions: combining CECT provides accurate prediction is beneficial clinicians developing individualized treatment strategies HCC.

Language: Английский

---

Classification of Ovarian Cancer with Multimodal Data Using AI Technology– A Review

Published: Oct. 4, 2024

Language: Английский

---

AI-Derived Blood Biomarkers for Ovarian Cancer Diagnosis: Systematic Review and Meta-Analysis (Preprint)

Published: Oct. 24, 2024

Emerging evidence underscores the potential application of artificial intelligence (AI) in discovering noninvasive blood biomarkers. However, diagnostic value AI-derived biomarkers for ovarian cancer (OC) remains inconsistent. We aimed to evaluate research quality and validity AI-based OC diagnosis. A systematic search was performed MEDLINE, Embase, IEEE Xplore, PubMed, Web Science, Cochrane Library databases. Studies examining accuracy AI were identified. The risk bias assessed using Quality Assessment Diagnostic Accuracy Studies–AI tool. Pooled sensitivity, specificity, area under curve (AUC) estimated a bivariate model meta-analysis. total 40 studies ultimately included. Most (n=31, 78%) included evaluated as low bias. Overall, pooled AUC 85% (95% CI 83%-87%), 91% 90%-92%), 0.95 0.92-0.96), respectively. For contingency tables with highest accuracy, 95% 90%-97%), 97% 95%-98%), 0.99 0.98-1.00), Stratification by algorithms revealed higher sensitivity specificity machine learning (sensitivity=85% specificity=92%) compared those deep (sensitivity=77% specificity=85%). In addition, serum reported substantially (94%) (96%) than plasma (sensitivity=83% specificity=91%). external validation demonstrated significantly (specificity=94%) without (specificity=89%), while reverse observed (74% vs 90%). No publication detected this demonstrate satisfactory performance diagnosis are anticipated become an effective modality future, potentially avoiding unnecessary surgeries. Future is warranted incorporate into models, well prioritize adoption methodologies. PROSPERO CRD42023481232; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023481232

Language: Английский

---