Machine learning approaches for neurological disease prediction: A systematic review DOI
Ana Fatima, Sarfaraz Masood

Expert Systems, Journal Year: 2024, Volume and Issue: unknown

Published: April 4, 2024

Abstract In this article, we present a systematic and exhaustive review regarding the trends, datasets employed, as well findings achieved in last 11 years neurological disorder prediction using machine learning models. work comparison between biomarkers used ML field with that are obtained through other non‐ml‐based research fields. This will help identifying potential gaps for domain. As study of disorders is far‐reaching task due to wide variety diseases, hence scope restricted three most prevalent is, Alzheimer's, Parkinson's, Autism Spectrum Disorder (ASD). From our analysis, it has been found over time deep techniques especially Convolutional Neural Networks have proved be beneficial disease task. For reason, Magnetic Resonance Imaging popular modality across all considered diseases. It also notable employment transfer approach maintenance global data centre helps dealing scarcity problems model training. The manuscript discusses challenges future field. To best knowledge, unlike studies, attempts put forth conclusion every article discussed highlighting salient aspects major studies particular problem.

Language: Английский

A Pseudo-Siamese Feature Fusion Generative Adversarial Network for Synthesizing High-Quality Fetal Four-Chamber Views DOI
Sibo Qiao, Silin Pan, Gang Luo

et al.

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2022, Volume and Issue: 27(3), P. 1193 - 1204

Published: Jan. 14, 2022

Four-chamber (FC) views are the primary ultrasound(US) images that cardiologists diagnose whether fetus has congenital heart disease (CHD) in prenatal diagnosis and screening. FC intuitively depict developmental morphology of fetal heart. Early CHD always been focus difficulty Furthermore, deep learning technology achieved great success medical image analysis. Hence, applying early screening helps improve diagnostic accuracy. However, lack large-scale high-quality brings incredible difficulties to models or cardiologists. we propose a Pseudo-Siamese Feature Fusion Generative Adversarial Network (PSFFGAN), synthesizing using sketch images. In addition, novel Triplet Loss Function (TGALF), which optimizes PSFFGAN fully extract cardiac anatomical structure information provided by synthesize corresponding with speckle noises, artifacts, other ultrasonic characteristics. The experimental results show synthesized our proposed have best objective evaluation values: SSIM 0.4627, MS-SSIM 0.6224, FID 83.92, respectively. More importantly, two professional evaluate healthy PSFFGAN, giving subjective score average qualified rate is 82% 79%, respectively, further proves effectiveness PSFFGAN.

Language: Английский

Citations

21
A Deep Learning Approach for the Estimation of Glomerular Filtration Rate DOI
Haishuai Wang, Benjamin Bowe,

Zhicheng Cui

et al.

IEEE Transactions on NanoBioscience, Journal Year: 2022, Volume and Issue: 21(4), P. 560 - 569

Published: Jan. 31, 2022

An accurate estimation of glomerular filtration rate (GFR) is clinically crucial for kidney disease diagnosis and predicting the prognosis chronic (CKD). Machine learning methodologies such as deep neural networks provide a potential avenue increasing accuracy in GFR estimation. We developed novel architecture, shallow network, to estimate (dlGFR short) examined its comparative performance with estimated from Modification Diet Renal Disease (MDRD) Chronic Kidney Epidemiology Collaboration (CKD-EPI) equations. The dlGFR model jointly trains network enable both linear transformation input features log target, non-linear feature embedding stage function classification. validate proposed methods on data multiple studies obtained NIDDK Central Database Repository. predicted values within 30% measured 88.3% accuracy, compared 87.1% 84.7% achieved by CKD-EPI MDRD equations (p = 0.051 p < 0.001, respectively). Our results suggest that are superior resulting traditional statistical estimating rate. Based these results, an end-to-end predication system has been deployed facilitate use algorithm.

Language: Английский

Citations

21
Deep learning approach to predict autism spectrum disorder: a systematic review and meta-analysis DOI Creative Commons
Yang Ding, Heng Zhang, Ting Qiu

et al.

BMC Psychiatry, Journal Year: 2024, Volume and Issue: 24(1)

Published: Oct. 28, 2024

The use of the deep learning (DL) approach has been suggested or applied to identify childhood autism spectrum disorder (ASD). capacity predict ASD, however, differs across investigations. Our study's objective was conduct a meta-analysis determine DL for ASD in children's classification accuracy. Eligibility criteria were designed according purpose meta-analysis; PubMed, EMBASE, Cochrane Library, and Web Science Database searched articles published up April 16, 2023, on accuracy methods classification. Using Revised Tool Quality Assessment Diagnostic Accuracy Studies (QUADAS-2) assess quality included studies. Sensitivity, specificity, areas under curve (AUC), summary receiver operating characteristic (SROC), corresponding 95% confidence intervals (CIs) compiled by using bivariate random-effects models. A total 11 predictive trials based models included, involving 9495 patients from 6 different databases. According models' results, overall sensitivity, AUC technique were, 0.95 (95% CI = 0.88–0.98), 0.93 0.85–0.97), 0.98 (95%CI: 0.97–0.99), respectively. Subgroup analysis results found that datasets did not cause heterogeneity (meta-regression P 0.55). Kaggle dataset's sensitivity specificity 0.94 0.82-1.00) 0.91 0.76-1.00), with 0.97 0.92-1.00) ABIDE dataset. techniques satisfactory However, major studies limited effectiveness this meta-analysis. Further need be performed demonstrate clinical practicability diagnosis.

Language: Английский

Citations

4
FLDS: An Intelligent Feature Learning Detection System for Visualizing Medical Images Supporting Fetal Four-Chamber Views DOI
Sibo Qiao, Shanchen Pang, Gang Luo

et al.

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2021, Volume and Issue: 26(10), P. 4814 - 4825

Published: June 22, 2021

Fetal congenital heart disease (CHD) is the most common type of fatal malformation. four-chamber (FC) view a significant and easily accessible ultrasound (US) image among fetal echocardiography images. Automatic detection four chambers considerably contributes to early diagnosis CHD. Furthermore, robust discriminative features are essential for detecting crucial visualizing medical images, especially FC views. However, it an incredibly challenging task due several key factors, such as numerous speckles in US with small size unfixed positions, category confusion caused by similarity cardiac chambers. These factors hinder process capturing features, hence destroying chambers' precise detection. Therefore, we propose intelligent feature learning system (FLDS) views detect A multistage residual hybrid attention module (MRHAM) presented this paper incorporated FLDS powerful helping accurately locate Extensive experiments demonstrate that our proposed outperforms current state-of-the-art, including precision 0.919, recall 0.971, F1 score 0.944, mAP 0.953, frames per second (FPS) 43. In addition, also validated on other nature images PASCAL VOC dataset, achieving higher 0.878 while input 608 × 608.

Language: Английский

Citations

27
Detecting autism in children through drawing characteristics using the visual-motor integration test DOI

P.S. Chen,

Jasin Wong, Eva E. Chen

et al.

Health Information Science and Systems, Journal Year: 2025, Volume and Issue: 13(1)

Published: Jan. 26, 2025

Language: Английский

Citations

0
Machine-Learning Model for Genetic Disorder Prediction DOI

Vanashree Agnihotri,

Prathamesh Hire,

Vaishali Ingle

et al.

Lecture notes in networks and systems, Journal Year: 2025, Volume and Issue: unknown, P. 133 - 144

Published: Jan. 1, 2025

Language: Английский

Citations

0
Exploring Filipino Medical Students’ Attitudes and Perceptions of Artificial Intelligence in Medical Education: A Mixed-Methods Study DOI Creative Commons
Robbi Miguel G. Falcon, Renne Margaret U. Alcazar,

Hannah G Babaran

et al.

MedEdPublish, Journal Year: 2025, Volume and Issue: 14, P. 282 - 282

Published: April 1, 2025

Artificial intelligence (AI) has many implications on the practice of medicine, especially for current medical students who have to consider impact AI information available patients and ethical aspects rendering healthcare as a whole. With fast pace development in healthcare, educators struggle incorporate curriculum. The generation will likely be first use tools their practice, hence this study aims investigate perceptions role education medicine using mixed methods parallel convergent design. findings revealed that had baseline understanding its but required further training practical use. Moreover, terms future (i.e., choice specialization, doctor-patient relationship) was evident must considered by order promote responsible physicians-in-training. In conclusion, from helped identify key areas focus integration into curriculum related both clinical practice.

Language: Английский

Citations

0
The Sensory Abnormality Mediated Partially the Efficacy of Repetitive Transcranial Magnetic Stimulation on Treating Comorbid Sleep Disorder in Autism Spectrum Disorder Children DOI Creative Commons
Lei Gao,

Chen Wang,

Xiao-rong Song

et al.

Frontiers in Psychiatry, Journal Year: 2022, Volume and Issue: 12

Published: Jan. 24, 2022

Sleep disorder emerges as a common comorbidity in children with autism spectrum (ASD), and the interaction between core symptoms of ASD its sleep remains unclear. Repetitive transcranial magnetic stimulation (rTMS) was used on bilateral dorsolateral prefrontal cortex (DLPFC) to investigate efficacy rTMS comorbid problems well mediation role intervention improvement. A total 41 Chinese who met criteria fifth edition American Diagnostic Statistical Manual Mental Disorders were recruited, 39 them (mean age: 9.0 ± 4.4 years old; male-female ratio 3.9: 1) completed study stimulating protocol high frequency left DLPFC low right DLPFC. They all assessed three times (before, at 4 weeks after, 8 after stimulation) by Children's Habits Questionnaire (CSHQ), Strengths Difficulties (SDQ), Childhood Autism Rating Scale, Behavior Questionnaire-2, Short Sensory Profile (SSP). The repeated-measures ANOVA showed that main effect "intervention time" CSHQ (F = 25.103, P < 0.001), SSP 6.345, 0.003), SDQ 9.975, 0.001) statistically significant. By Bayesian analysis, we only found score mediated treating (αβ 5.11 1.51, 95% CI: 2.50-8.41). percentage 37.94%. Our results indicated modulation for both autistic disturbances. sensory abnormality improvement ASD.

Language: Английский

Citations

13
Genomic Machine Learning Meta-regression: Insights on Associations of Study Features With Reported Model Performance DOI
Eric J. Barnett,

Daniel G. Onete,

Asif Salekin

et al.

IEEE/ACM Transactions on Computational Biology and Bioinformatics, Journal Year: 2023, Volume and Issue: 21(1), P. 169 - 177

Published: Dec. 18, 2023

Many studies have been conducted with the goal of correctly predicting diagnostic status a disorder using combination genomic data and machine learning. It is often hard to judge which components study led better results whether reported represent true improvement or an uncorrected bias inflating performance. We extracted information about methods used other differentiating features in learning models. these linear regressions model tested for univariate multivariate associations as well interactions between features. Of models reviewed, 46% feature selection that can lead leakage. Across our models, number hyperparameter optimizations reported, leakage due selection, type, modeling autoimmune were significantly associated increase found significant, negative interaction training size. Our suggest susceptible are prevalent among research, resulting inflated Best practice guidelines promote avoidance recognition may help field avoid biased results.

Language: Английский

Citations

8
DeepASD: a deep adversarial-regularized graph learning method for ASD diagnosis with multimodal data DOI Creative Commons
W.F. Chen, Jianjun Yang,

Zhongquan Sun

et al.

Translational Psychiatry, Journal Year: 2024, Volume and Issue: 14(1)

Published: Sept. 14, 2024

Language: Английский

Citations

2