Detecting B-cell lymphoma-6 overexpression status in primary central nervous system lymphoma using multiparametric MRI-based machine learning DOI Creative Commons

Mingxiao Wang,

Guoli Liu, Nan Zhang

et al.

Neuroradiology, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 24, 2025

In primary central nervous system lymphoma (PCNSL), B-cell lymphoma-6 (BCL-6) is an unfavorable prognostic biomarker. We aim to non-invasively detect BCL-6 overexpression in PCNSL patients using multiparametric MRI and machine learning techniques. 65 (101 lesions) with (PCNSL) diagnosed from January 2013 July 2023, all were randomly divided into a training set validation according ratio of 8 2. ADC map derived DWI (b = 0/1000 s/mm2), fast spin echo T2WI, T2FLAIR, collected at 3.0 T. A total 2234 radiomics features the tumor segmentation area extracted LASSO used select features. Logistic regression (LR), Naive bayes (NB), Support vector (SVM), K-nearest Neighbor, (KNN) Multilayer Perceptron (MLP), for learning, sensitivity, specificity, accuracy F1-score, under curve (AUC) was evaluate detection performance five classifiers, 6 groups combinations different sequences fitted 5 optimal classifier obtained. status could be identified varying degrees 30 models based on radiomics, model improved by combining classifiers. (SVM) combined three sequence group had largest AUC (0.95) satisfactory (0.87) set. Multiparametric promising detecting overexpression.

Language: Английский

MRI-based brain tumor detection using convolutional deep learning methods and chosen machine learning techniques DOI Creative Commons
Soheila Saeedi, Sorayya Rezayi, Hamidreza Keshavarz

et al.

BMC Medical Informatics and Decision Making, Journal Year: 2023, Volume and Issue: 23(1)

Published: Jan. 23, 2023

Detecting brain tumors in their early stages is crucial. Brain are classified by biopsy, which can only be performed through definitive surgery. Computational intelligence-oriented techniques help physicians identify and classify tumors. Herein, we proposed two deep learning methods several machine approaches for diagnosing three types of tumor, i.e., glioma, meningioma, pituitary gland tumors, as well healthy brains without using magnetic resonance images to enable detect with high accuracy stages.

Language: Английский

Citations

238
Machine Learning Models for the Identification of Prognostic and Predictive Cancer Biomarkers: A Systematic Review DOI Open Access
Qasem Al-Tashi, Maliazurina Saad, Amgad Muneer

et al.

International Journal of Molecular Sciences, Journal Year: 2023, Volume and Issue: 24(9), P. 7781 - 7781

Published: April 24, 2023

The identification of biomarkers plays a crucial role in personalized medicine, both the clinical and research settings. However, contrast between predictive prognostic can be challenging due to overlap two. A biomarker predicts future outcome cancer, regardless treatment, effectiveness therapeutic intervention. Misclassifying as (or vice versa) have serious financial personal consequences for patients. To address this issue, various statistical machine learning approaches been developed. aim study is present an in-depth analysis recent advancements, trends, challenges, prospects identification. systematic search was conducted using PubMed identify relevant studies published 2017 2023. selected were analyzed better understand concept identification, evaluate methods, assess level activity, highlight application these methods cancer treatment. Furthermore, existing obstacles concerns are discussed prospective areas. We believe that review will serve valuable resource researchers, providing insights into used discovery identifying opportunities.

Language: Английский

Citations

70
A Novel Hybrid Dynamic Harris Hawks Optimized Gated Recurrent Unit Approach for Breast Cancer Prediction DOI Creative Commons

Rajesh Natarajan,

Sujatha Krishna,

H. L. Gururaj

et al.

International Journal of Computational Intelligence Systems, Journal Year: 2025, Volume and Issue: 18(1)

Published: Jan. 14, 2025

Language: Английский

Citations

2
Multi-Parametric MRI-Based Radiomics Models for Predicting Molecular Subtype and Androgen Receptor Expression in Breast Cancer DOI Creative Commons
Yühong Huang,

Lihong Wei,

Yalan Hu

et al.

Frontiers in Oncology, Journal Year: 2021, Volume and Issue: 11

Published: Aug. 18, 2021

To investigate whether radiomics features extracted from multi-parametric MRI combining machine learning approach can predict molecular subtype and androgen receptor (AR) expression of breast cancer in a non-invasive way.Patients diagnosed with clinical T2-4 stage March 2016 to July 2020 were retrospectively enrolled. The subtypes AR pre-treatment biopsy specimens assessed. A total 4,198 the pre-biopsy (including dynamic contrast-enhancement T1-weighted images, fat-suppressed T2-weighted apparent diffusion coefficient map) each patient. We applied several feature selection strategies including least absolute shrinkage operator (LASSO), recursive elimination (RFE), maximum relevance minimum redundancy (mRMR), Boruta Pearson correlation analysis, select most optimal features. then built 120 diagnostic models using distinct classification algorithms sets divided by sequences testing dataset leave-one-out cross-validation (LOOCV). performances binary assessed via area under receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative (NPV). And multiclass AUC, overall precision, recall rate, F1-score.A 162 patients (mean age, 46.91 ± 10.08 years) enrolled this study; 30 low-AR 132 high-AR expression. HR+/HER2- cancers 56 cases (34.6%), HER2+ 81 (50.0%), TNBC 25 (15.4%). There was no significant difference clinicopathologic characteristics between groups (P > 0.05), except menopausal status, ER, PR, HER2, Ki-67 index = 0.043, <0.001, 0.015, 0.006, respectively). No observed among three status <0.001 0.012, Multilayer Perceptron (MLP) showed best performance discriminating expression, an AUC 0.907 accuracy 85.8% dataset. highest obtained for vs. non-TNBC (AUC: 0.965, accuracy: 92.6%), HER2- 0.840, 79.0%), others 0.860, 82.1%) MLP as well. micro-AUC model 0.896, 0.735.Multi-parametric MRI-based approaches provide promising method non-invasively.

Language: Английский

Citations

61
Breast cancer diagnosis in an early stage using novel deep learning with hybrid optimization technique DOI Open Access
Kranti Kumar Dewangan, Deepak Kumar Dewangan, Satya Prakash Sahu

et al.

Multimedia Tools and Applications, Journal Year: 2022, Volume and Issue: 81(10), P. 13935 - 13960

Published: Feb. 25, 2022

Language: Английский

Citations

46
Predicting breast cancer types on and beyond molecular level in a multi-modal fashion DOI Creative Commons
Tianyu Zhang, Tao Tan, Luyi Han

et al.

npj Breast Cancer, Journal Year: 2023, Volume and Issue: 9(1)

Published: March 22, 2023

Accurately determining the molecular subtypes of breast cancer is important for prognosis patients and can guide treatment selection. In this study, we develop a deep learning-based model predicting directly from diagnostic mammography ultrasound images. Multi-modal learning with intra- inter-modality attention modules (MDL-IIA) proposed to extract relations between task. MDL-IIA leads best performance compared other cohort models in 4-category Matthews correlation coefficient (MCC) 0.837 (95% confidence interval [CI]: 0.803, 0.870). The also discriminate Luminal Non-Luminal disease an area under receiver operating characteristic curve 0.929 CI: 0.903, 0.951). These results significantly outperform clinicians' predictions based on radiographic imaging. Beyond molecular-level test, gene-level ground truth, our method bypass inherent uncertainty immunohistochemistry test. This work thus provides noninvasive predict cancer, potentially guiding selection providing decision support clinicians.

Language: Английский

Citations

35
A robust and efficient AI assistant for breast tumor segmentation from DCE-MRI via a spatial-temporal framework DOI Creative Commons
Jiadong Zhang, Zhiming Cui, Zhenwei Shi

et al.

Patterns, Journal Year: 2023, Volume and Issue: 4(9), P. 100826 - 100826

Published: Aug. 16, 2023

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) allows screening, follow up, and diagnosis for breast tumor with high sensitivity. Accurate segmentation from DCE-MRI can provide crucial information of location shape, which significantly influences the downstream clinical decisions. In this paper, we aim to develop an artificial intelligence (AI) assistant automatically segment tumors by capturing dynamic changes in multi-phase a spatial-temporal framework. The main advantages our AI include (1) robustness, i.e., model handle MR data different phase numbers intervals, as demonstrated on large-scale dataset seven medical centers, (2) efficiency, reduces time required manual annotation factor 20, while maintaining accuracy comparable that physicians. More importantly, fundamental step build AI-assisted cancer system, will promote application more diagnostic practices regarding cancer.

Language: Английский

Citations

29
Machine learning algorithms reveal potential miRNAs biomarkers in gastric cancer DOI Creative Commons
Hanieh Azari, Elham Nazari,

Reza Mohit

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: April 15, 2023

Abstract Gastric cancer is the high mortality rate cancers globally, and current survival 30% even with use of combination therapies. Recently, mounting evidence indicates potential role miRNAs in diagnosis assessing prognosis cancers. In state-of-art research cancer, machine-learning (ML) has gained increasing attention to find clinically useful biomarkers. The present study aimed identify diagnostic prognostic GC application ML. Using TCGA database ML algorithms such as Support Vector Machine (SVM), Random Forest, k-NN, etc., a panel 29 was obtained. Among algorithms, SVM chosen (AUC:88.5%, Accuracy:93% GC). To common molecular mechanisms miRNAs, their gene targets were predicted using online databases miRWalk, miRDB, Targetscan. Functional enrichment analyzes performed Gene Ontology (GO) Kyoto Database Genes Genomes (KEGG), well identification protein–protein interactions (PPI) STRING database. Pathway analysis target genes revealed involvement several cancer-related pathways including miRNA mediated inhibition translation, regulation expression by genetic imprinting, Wnt signaling pathway. Survival ROC curve showed that levels hsa-miR-21, hsa-miR-133a, hsa-miR-146b, hsa-miR-29c associated higher potentially earlier detection patients. A dysregulated may serve reliable biomarkers for gastric identified machine learning, which represents powerful tool biomarker identification.

Language: Английский

Citations

28
Radiomics and artificial intelligence in breast imaging: a survey DOI
Tianyu Zhang, Tao Tan, Riccardo Samperna

et al.

Artificial Intelligence Review, Journal Year: 2023, Volume and Issue: 56(S1), P. 857 - 892

Published: July 8, 2023

Language: Английский

Citations

23
Research progress on artificial intelligence technology-assisted diagnosis of thyroid diseases DOI Creative Commons
Lina Yang, Xinyuan Wang,

Shixia Zhang

et al.

Frontiers in Oncology, Journal Year: 2025, Volume and Issue: 15

Published: Feb. 20, 2025

With the rapid development of "Internet + Medical" model, artificial intelligence technology has been widely used in analysis medical images. Among them, using deep learning algorithms to identify features ultrasound and pathological images realize intelligent diagnosis diseases entered clinical verification stage. This study is based on application research reviews early screening thyroid diseases. The cure rate disease high stage, but once it deteriorates into cancer, risk death treatment costs patient increase. At present, still depends examination equipment experience doctors, there a certain misdiagnosis rate. Based above background, particularly important explore that can achieve objective lesions stages. paper provides comprehensive review recent technology. It integrates findings multiple studies traditional machine are as objects. convolutional neural network model recognition accuracy for nodules cell lesions. U-Net significantly improve nodule when segmentation algorithm. article focuses reviewing sections, hoping provide researchers with ideas help clinicians cancer.

Language: Английский

Citations

1