Machine learning and new insights for breast cancer diagnosis DOI Creative Commons

Guo Ya,

Heng Zhang,

Leilei Yuan

et al.

Journal of International Medical Research, Journal Year: 2024, Volume and Issue: 52(4)

Published: April 1, 2024

Breast cancer (BC) is the most prominent form of among females all over world. The current methods BC detection include X-ray mammography, ultrasound, computed tomography, magnetic resonance imaging, positron emission tomography and breast thermographic techniques. More recently, machine learning (ML) tools have been increasingly employed in diagnostic medicine for its high efficiency intervention. subsequent imaging features mathematical analyses can then be used to generate ML models, which stratify, differentiate detect benign malignant lesions. Given marked advantages, radiomics a frequently tool recent research clinics. Artificial neural networks deep (DL) are novel forms that evaluate data using computer simulation human brain. DL directly processes unstructured information, such as images, sounds language, performs precise clinical image stratification, medical record tumour diagnosis. Herein, this review thoroughly summarizes prior investigations on application images intervention radiomics, namely ML. aim was provide guidance scientists regarding use artificial intelligence clinic.

Language: Английский

A Comprehensive Review on Synergy of Multi-Modal Data and AI Technologies in Medical Diagnosis DOI Creative Commons
Xi Xu, Jianqiang Li, Zhichao Zhu

et al.

Bioengineering, Journal Year: 2024, Volume and Issue: 11(3), P. 219 - 219

Published: Feb. 25, 2024

Disease diagnosis represents a critical and arduous endeavor within the medical field. Artificial intelligence (AI) techniques, spanning from machine learning deep to large model paradigms, stand poised significantly augment physicians in rendering more evidence-based decisions, thus presenting pioneering solution for clinical practice. Traditionally, amalgamation of diverse data modalities (e.g., image, text, speech, genetic data, physiological signals) is imperative facilitate comprehensive disease analysis, topic burgeoning interest among both researchers clinicians recent times. Hence, there exists pressing need synthesize latest strides multi-modal AI technologies realm diagnosis. In this paper, we narrow our focus five specific disorders (Alzheimer’s disease, breast cancer, depression, heart epilepsy), elucidating advanced endeavors their treatment through lens artificial intelligence. Our survey not only delineates detailed diagnostic methodologies across varying but also underscores commonly utilized public datasets, intricacies feature engineering, prevalent classification models, envisaged challenges future endeavors. essence, research contribute advancement methodologies, furnishing invaluable insights decision making.

Language: Английский

Citations

37
Development of a machine learning-based radiomics signature for estimating breast cancer TME phenotypes and predicting anti-PD-1/PD-L1 immunotherapy response DOI Creative Commons
Xiaorui Han, Yuan Guo,

Huifen Ye

et al.

Breast Cancer Research, Journal Year: 2024, Volume and Issue: 26(1)

Published: Jan. 29, 2024

Abstract Backgrounds Since breast cancer patients respond diversely to immunotherapy, there is an urgent need explore novel biomarkers precisely predict clinical responses and enhance therapeutic efficacy. The purpose of our present research was construct independently validate a biomarker tumor microenvironment (TME) phenotypes via machine learning-based radiomics way. interrelationship between the biomarker, TME recipients’ response also revealed. Methods In this retrospective multi-cohort investigation, five separate cohorts were recruited measure signature, which constructed validated by integrating RNA-seq data with DCE-MRI images for predicting immunotherapy response. Initially, we using 1089 in TCGA database. Then, parallel 94 obtained from TCIA applied develop radiomics-based signature random forest learning. repeatability then internal validation set. Two additional independent external sets analyzed reassess signature. Immune phenotype cohort ( n = 158) divided based on CD8 cell infiltration into immune-inflamed immune-desert phenotypes; these utilized examine relationship immune Finally, Immunotherapy-treated 77 cases who received anti-PD-1/PD-L1 treatment evaluate predictive efficiency terms outcomes. Results separated two heterogeneous clusters: Cluster A, "immune-inflamed" cluster, containing substantial innate adaptive infiltration, B, "immune-desert" modest infiltration. We ([AUC] 0.855; 95% CI 0.777–0.932; p < 0.05) verified it set (0.844; 0.606–1; 0.05). known cohort, can identify either or (0.814; 0.717–0.911; objective had higher baseline scores than those stable progressing disease 0.05); moreover, achieved AUC 0.784 (0.643–0.926; Conclusions Our imaging practicable beneficial anti-PD-1/PD-L1-treated patients. It particularly effective identifying may aid its transformation phenotype.

Language: Английский

Citations

9
Artificial intelligence: opportunities and challenges in the clinical applications of triple-negative breast cancer DOI

Jiamin Guo,

Junjie Hu, Yichen Zheng

et al.

British Journal of Cancer, Journal Year: 2023, Volume and Issue: 128(12), P. 2141 - 2149

Published: March 4, 2023

Language: Английский

Citations

22
Update on DWI for Breast Cancer Diagnosis and Treatment Monitoring DOI Open Access
Roberto Lo Gullo, Savannah C. Partridge, Hee Jung Shin

et al.

American Journal of Roentgenology, Journal Year: 2023, Volume and Issue: 222(1)

Published: Oct. 18, 2023

DWI is a noncontrast MRI technique that measures the diffusion of water molecules within biologic tissue. increasingly incorporated into routine breast examinations. Currently, main applications are cancer detection and characterization, prognostication, prediction treatment response to neoadjuvant chemotherapy. In addition, promising as alternative for screening. Problems with suboptimal resolution image quality have restricted mainstream use imaging, but these shortcomings being addressed through several technologic advancements. this review, we present an up-to-date assessment including summary clinical literature recommendations future use.

Language: Английский

Citations

20
Disulfidptosis-associated lncRNAs predict breast cancer subtypes DOI Creative Commons
Qing Xia, Qibin Yan, Zehua Wang

et al.

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

Published: Sept. 27, 2023

Abstract Disulfidptosis is a newly discovered mode of cell death. However, its relationship with breast cancer subtypes remains unclear. In this study, we aimed to construct disulfidptosis-associated subtype prediction model. We obtained 19 disulfidptosis-related genes from published articles and performed correlation analysis lncRNAs differentially expressed in cancer. then used the random forest algorithm select important establish identified 132 significantly associated disulfidptosis (FDR < 0.01, |R|> 0.15) selected first four build model (training set AUC = 0.992). The accurately predicted (test 0.842). Among key lncRNAs, LINC02188 had highest expression Basal subtype, while LINC01488 GATA3-AS1 lowest Basal. Her2 LINC00511 level compared other lncRNAs. LumA LumB subtypes, these subtypes. Normal Our study also found that were closely related RNA methylation modification angiogenesis 0.05, 0.1), as well immune infiltrating cells ( P .adj 0.1). based on can predict provide new direction for research clinical therapeutic targets

Language: Английский

Citations

17
Development and validation of a clinicoradiomic nomogram to assess the HER2 status of patients with invasive ductal carcinoma DOI Creative Commons

Aqiao Xu,

Xiufeng Chu, Shengjian Zhang

et al.

BMC Cancer, Journal Year: 2022, Volume and Issue: 22(1)

Published: Aug. 10, 2022

The determination of HER2 expression status contributes significantly to HER2-targeted therapy in breast carcinoma. However, an economical, efficient, and non-invasive assessment is lacking. We aimed develop a clinicoradiomic nomogram based on radiomics scores extracted from multiparametric MRI (mpMRI, including ADC-map, T2W1, DCE-T1WI) clinical risk factors assess status.We retrospectively collected 214 patients with pathologically confirmed invasive ductal carcinoma between January 2018 March 2021 Fudan University Shanghai Cancer Center, randomly divided this cohort into training set (n = 128, 42 HER2-positive 86 HER2-negative cases) validation 86, 28 58 at ratio 6:4. original transformed pretherapy mpMRI images were treated by semi-automated segmentation manual modification the DeepWise scientific research platform v1.6 ( http://keyan.deepwise.com/ ), then feature extraction was implemented PyRadiomics library. Recursive elimination (RFE) logistic regression (LR) LASSO adpoted identify optimal features before modeling. LR, Linear Discriminant Analysis (LDA), support vector machine (SVM), random forest (RF), naive Bayesian (NB) XGBoost (XGB) algorithms used construct signatures. Independent predictors identified through univariate analysis (age, tumor location, ki-67 index, histological grade, lymph node metastasis). Then, signature best diagnostic performance (Rad score) further combined significant model (nomogram) using multivariate regression. discriminative power constructed models evaluated AUC, DeLong test, calibration curve, decision curve (DCA).70 (32.71%) enrolled cases HER2-positive, while 144 (67.29%) HER2-negative. Eleven retained 6 radiomcis classifiers which RF classifier showed highest AUC 0.887 (95%CI: 0.827-0.947) acheived 0.840 0.758-0.922) set. A that incorporated Rad score two selected (Ki-67 index grade) yielded better discrimination compared (p 0.374, Delong test), 0.945 0.904-0.987) 0.868 0.789-0.948; p 0.123) Moreover, p-value 0.732 Hosmer-Lemeshow test demonstrated good agreement, DCA verified benefits nomogram.Post largescale validation, may have potential be as tool for prediction.

Language: Английский

Citations

27
Intra- and Peritumoral Radiomics Model Based on Early DCE-MRI for Preoperative Prediction of Molecular Subtypes in Invasive Ductal Breast Carcinoma: A Multitask Machine Learning Study DOI Creative Commons
Shuhai Zhang, Xiaolei Wang, Zhao Yang

et al.

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

Published: June 24, 2022

Purpose The aim of this study is to investigate radiomics features extracted from the optimal peritumoral region and intratumoral area on early phase dynamic contrast-enhanced MRI (DCE-MRI) for predicting molecular subtypes invasive ductal breast carcinoma (IDBC). Methods A total 422 IDBC patients with immunohistochemical fluorescence in situ hybridization results two hospitals (Center 1: 327 cases, Center 2: 95 cases) who underwent preoperative DCE-MRI were retrospectively enrolled. After image preprocessing, radiomic four regions centers, selected region. Based intratumoral, features, clinical–radiological characteristics, five models constructed through support vector machine (SVM) multiple classification tasks related visualized by nomogram. performance was evaluated receiver operating characteristic curves, confusion matrix, calibration decision curve analysis. Results 6-mm size defined hormone receptor (HR)-positive vs others, triple-negative cancer (TNBC) HR-positive human epidermal growth factor 2 (HER2)-enriched TNBC, 8 mm applied HER2-enriched others. combined three binary (HR-positive TNBC others) obtained AUCs 0.838, 0.848, 0.930 training cohort, respectively; 0.827, 0.813, 0.879 internal test 0.791, 0.707, 0.852 external respectively. Conclusion Radiomics had a potential predict HR-positive, HER2-enriched, preoperatively.

Language: Английский

Citations

25
Radiomics and artificial intelligence analysis by T2-weighted imaging and dynamic contrast-enhanced magnetic resonance imaging to predict Breast Cancer Histological Outcome DOI
Antonella Petrillo, Roberta Fusco,

Maria Luisa Barretta

et al.

La radiologia medica, Journal Year: 2023, Volume and Issue: 128(11), P. 1347 - 1371

Published: Oct. 6, 2023

Language: Английский

Citations

14
Development and Validation of a Pathomics Model Using Machine Learning to Predict <i>CXCL8</i> Expression and Prognosis in Head and Neck Cancer DOI Creative Commons
Weihua Wang, Suyu Ruan, Yuhang Xie

et al.

Clinical and Experimental Otorhinolaryngology, Journal Year: 2024, Volume and Issue: 17(1), P. 85 - 97

Published: Jan. 22, 2024

Objectives. The necessity to develop a method for prognostication and identify novel biomarkers personalized medicine in patients with head neck squamous cell carcinoma (HNSCC) cannot be overstated. Recently, pathomics, which relies on quantitative analysis of medical imaging, has come the forefront. CXCL8, an essential inflammatory cytokine, been shown correlate overall survival (OS). This study examined relationship between <i>CXCL8</i> mRNA expression pathomics features aimed explore biological underpinnings <i>CXCL8</i>.Methods. Clinical information transcripts per million sequencing data were obtained from Cancer Genome Atlas (TCGA)-HNSCC dataset. We identified correlations patient rates using Kaplan-Meier curve. A retrospective 313 samples diagnosed HNSCC TCGA database was conducted. Pathomics extracted hematoxylin eosin–stained images, then minimum redundancy maximum relevance, recursive feature elimination (mRMR-RFE) applied, followed by screening logistic regression algorithm.Results. curves indicated that high significantly associated decreased OS. model incorporated 16 radiomics mRMR-RFE training set demonstrated strong performance testing set. Calibration plots showed probability gene predicted good agreement actual observations, suggesting model’s clinical applicability.Conclusion. serves as effective tool predicting prognosis can aid decision-making. Elevated levels may lead reduced DNA damage are pro-inflammatory tumor microenvironment, offering potential therapeutic target.

Language: Английский

Citations

5
Application of radiomics to meningiomas: A systematic review DOI
Ruchit V. Patel, Shun Yao, Raymond Y. Huang

et al.

Neuro-Oncology, Journal Year: 2023, Volume and Issue: 25(6), P. 1166 - 1176

Published: Feb. 1, 2023

Abstract Background Quantitative imaging analysis through radiomics is a powerful technology to non-invasively assess molecular correlates and guide clinical decision-making. There has been growing interest in image-based phenotyping for meningiomas given the complexities management. Methods We systematically reviewed meningioma analyses published PubMed, Embase, Web of Science until December 20, 2021. compiled performance data assessed publication quality using score (RQS). Results A total 170 publications were grouped into 5 categories applications meningiomas: Tumor detection segmentation (21%), classification across neurologic diseases (54%), grading (14%), feature correlation (3%), prognostication (8%). majority focused on technical model development (73%) versus (27%), with increasing adoption deep learning. Studies utilized either private institutional (50%) or public (49%) datasets, only 68% validation dataset. For segmentation, radiomic models had mean accuracy 93.1 ± 8.1% dice coefficient 88.8 7.9%. Meningioma 95.2 4.0%. area-under-the-curve (AUC) 0.85 0.08. Correlation biological features AUC 0.89 0.07. Prognostication course 0.83 While studies higher RQS compared studies, was low overall 6.7 5.9 (possible range −8 36). Conclusions global growth radiomics, driven by accessibility novel computational methodology. Translatability toward complex tasks such as requires that improve quality, develop comprehensive patient engage prospective trials.

Language: Английский

Citations

12