Fusion of shallow and deep features from 18F-FDG PET/CT for predicting EGFR-sensitizing mutations in non-small cell lung cancer

Quantitative Imaging in Medicine and Surgery, Journal Year: 2024, Volume and Issue: 14(8), P. 5460 - 5472

Published: Feb. 23, 2024

Non-small cell lung cancer (NSCLC) patients with epidermal growth factor receptor-sensitizing (EGFR-sensitizing) mutations exhibit a positive response to tyrosine kinase inhibitors (TKIs). Given the limitations of current clinical predictive methods, it is critical explore radiomics-based approaches. In this study, we leveraged deep-learning technology multimodal radiomics data more accurately predict EGFR-sensitizing mutations.

Language: Английский

---

Impact of [18F]FDG PET/CT Radiomics and Artificial Intelligence in Clinical Decision Making in Lung Cancer: Its Current Role

Seminars in Nuclear Medicine, Journal Year: 2025, Volume and Issue: unknown

Published: March 1, 2025

Lung cancer remains one of the most prevalent cancers globally and leading cause cancer-related deaths, accounting for nearly one-fifth all fatalities. Fluoro-2-deoxy-D-glucose positron emission tomography/computed tomography ([18F]FDG PET/CT) plays a vital role in assessing lung managing disease progression. While traditional PET/CT imaging relies on qualitative analysis basic quantitative parameters, radiomics offers more advanced approach to analyzing tumor phenotypes. Recently, has gained attention its potential enhance prognostic diagnostic capabilities [18F]FDG various cancers. This review explores expanding PET/CT-based radiomics, particularly when integrated with artificial intelligence (AI), cancer, especially non-small cell (NSCLC). We how AI improve diagnostics, staging, subtype identification, molecular marker detection, which influence treatment decisions. Additionally, we address challenges clinical integration, such as protocol standardization, feature reproducibility, need extensive prospective studies. Ultimately, hold great promise enabling personalized effective treatments, potentially transforming management.

Language: Английский

---

Prediction of EGFR mutation status and its subtypes in non-small cell lung cancer based on 18F-FDG PET/CT radiological features

Nuclear Medicine Communications, Journal Year: 2025, Volume and Issue: 46(4), P. 326 - 336

Published: Jan. 20, 2025

Prediction of epidermal growth factor receptor (EGFR) mutation status and subtypes in patients with non-small cell lung cancer (NSCLC) based on 18 F-fluorodeoxyglucose ( F-FDG) PET/computed tomography (CT) radiomics features. Retrospective analysis 201 NSCLC F-FDG PET/CT EGFR genetic testing was carried out. Radiomics features clinical factors were used to construct a combined model for identifying status. Mutation/wild-type models trained training cohort n = 129) validated an internal validation 41) vs external 50). A second predicting the 19/21 locus also built evaluated subset mutations (training cohort, 55; 14). The predictive performance net benefit assessed by area under curve (AUC) subjects, nomogram, calibration decision curve. AUC distinguishing 0.864 0.806 0.791 test sets respectively, site 0.971 0.867 respectively. curves individual showed better predictions (Brier score <0.25). Decision that had application. can predict patients, guiding targeted therapy, facilitate precision medicine development.

Language: Английский

---

Artificial intelligence for tumor [18F]FDG-PET imaging: Advancement and future trends—part I

Seminars in Nuclear Medicine, Journal Year: 2025, Volume and Issue: unknown

Published: March 1, 2025

The advent of sophisticated image analysis techniques has facilitated the extraction increasingly complex data, such as radiomic features, from various imaging modalities, including [18F]FDG PET/CT, a well-established cornerstone oncological imaging. Furthermore, use artificial intelligence (AI) algorithms shown considerable promise in enhancing interpretation these quantitative parameters. Additionally, AI-driven models enable integration parameters multiple modalities along with clinical facilitating development comprehensive significant impact. However, challenges remain regarding standardization and validation AI-powered models, well their implementation real-world practice. variability acquisition protocols, segmentation methods, feature approaches across different institutions necessitates robust harmonization efforts to ensure reproducibility utility. Moreover, successful translation AI into practice requires prospective large cohorts, seamless existing workflows assess ability enhance clinicians' performance. This review aims provide an overview literature highlight three key applications: diagnostic impact, prediction treatment response, long-term patient prognostication. In first part, we will focus on head neck, lung, breast, gastroesophageal, colorectal, gynecological malignancies.

Language: Английский

---

High-dimensional multinomial multiclass severity scoring of COVID-19 pneumonia using CT radiomics features and machine learning algorithms

Scientific Reports, Journal Year: 2022, Volume and Issue: 12(1)

Published: Sept. 1, 2022

We aimed to construct a prediction model based on computed tomography (CT) radiomics features classify COVID-19 patients into severe-, moderate-, mild-, and non-pneumonic. A total of 1110 were studied from publicly available dataset with 4-class severity scoring performed by radiologist (based CT images clinical features). The entire lungs segmented followed resizing, bin discretization radiomic extraction. utilized two feature selection algorithms, namely bagging random forest (BRF) multivariate adaptive regression splines (MARS), each coupled classifier, multinomial logistic (MLR), multiclass classification models. was divided 50% (555 samples), 20% (223 30% (332 samples) for training, validation, untouched test datasets, respectively. Subsequently, nested cross-validation train/validation select the tune All predictive power indices reported testing set. performance multi-class models assessed using precision, recall, F1-score, accuracy 4 × confusion matrices. In addition, areas under receiver operating characteristic curves (AUCs) classifications calculated compared both Using BRF, 23 selected, 11 first-order, 9 GLCM, 1 GLRLM, GLDM, shape. Ten selected MARS algorithm, 3 GLSZM, shape, GLCM features. mean absolute deviation, skewness, variance first-order flatness cluster prominence Gray Level Non Uniformity Normalize GLRLM BRF algorithms. or significantly associated four-class outcomes as within MLR (All p values < 0.05). + resulted in pseudo-R

Language: Английский

---

Myocardial Perfusion SPECT Imaging Radiomic Features and Machine Learning Algorithms for Cardiac Contractile Pattern Recognition

Journal of Digital Imaging, Journal Year: 2022, Volume and Issue: 36(2), P. 497 - 509

Published: Nov. 14, 2022

Abstract A U-shaped contraction pattern was shown to be associated with a better Cardiac resynchronization therapy (CRT) response. The main goal of this study is automatically recognize left ventricular contractile patterns using machine learning algorithms trained on conventional quantitative features (ConQuaFea) and radiomic extracted from Gated single-photon emission computed tomography myocardial perfusion imaging (GSPECT MPI). Among 98 patients standard resting GSPECT MPI included in study, 29 received CRT 69 did not (also had inclusion criteria but receive treatment yet at the time data collection, or refused treatment). total non-CRT were employed for training, testing. models built utilizing three distinct feature sets (ConQuaFea, radiomics, ConQuaFea + radiomics (combined)), which chosen Recursive elimination (RFE) selection (FS), then seven different (ML) classifiers. In addition, outcome prediction assessed by as study’s final phase. MLP classifier highest performance among (AUC, SEN, SPE = 0.80, 0.85, 0.76). RF achieved best terms AUC, values 0.65, 0.62, 0.68, respectively, models. GB approaches 0.78, 0.92, 0.63 0.74, 0.93, 0.56, combined promising obtained when detect learning.

Language: Английский

---

Time-to-event overall survival prediction in glioblastoma multiforme patients using magnetic resonance imaging radiomics

La radiologia medica, Journal Year: 2023, Volume and Issue: 128(12), P. 1521 - 1534

Published: Sept. 26, 2023

Abstract Purpose Glioblastoma Multiforme (GBM) represents the predominant aggressive primary tumor of brain with short overall survival (OS) time. We aim to assess potential radiomic features in predicting time-to-event OS patients GBM using machine learning (ML) algorithms. Materials and methods One hundred nineteen GBM, who had T1-weighted contrast-enhanced T2-FLAIR MRI sequences, along clinical data time, were enrolled. Image preprocessing included 64 bin discretization, Laplacian Gaussian (LOG) filters three Sigma values eight variations Wavelet Transform. Images then segmented, followed by extraction 1212 features. Seven feature selection (FS) six ML algorithms utilized. The combination preprocessing, FS, (12 × 7 6 = 504 models) was evaluated multivariate analysis. Results Our analysis showed that best prognostic FS/ML combinations are Mutual Information (MI)/Cox Boost, MI/Generalized Linear Model Boosting (GLMB) Network (GLMN), all which done via LOG (Sigma 1 mm) method (C-index 0.77). filter mm method, MI, GLMB GLMN achieved significantly higher C-indices than other (all p < 0.05, mean 0.65, 0.70, 0.64, respectively). Conclusion capable MRI-based radiomics variables might appear promising assisting clinicians prediction GBM. Further research is needed establish applicability management clinic.

Language: Английский

---

Machine Learning and Texture Analysis of [18F]FDG PET/CT Images for the Prediction of Distant Metastases in Non-Small-Cell Lung Cancer Patients

Biomedicines, Journal Year: 2024, Volume and Issue: 12(3), P. 472 - 472

Published: Feb. 20, 2024

The aim of our study was to predict the occurrence distant metastases in non-small-cell lung cancer (NSCLC) patients using machine learning methods and texture analysis 18F-labeled 2-deoxy-d-glucose Positron Emission Tomography/Computed Tomography {[18F]FDG PET/CT} images. In this retrospective single-center study, we evaluated 79 with advanced NSCLC who had undergone [18F]FDG PET/CT scan at diagnosis before any therapy. Patients were divided into two independent training (n = 44) final testing 35) cohorts. Texture features primary tumors lymph node extracted from images LIFEx program. Six applied dataset entire panel features. Dedicated selection used generate different combinations five performance selected determined accuracy, confusion matrix, receiver operating characteristic (ROC) curves, area under curve (AUC). A total 104 78 lesions analyzed cohorts, respectively. support vector (SVM) decision tree showed highest accuracy cohort. Seven obtained introduced models subsequently cohorts SVM tree. AUC method higher than those best combination included shape sphericity, gray level run length matrix_run non-uniformity (GLRLM_RLNU), Total Lesion Glycolysis (TLG), Metabolic Tumor Volume (MTV), compacity. these could an 74.4% 0.63 patients.

Language: Английский

---

Predictive value of 18F-FDG PET/CT-based radiomics model for neoadjuvant chemotherapy efficacy in breast cancer: a multi-scanner/center study with external validation

European Journal of Nuclear Medicine and Molecular Imaging, Journal Year: 2023, Volume and Issue: 50(7), P. 1869 - 1880

Published: Feb. 20, 2023

Language: Английский

---

Post-revascularization Ejection Fraction Prediction for Patients Undergoing Percutaneous Coronary Intervention Based on Myocardial Perfusion SPECT Imaging Radiomics: a Preliminary Machine Learning Study

Journal of Digital Imaging, Journal Year: 2023, Volume and Issue: 36(4), P. 1348 - 1363

Published: April 14, 2023

In this study, the ability of radiomics features extracted from myocardial perfusion imaging with SPECT (MPI-SPECT) was investigated for prediction ejection fraction (EF) post-percutaneous coronary intervention (PCI) treatment. A total 52 patients who had undergone pre-PCI MPI-SPECT were enrolled in study. After normalization images, left ventricle, initially automatically segmented by k-means and active contour methods, finally edited approved an expert radiologist. More than 1700 2D 3D each patient's scan. cross-combination three feature selections seven classifier methods implemented. Three classes no or dis-improvement (class 1), improved EF 0 to 5% 2), over 3) predicted using tenfold cross-validation. Lastly, models evaluated based on accuracy, AUC, sensitivity, specificity, precision, F-score. Neighborhood component analysis (NCA) selected most predictive signatures, including Gabor, first-order, NGTDM features. Among classifiers, best performance achieved fine KNN classifier, which yielded mean F-score 0.84, 0.83, 0.75, 0.87, 0.78, 0.76, respectively, 100 iterations classification, within 10-fold The MPI-SPECT-based radiomic are well suited predicting post-revascularization therefore provide a helpful approach deciding appropriate

Language: Английский

---