A Radiomics Model Combining Machine Learning and Neural Networks for High-Accuracy Prediction of Cervical Lymph Node Metastasis on Ultrasound of Head and Neck Squamous Cell Carcinoma

Oral Surgery Oral Medicine Oral Pathology and Oral Radiology, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

Language: Английский

---

PSMA PET/CT based multimodal deep learning model for accurate prediction of pelvic lymph-node metastases in prostate cancer patients identified as candidates for extended pelvic lymph node dissection by preoperative nomograms

European Journal of Nuclear Medicine and Molecular Imaging, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 27, 2025

Language: Английский

---

Development and validation of a machine learning model to predict the risk of lymph node metastasis in early-stage supraglottic laryngeal cancer

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

Published: Jan. 29, 2025

Background Cervical lymph node metastasis (LNM) is a significant factor that leads to poor prognosis in laryngeal cancer. Early-stage supraglottic cancer (SGLC) prone LNM. However, research on risk factors for predicting cervical LNM early-stage SGLC limited. This study seeks create and validate predictive model through the application of machine learning (ML) algorithms. Methods The training set internal validation data were extracted from Surveillance, Epidemiology, End Results (SEER) database. Data 78 patients collected Fujian Provincial Hospital independent external validation. We identified four variables associated with developed six ML models based these predict patients. In two cohorts, 167 (47.44%) 26 (33.33%) experienced LNM, respectively. Age, T stage, grade, tumor size as predictors All performed well, both validations, eXtreme Gradient Boosting (XGB) outperformed other models, AUC values 0.87 0.80, decision curve analysis demonstrated have excellent clinical applicability. Conclusions Our indicates combining algorithms can effectively diagnosed SGLC. first apply

Language: Английский

---

Deep Learning Radiomics Based on MRI for Differentiating Benign and Malignant Parapharyngeal Space Tumors

The Laryngoscope, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 11, 2025

The study aims to establish a pre-academic diagnostic tool based on deep learning and conventional radiomics features guide the clinical decision-making of parapharyngeal space (PPS) tumors. This retrospective included 217 patients with PPS tumors, from two medical centers in China March 1, 2011, October 2023. cohort was divided into training set (n = 145) test 72). A (DL) model (Rad) neck MRI were constructed distinguish malignant tumors (MTs) benign (BTs) (DLR) which integrates further developed. area under receiver operating characteristic curve (AUC), specificity, sensitivity used evaluate performance. Decision analysis (DCA) applied assess utility. Compared Rad DL models, DLR showed excellent performance this study, highest AUC 0.899 0.821 set, respectively. DCA confirmed utility distinguishing pathological types demonstrated high predictive ability diagnosing MTs BTs could serve as powerful aid preoperative diagnosis III Laryngoscope, 2025.

Language: Английский

---

Identification of Lesion Bioactivity in Hepatic Cystic Echinococcosis Using a Transformer-Based Fusion Model

Journal of Infection, Journal Year: 2025, Volume and Issue: unknown, P. 106455 - 106455

Published: March 1, 2025

Differentiating whether hepatic cystic echinococcosis (HCE) lesions exhibit biological activity is essential for developing effective treatment plans. This study evaluates the performance of a Transformer-based fusion model in predicting HCE lesion activity. analyzed CT images and clinical variables from 700 patients across three hospitals 2018 to 2023. Univariate multivariate logistic regression analyses were conducted selection construct model. Radiomic features extracted using Pyradiomics develop radiomics Additionally, 2D deep learning 3D trained images. The was constructed feature-level fusion, decision-level Transformer network architecture, allowing analysis discriminative ability correlation among radiomic features, while comparing classification multimodal models. In comparison exhibited superior identifying lesions. demonstrated highest both test set external validation set, achieving AUC values 0.997 (0.992-1.000) 0.944 (0.911-0.977), respectively, thereby outperforming models, enabling precise differentiation integrates facilitating accurate exhibiting significant potential application.

Language: Английский

---

CT-based radiomics-deep learning model predicts occult lymph node metastasis in early-stage lung adenocarcinoma patients: A multicenter study

Chinese Journal of Cancer Research, Journal Year: 2025, Volume and Issue: 37(1), P. 12 - 27

Published: Jan. 1, 2025

The neglect of occult lymph nodes metastasis (OLNM) is one the pivotal causes early non-small cell lung cancer (NSCLC) recurrence after local treatments such as stereotactic body radiotherapy (SBRT) or surgery. This study aimed to develop and validate a computed tomography (CT)-based radiomics deep learning (DL) fusion model for predicting non-invasive OLNM. Patients with radiologically node-negative adenocarcinoma from two centers were retrospectively analyzed. We developed clinical, radiomics, radiomics-clinical models using logistic regression. A DL was established three-dimensional squeeze-and-excitation residual network-34 (3D SE-ResNet34) created by integrating seleted features features. Model performance assessed area under curve (AUC) receiver operating characteristic (ROC) curve, calibration curves, decision analysis (DCA). Five predictive compared; SHapley Additive exPlanations (SHAP) Gradient-weighted Class Activation Mapping (Grad-CAM) employed visualization interpretation. Overall, 358 patients included: 186 in training cohort, 48 internal validation 124 external testing cohort. incorporating 3D SE-Resnet34 achieved highest AUC 0.947 dataset, strong cohorts (AUCs 0.903 0.907, respectively), outperforming single-modal models, clinical combined (DeLong test: P<0.05). DCA confirmed its utility, curves demonstrated excellent agreement between predicted observed OLNM probabilities. Features interpretation highlighted importance textural characteristics surrounding tumor regions stratifying risk. reliably accurately predicts early-stage adenocarcinoma, offering tool refine staging guide personalized treatment decisions. These results may aid clinicians optimizing surgical strategies.

Language: Английский

---

Evaluating fusion models for predicting occult lymph node metastasis in tongue squamous cell carcinoma.

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

Published: March 5, 2025

Language: Английский

---

Dual-Modality Virtual Biopsy System Integrating MRI and MG for Noninvasive Predicting HER2 Status in Breast Cancer

Academic Radiology, Journal Year: 2025, Volume and Issue: unknown

Published: March 1, 2025

Language: Английский

---

Multi-dimensional interpretable deep learning-radiomics based on intra-tumoral and spatial habitat for preoperative prediction of thymic epithelial tumours risk categorisation

Acta Oncologica, Journal Year: 2025, Volume and Issue: 64, P. 391 - 405

Published: March 13, 2025

Background and purpose: This study aims to develop compare combined models based on enhanced CT-based radiomics, multi-dimensional deep learning, clinical-conventional imaging spatial habitat analysis achieve accurate prediction of thymoma risk classification. Materials Methods: 205 consecutive patients with confirmed by surgical pathology were recruited from three medical centers. Venous phase CT images used delineate the tumor, 2D 3D learning whole tumor established feature extraction was performed. The tumors divided into different sub-regions K-means clustering method corresponding features obtained. data collected evaluated, univariate multivariate for screening. above types fused each other construct a variety models. Quantitative indicators such as area under receiver operating characteristic (ROC) curve (AUC) calculated evaluate performance model. Results: AUC RDLCSM developed LightGBM classifier 0.953 in training cohort, 0.930 internal validation 0.924 0.903 two external cohorts, respectively. performs better than RDLM (AUC range: 0.831-0.890) 2DLCSM 0.785-0.916) KNN. In addition, had highest accuracy (0.818-0.882) specificity (0.926-1.000). Interpretation: RDLCSM, which combines whole-tumor clinical-visual radiology, subregional omics, can be non-invasive tool predict

Language: Английский

---

Deep learning based on intratumoral heterogeneity predicts histopathologic grade of hepatocellular carcinoma

BMC Cancer, Journal Year: 2025, Volume and Issue: 25(1)

Published: March 18, 2025

The potential of medical imaging to non-invasively assess intratumoral heterogeneity (ITH) is increasingly being recognized. This study aimed investigate the value ITH-based deep learning model for preoperative prediction histopathologic grade in hepatocellular carcinoma (HCC). A total 858 patients from primary cohort and two external cohorts were included. 3.0T or 1.5T axial portal venous phase MRI images collected. We conducted radiomics feature-driven K-means clustering automatic partition reveal ITH. 2.5D 3D models based on ResNet architecture trained extract hidden features each subregion. selected used train Random Forest classifier, which constructed feature-fusion model. extracted voxel-level unsupervised clustered by generate three subregions. In learning, ITH had superior predictive efficacy than whole-tumor (AUC: 0.82 vs. 0.72; p = 0.004). Even validation test sets, this maintained a high AUC 0.78–0.83, net reclassification indices indicated that it could improve 25–28%. Regarding prognostic value, overall survival (OS) recurrence-free (RFS) be significantly stratified model, multivariable Cox regressions its signature was identified as risk predictor OS RFS (p < 0.05). provided non-invasive method classifying tumor differentiation HCC, may serve promising strategy stratification management.

Language: Английский

---