Towards clinical application of image mining: a systematic review on artificial intelligence and radiomics

European Journal of Nuclear Medicine and Molecular Imaging, Journal Year: 2019, Volume and Issue: 46(13), P. 2656 - 2672

Published: June 18, 2019

The aim of this systematic review was to analyse literature on artificial intelligence (AI) and radiomics, including all medical imaging modalities, for oncological non-oncological applications, in order assess how far the image mining research stands from routine application. To do this, we applied a trial phases classification inspired drug development process. Among articles considered inclusion PubMed were multimodality AI radiomics investigations, with validation analysis aimed at relevant clinical objectives. Quality assessment selected papers performed according QUADAS-2 criteria. We developed criteria studies. Overall 34,626 retrieved, 300 applying inclusion/exclusion criteria, 171 high-quality (QUADAS-2 ≥ 7) identified analysed. In 27/171 (16%), 141/171 (82%), 3/171 (2%) studies an AI-based algorithm, model, combined radiomics/AI approach, respectively, described. A total 26/27(96%) 1/27 (4%) classified as phase II III, respectively. Consequently, 13/141 (9%), 10/141 (7%), 111/141 (79%), 7/141 (5%) 0, I, II, All three categorised trials. results are promising but still not mature enough tools be implemented setting widely used. transfer learning well-known process, some specific adaptations discipline could represent most effective way algorithms become standard care tools.

Language: Английский

---

Development and Validation of a Preoperative Magnetic Resonance Imaging Radiomics–Based Signature to Predict Axillary Lymph Node Metastasis and Disease-Free Survival in Patients With Early-Stage Breast Cancer

JAMA Network Open, Journal Year: 2020, Volume and Issue: 3(12), P. e2028086 - e2028086

Published: Dec. 8, 2020

Axillary lymph node metastasis (ALNM) status, typically estimated using an invasive procedure with a high false-negative rate, strongly affects the prognosis of recurrence in breast cancer. However, preoperative noninvasive tools to accurately predict ALNM status and disease-free survival (DFS) are lacking.To develop validate dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) radiomic signatures for identification assess individual DFS patients early-stage cancer.This retrospective prognostic study included histologically confirmed cancer diagnosed at 4 hospitals China from July 3, 2007, September 21, 2019, randomly divided (7:3) into development vaidation cohorts. All underwent MRI scans, were treated surgery sentinel biopsy or ALN dissection, pathologically examined determine status. Data analysis was conducted February 15, March 20, 2020.Clinical DCE-MRI signatures.The primary end points DFS.This 1214 women (median [IQR] age, 47 [42-55] years), split (849 [69.9%]) validation (365 [30.1%]) The signature identified cohorts areas under curve (AUCs) 0.88 0.85, respectively, clinical-radiomic nomogram predicted (AUC, 0.92 0.90, respectively) based on least absolute shrinkage selection operator (LASSO)-logistic regression model. 3-year 0.81 0.73, respectively), could discriminate high-risk low-risk cohort (hazard ratio [HR], 0.04; 95% CI, 0.01-0.11; P < .001) (HR, 0.004-0.32; random forest-Cox associated 0.89 respectively). decision demonstrated that displayed better clinical predictive usefulness than alone.This described application MRI-based machine learning cancer, presenting novel individualized nomograms be used DFS. useful decision-making personalized surgical interventions therapeutic regimens

Language: Английский

---

Deep Learning vs. Radiomics for Predicting Axillary Lymph Node Metastasis of Breast Cancer Using Ultrasound Images: Don't Forget the Peritumoral Region

Frontiers in Oncology, Journal Year: 2020, Volume and Issue: 10

Published: Jan. 31, 2020

Objective: Axillary lymph node (ALN) metastasis status is important in guiding treatment breast cancer. The aims were to assess how deep convolutional neural network (CNN) performed compared with radiomics analysis predicting ALN using ultrasound, and investigate the value of both intratumoral peritumoral regions prediction. Methods: We retrospectively enrolled 479 cancer patients 2,395 ultrasound images. Based on intratumoral, peritumoral, combined intra- regions, three CNNs built DenseNet, models random forest, respectively. By combining molecular subtype, another built. All training cohort (343 1,715 images) evaluated testing (136 680 ROC analysis. Another prospective 16 was further test models. Results: AUCs image-only training/testing cohorts 0.957/0.912 for region, 0.944/0.775 0.937/0.748 which numerically higher than their corresponding 0.940/0.886, 0.920/0.724, 0.913/0.693. overall performance image-molecular terms slightly increased 0.962/0.933, 0.951/0.813, 0.931/0.794, region significantly better those either or (p < 0.05). In study, CNN model achieved highest AUC 0.95 among all Conclusions: showed For models, performance.

Language: Английский

---

Machine learning in breast MRI

Journal of Magnetic Resonance Imaging, Journal Year: 2019, Volume and Issue: 52(4), P. 998 - 1018

Published: July 5, 2019

Machine-learning techniques have led to remarkable advances in data extraction and analysis of medical imaging. Applications machine learning breast MRI continue expand rapidly as increasingly accurate 3D lesion segmentation allows the combination radiologist-level interpretation (eg, BI-RADS lexicon), from advanced multiparametric imaging techniques, patient-level such genetic risk markers. Advances feature rapid dataset analysis, which offers promise large pooled multiinstitutional analysis. The object this review is provide an overview machine-learning deep-learning for MRI, including supervised unsupervised methods, anatomic segmentation, segmentation. Finally, it explores role learning, current limitations, future applications texture radiomics, radiogenomics. Level Evidence: 3 Technical Efficacy Stage: 2 J. Magn. Reson. Imaging 2019. 2020;52:998-1018.

Language: Английский

---

Radiomics Analysis of Dynamic Contrast-Enhanced Magnetic Resonance Imaging for the Prediction of Sentinel Lymph Node Metastasis in Breast Cancer

Frontiers in Oncology, Journal Year: 2019, Volume and Issue: 9

Published: Sept. 30, 2019

Purpose: To investigate whether a combination of radiomics and automatic machine learning applied to dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) primary breast cancer can non-invasively predict axillary sentinel lymph node (SLN) metastasis. Methods: 62 patients who received DCE-MRI scan were enrolled. Tumor resection biopsy performed within 1 week after the examination. According time signal intensity curve, volumes interest (VOIs) delineated on whole tumor in images with strongest enhanced phase. Datasets randomly divided into two sets including training set (~80%) validation (~20%). A total 1,409 quantitative features extracted from each VOI. The select K best least absolute shrinkage selection operator (Lasso) used obtain optimal features. Three classification models based logistic regression (LR), XGboost, support vector (SVM) classifiers constructed. Receiver Operating Curve (ROC) analysis was analyze prediction performance models. Both feature construction firstly set, then further tested by same thresholds. Results: There is no significant difference between all clinical pathological variables without SLN metastasis (P > 0.05), except histological grade = 0.03). Six obtained as for construction. In respect accuracy MSE, SVM demonstrated highest performance, an accuracy, AUC, sensitivity (for positive SLN), specificity SLN) Mean Squared Error (MSE) 0.85, 0.83, 0.71, 1, 0.26, respectively. Conclusions: We feasibility combining artificial intelligence tumors cancer. This non-invasive approach could be very promising application.

Language: Английский

---

Robustness of radiomic features in magnetic resonance imaging: review and a phantom study

Visual Computing for Industry Biomedicine and Art, Journal Year: 2019, Volume and Issue: 2(1)

Published: Nov. 20, 2019

Abstract Radiomic analysis has exponentially increased the amount of quantitative data that can be extracted from a single image. These imaging biomarkers aid in generation prediction models aimed to further personalized medicine. However, generalizability model is dependent on robustness these features. The purpose this study review current literature regarding radiomic features magnetic resonance imaging. Additionally, phantom performed systematically evaluate behavior under various conditions (signal noise ratio, region interest delineation, voxel size change and normalization methods) using intraclass correlation coefficients. include first order, shape, gray level cooccurrence matrix run length matrix. Many are found non-robust changing parameters. Feature assessment prior feature selection, especially case combining multi-institutional data, may warranted. Further investigation needed area research.

Language: Английский

---

Intratumoral and Peritumoral Radiomics Based on Functional Parametric Maps from Breast DCE‐MRI for Prediction of HER‐2 and Ki‐67 Status

Journal of Magnetic Resonance Imaging, Journal Year: 2021, Volume and Issue: 54(3), P. 703 - 714

Published: May 6, 2021

Radiomics has been applied to breast magnetic resonance imaging (MRI) for gene status prediction. However, the features of peritumoral regions were not thoroughly investigated.To evaluate use intratumoral and from functional parametric maps based on dynamic contrast-enhanced MRI (DCE-MRI) prediction HER-2 Ki-67 status.Retrospective.A total 351 female patients (average age, 51 years) with pathologically confirmed cancer assigned training (n = 243) validation 108) cohorts.3.0T, T1 gradient echo.Radiomic extracted six calculated using time-intensity curves DCE-MRI. The intraclass correlation coefficients (ICCs) used determine reproducibility feature extraction. Based intratumoral, peritumoral, combined intra- regions, three radiomics signatures (RSs) built least absolute shrinkage selection operator (LASSO) logistic regression model, respectively.Wilcoxon rank-sum test, minimum redundancy maximum relevance, LASSO, receiver operating characteristic curve (ROC) analysis, DeLong test.The RSs achieved areas under ROC (AUCs) 0.683 (95% confidence interval [CI], 0.574-0.793) 0.690 CI, 0.577-0.804), 0.714 0.616-0.812) 0.692 0.590-0.794) in cohort, respectively. yielded AUCs 0.713 0.604-0.823) 0.749 0.656-0.841), There no significant differences performance among RSs. Most (69.7%) had good agreement (ICCs >0.8).Radiomic DCE-MRI potential identify status.3 Technical Efficacy Stage: 2.

Language: Английский

---

Preoperative prediction of lymphovascular invasion in invasive breast cancer with dynamic contrast‐enhanced‐MRI‐based radiomics

Journal of Magnetic Resonance Imaging, Journal Year: 2019, Volume and Issue: 50(3), P. 847 - 857

Published: Feb. 17, 2019

Background Lymphovascular invasion (LVI) status facilitates the selection of optimal therapeutic strategy for breast cancer patients, but in clinical practice LVI is determined pathological specimens after resection. Purpose To explore use dynamic contrast‐enhanced (DCE)‐magnetic resonance imaging (MRI)‐based radiomics preoperative prediction invasive cancer. Study Type Prospective. Population Ninety training cohort patients (22 LVI‐positive and 68 LVI‐negative) 59 validation 37 were enrolled. Field Strength/Sequence 1.5 T 3.0 T, 1 ‐weighted DCE‐MRI. Assessment Axillary lymph node (ALN) each patient was evaluated based on MR images (defined as MRI ALN status), DCE semiquantitative parameters lesions calculated. Radiomic features extracted from first postcontrast A signature constructed with 10‐fold cross‐validation. The independent risk factors identified models developed. Their performances usefulness cohort. Statistical Tests Mann–Whitney U ‐test, chi‐square test, kappa statistics, least absolute shrinkage operator (LASSO) regression, logistic receiver operating characteristic (ROC) analysis, DeLong decision curve analysis (DCA). Results Two radiomic selected to construct signature. (odds ratio, 10.452; P < 0.001) 2.895; = 0.031) LVI. value area under (AUC) a model combining both (0.763) higher than that alone (0.665; 0.029) similar (0.752; 0.857). DCA showed combined added more net benefit either feature alone. Data Conclusion DCE‐MRI‐based combination effective predicting before surgery. Level Evidence: Technical Efficacy Stage: 2 J. Magn. Reson. Imaging 2019;50:847–857.

Language: Английский

---

Deep learning radiomics of ultrasonography: Identifying the risk of axillary non-sentinel lymph node involvement in primary breast cancer

EBioMedicine, Journal Year: 2020, Volume and Issue: 60, P. 103018 - 103018

Published: Sept. 24, 2020

BackgroundCompletion axillary lymph node dissection is overtreatment for patients with sentinel (SLN) metastasis in whom the metastatic risk of residual non-SLN (NSLN) low. However, National Comprehensive Cancer Network panel posits that none previous studies has successfully identified such subset patients. Here, we develop a multicentre deep learning radiomics ultrasonography model (DLRU) to predict SLN and NSLN metastasis.MethodsIn total, 937 eligible breast cancer ultrasound images were enrolled from two hospitals as training set (n = 542) independent test 395) respectively. Using images, developed validated prediction combined sequentially identify NSLN, thereby, classifying relevant management groups.FindingsIn set, DLRU yields best performance identifying disease SLNs (sensitivity=98.4%, 95% CI 96.6–100) NSLNs 95.6–99.9). The also accurately stratifies without or into corresponding low-risk (LR)-SLN high-risk (HR)-SLN&LR-NSLN category negative predictive value 97% (95% 94.2–100) 91.7% 88.8–97.9), Moreover, compared current clinical management, appropriately assigned 51% (39.6%/77.4%) overtreated entire study cohort LR group, perhaps avoiding overtreatment.InterpretationThe indicates it may offer simple preoperative tool promote personalized cancer.FundingThe Nature Science Foundation China; Outstanding Youth Fund Project Natural Scientific research project Heilongjiang Health Committee; Postgraduate Research &Practice Innovation Program Harbin Medical University.

Language: Английский

---

Radiomics in breast MRI: current progress toward clinical application in the era of artificial intelligence

La radiologia medica, Journal Year: 2021, Volume and Issue: 127(1), P. 39 - 56

Published: Oct. 26, 2021

Language: Английский

---