Staging of cervical cancer based on tumor heterogeneity characterized by texture features on (18)F-FDG PET images

Patch-Based Texture Feature Extraction Towards Improved Clinical Task Performance

Texture features can capture microstructural patterns and tissue heterogeneity, playing a pivotal role in medical image analysis. Compared to deep learning-based features, texture features offer superior interpretability in clinical applications. However, as conventional texture features focus strictly on voxel-level statistical information, they fail to account for critical spatial heterogeneity between small tissue volumes, which may hold significant importance. To overcome this limitation, we propose novel 3D patch-based texture features and develop a radiomics analysis framework to validate the efficacy of our proposed features. Specifically, multi-scale 3D patches were created to construct patch patterns via k-means clustering. The multi-resolution images were discretized based on labels of the patterns, and then texture features were extracted to quantify the spatial heterogeneity between patches. Twenty-five cross-combination models of five feature selection methods and five classifiers were constructed. Our methodology was evaluated using two independent MRI datasets. Specifically, 145 breast cancer patients were included for axillary lymph node metastasis prediction, and 63 cervical cancer patients were enrolled for histological subtype prediction. Experimental results demonstrated that the proposed 3D patch-based texture features achieved an AUC of 0.76 in the breast cancer lymph node metastasis prediction task and an AUC of 0.94 in cervical cancer histological subtype prediction, outperforming conventional texture features (0.74 and 0.83, respectively). Our proposed features have successfully captured multi-scale patch-level texture representations, which could enhance the application of imaging biomarkers in the precise prediction of cancers and personalized therapeutic interventions.

Radiomics analysis in differentiating osteosarcoma and chondrosarcoma based on T2-weighted imaging and contrast-enhanced T1-weighted imaging

This study was performed to investigate the diagnostic value of radiomics models constructed by fat suppressed T2-weighted imaging (T2WI-FS) and contrast-enhanced T1-weighted imaging (CET1) based on magnetic resonance imaging (MRI) for differentiation of osteosarcoma (OS) and chondrosarcoma (CS). In this retrospective cohort study, we included all inpatients with pathologically confirmed OS or CS from Second Xiangya Hospital of Central South University (Hunan, China) as of October 2020. Demographic and imaging variables were extracted from electronic medical records and compared between OS and CS group. Totals of 530 radiomics features were extracted from CET1 and T2WI-FS sequences based on MRI. The least absolute shrinkage and selection operator (LASSO) method was used for screening and dimensionality reduction of the radiomics model. Multivariate logistic regression analysis was performed to construct the radiomics model, and receiver operating characteristic curve (ROC) was generated to evaluate the diagnostic accuracy of the radiomics model. The training cohort and validation cohort included 87 and 29 patients, respectively. 8 CET1 features and 15 T2WI-FS features were screened based on the radiomics features. In the training group, the area under the receiver-operator characteristic curve (AUC) value for CET1 and T2WI-FS sequences in the radiomics model was 0.894 (95% CI 0.817-0.970) and 0.970 (95% CI 0.940-0.999), respectively. In the validation group, the AUC value for CET1 and T2WI-FS sequences in the radiomics model was 0.821 (95% CI 0.642-1.000) and 0.899 (95% CI 0.785-1.000), respectively. In this study, we developed a radiomics model based on T2WI-FS and CET1 sequences to differentiate between OS and CS. This model exhibits good performance and can help clinicians make decisions and optimize the use of healthcare resources.

Efficient bladder cancer diagnosis using an improved RIME algorithm with Orthogonal Learning

Bladder cancer (BC) diagnosis presents a critical challenge in biomedical research, necessitating accurate tumor classification from diverse datasets for effective treatment planning. This paper introduces a novel wrapper feature selection (FS) method that leverages a hybrid optimization algorithm combining Orthogonal Learning (OL) with a rime optimization algorithm (RIME), termed mRIME. The mRIME algorithm is designed to avoid local optima, streamline the search process, and select the most relevant features without compromising classifier performance. It also introduces mRIME-SVM, a novel hybrid model integrating modified mRIME for FS with Support Vector Machine (SVM) for classification. The mRIME algorithm is employed as an FS method and is also utilized to fine-tune the hyperparameters of it the It SVM, enhancing the overall classification accuracy. Specifically, mRIME navigates complex search spaces to optimize FS without compromising classifier performance. Evaluated on eight diverse BC datasets, mRIME-SVM outperforms popular metaheuristic algorithms, ensuring precise and reliable diagnostic outcomes. Moreover, the proposed mRIME was employed for tackling global optimization problems. It has been thoroughly assessed using the IEEE Congress on Evolutionary Computation 2022 (CEC'2022) test suite. Comparative analyzes with Gray wolf optimization (GWO), Whale optimization algorithm (WOA), Harris hawks optimization (HHO), Golden Jackal Optimization (GJO), Hunger Game optimization algorithm (HGS), Sinh Cosh Optimizer (SCHO), and the original RIME highlight mRIME's competitiveness and efficacy across diverse optimization tasks. Leveraging mRIME's success, mRIME-SVM achieves high classification accuracy on nine BC datasets, surpassing existing models. Results underscore mRIME's competitiveness and applicability across diverse optimization tasks, extending its utility to enhance BC classification. This study contributes to advancing BC diagnostics with a robust computational framework, promising broader applications in bioinformatics and AI-driven medical research.

MRI-based radiomics features for the non-invasive prediction of FIGO stage in cervical carcinoma: A multi-center study

PURPOSE

To develop and validate a model based on MRI radiomics modals for predicting surgical high FIGO(IB3 and ≥ IIA2) and low FIGO(IB1, IB2, and IIA1) stages in patients with cervical carcinoma (CC).

METHODS

A total of 296 early-stage patients with CC (preoperative FIGO stages IB-IIA) confirmed by surgery and pathology were included in this retrospective study from two institutions For each patient,we extracted radiomics features from spectral attenuated inversion-recovery T2-weighted (SPAIR-T2W) and contrast-enhanced T1-weighted (CE-T1W) images.Manual segmentation was performed using the 3D Slicer software, while radiomics features were extracted, screened using the R software. A 2-stage feature extraction strategy involving univariate analysis and the Least Absolute Shrinkage Selection Operator technique was performed. A support vector machine-based model was eventually constructed. Predictive accuracy of the training and validation datasets was assessed in terms of area under the ROC curve (AUC).

RESULTS

A total of 1130 features were extracted from SPAIR-T2WI and CET1WI images respectively, in which 8 and 7 features significantly were associated with FIGO staging. AUCs of the SPAIR-T2W and CE-T1W models were were 0.803 and 0.790, respectively, in the internal validation group. In the external validation group, the AUCs were 0.767 and 0.749, respectively, which increased to 0.771 in the combined model.

CONCLUSION

Our study demonstrated the feasibility of radiomics features from SPAIR-T2W and CE-T1W images for the prediction of surgical FIGO stage in CC. Our proposed model thereby carries the potential as a non-invasive tool for the staging and treatment planning of this disease.

ADVANCES IN KNOWLEDGE

A radiomics model provide a non-invasive and objective method for the detection of FIGO staging in patients with cervical cancer before surgery, thus providing a reference for the selection of treatment options for patients.

Development and validation of a 18F-FDG PET/CT radiomics nomogram for predicting progression free survival in locally advanced cervical cancer: a retrospective multicenter study

BACKGROUND

The existing staging system cannot meet the needs of accurate survival prediction. Accurate survival prediction for locally advanced cervical cancer (LACC) patients who have undergone concurrent radiochemotherapy (CCRT) can improve their treatment management. Thus, this present study aimed to develop and validate radiomics models based on pretreatment 18Fluorine-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET)-computed tomography (CT) images to accurately predict the prognosis in patients.

METHODS

The data from 190 consecutive patients with LACC who underwent pretreatment 18F-FDG PET-CT and CCRT at two cancer hospitals were retrospectively analyzed; 176 patients from the same hospital were randomly divided into training (n = 117) and internal validation (n = 50) cohorts. Clinical features were selected from the training cohort using univariate and multivariate Cox proportional hazards models; radiomic features were extracted from PET and CT images and filtered using least absolute shrinkage and selection operator and Cox proportional hazard regression. Three prediction models and a nomogram were then constructed using the previously selected clinical, CT and PET radiomics features. The external validation cohort that was used to validate the models included 23 patients with LACC from another cancer hospital. The predictive performance of the constructed models was evaluated using receiver operator characteristic curves, Kaplan Meier curves, and a nomogram.

RESULTS

In total, one clinical, one PET radiomics, and three CT radiomics features were significantly associated with progression-free survival in the training cohort. Across all three cohorts, the combined model displayed better efficacy and clinical utility than any of these parameters alone in predicting 3-year progression-free survival (area under curve: 0.661, 0.718, and 0.775; C-index: 0.698, 0.724, and 0.705, respectively) and 5-year progression-free survival (area under curve: 0.661, 0.711, and 0.767; C-index, 0.698, 0.722, and 0.676, respectively). On subsequent construction of a nomogram, the calibration curve demonstrated good agreement between actually observed and nomogram-predicted values.

CONCLUSIONS

In this study, a clinico-radiomics prediction model was developed and successfully validated using an independent external validation cohort. The nomogram incorporating radiomics and clinical features could be a useful clinical tool for the early and accurate assessment of long-term prognosis in patients with LACC patients who undergo concurrent chemoradiotherapy.

Radiomics-based T-staging of hollow organ cancers

Cancer growing in hollow organs has become a serious threat to human health. The accurate T-staging of hollow organ cancers is a major concern in the clinic. With the rapid development of medical imaging technologies, radiomics has become a reliable tool of T-staging. Due to similar growth characteristics of hollow organ cancers, radiomics studies of these cancers can be used as a common reference. In radiomics, feature-based and deep learning-based methods are two critical research focuses. Therefore, we review feature-based and deep learning-based T-staging methods in this paper. In conclusion, existing radiomics studies may underestimate the hollow organ wall during segmentation and the depth of invasion in staging. It is expected that this survey could provide promising directions for following research in this realm.

Recurrence risk stratification for locally advanced cervical cancer using multi-modality transformer network

Objectives

Recurrence risk evaluation is clinically significant for patients with locally advanced cervical cancer (LACC). We investigated the ability of transformer network in recurrence risk stratification of LACC based on computed tomography (CT) and magnetic resonance (MR) images.

Methods

A total of 104 patients with pathologically diagnosed LACC between July 2017 and December 2021 were enrolled in this study. All patients underwent CT and MR scanning, and their recurrence status was identified by the biopsy. We randomly divided patients into training cohort (48 cases, non-recurrence: recurrence = 37: 11), validation cohort (21 cases, non-recurrence: recurrence = 16: 5), and testing cohort (35 cases, non-recurrence: recurrence = 27: 8), upon which we extracted 1989, 882 and 315 patches for model's development, validation and evaluation, respectively. The transformer network consisted of three modality fusion modules to extract multi-modality and multi-scale information, and a fully-connected module to perform recurrence risk prediction. The model's prediction performance was assessed by six metrics, including the area under the receiver operating characteristic curve (AUC), accuracy, f1-score, sensitivity, specificity and precision. Univariate analysis with F-test and T-test were conducted for statistical analysis.

Results

The proposed transformer network is superior to conventional radiomics methods and other deep learning networks in both training, validation and testing cohorts. Particularly, in testing cohort, the transformer network achieved the highest AUC of 0.819 ± 0.038, while four conventional radiomics methods and two deep learning networks got the AUCs of 0.680 ± 0.050, 0.720 ± 0.068, 0.777 ± 0.048, 0.691 ± 0.103, 0.743 ± 0.022 and 0.733 ± 0.027, respectively.

Conclusions

The multi-modality transformer network showed promising performance in recurrence risk stratification of LACC and may be used as an effective tool to help clinicians make clinical decisions.

Development of a Novel Deep Learning-Based Prediction Model for the Prognosis of Operable Cervical Cancer

BACKGROUND

Cervical cancer ranks as the 4th most common female cancer worldwide. Early stage cervical cancer patients can be treated with operation, but clinical staging system is not a good predictor of patients' survival. We aimed to develop a novel prognostic model to predict the prognosis for operable cervical cancer patients with better accuracy than clinical staging system.

METHODS

A total of 13,952 operable cervical cancer patients were retrospectively enrolled in this study. The whole dataset was randomly split into a training set (n = 9,068, 65%), validation set (n = 2,442, 17.5%), and testing set (n = 2,442, 17.5%). Cox proportional hazard (CPH) model and random survival forest (RSF) model were used as baseline models for the prediction of overall survival (OS). Then, a deep survival learning model (DSLM) was developed for OS prediction. Finally, a novel prognostic model was explored based on this DSLM.

RESULTS

The C-indexes for the CPH and RSF model were 0.731 and 0.753, respectively. DSLM, which had four layers that had 50 neurons in each layer, achieved a C-index of 0.782 in the validation set and a C-index of 0.758 in the testing set. The novel prognostic model based on DSLM showed better performances than the conventional clinical staging system (area under receiver operating curves were 0.826 and 0.689, respectively). Personalized survival curves for individual patient using this novel model also showed notably different survival slopes.

CONCLUSIONS

Our study developed a novel, practical, personalized prognostic model for operable cervical cancer patients. This novel prognostic model may have the potential to provide a more prognostic information to oncologists.

Application and Clinical Value of Machine Learning-Based Cervical Cancer Diagnosis and Prediction Model in Adjuvant Chemotherapy for Cervical Cancer: A Single-Center, Controlled, Non-Arbitrary Size Case-Control Study

Objective

A case-control study was conducted to explore the application and clinical value of machine learning-based cervical cancer (CC) diagnosis and prediction model in adjuvant chemotherapy of CC.

Methods

From August 2019 to August 2021, 46 patients with stage IA CC (study group) and 55 patients with high-grade squamous intraepithelial lesions (HSIL) (control group) were retrospectively analyzed. All patients completed routine MRI examinations, the ADC values of diseased CC and normal cervix and cervical tissues in different stages were compared, and the changes of ADC values in CC tissues before and after chemotherapy were analyzed. The training set (IA = 37, HSIL = 44) and test set (IA = 9, HSIL = 11) are set in a ratio of 4 : 1. The preoperative MRI images were collected and uploaded to the radiomics cloud platform after preprocessing, and the cervix was manually delineated layer by layer on OSag-T2WI, OAx-T1WI, and OAx-T2FS, respectively, to obtain a three-dimensional volume of interest (VOI) of the cervix to extract omics features. Variance Threshold analysis, univariate feature selection (SelectKBest), and least absolute shrinkage and selection operator (LASSO) are adopted to reduce the dimension of data and enroll features. The arbitrary forest model was adopted for machine learning, the ROC curve was drawn, and the diagnostic performance of different sequence omics models was analyzed.

Results

Compared with ADC of stage A CC and HSIL, the ADC value of CC was remarkably lower than that of normal CC (P < 0.05). The ROC curve analysis of ADC value to differentiate CC and normal cervix indicated that the AUC was 0.838 and the 95% confidence interval was 0.721–0.955. According to the maximum Youden index of 0.848, the optimal critical value of ADC was 1.267 × 10⁻³ mm²/s and the sensitivity and specificity were 92.21% and 9.48%, respectively. All results are indicated in Table 2. After CC treatment, 12 patients were effective (CR + PR) and 4 patients were ineffective (PD + SD). When the b value was 1000 s/mm², the ADC value of the effective patients after the second chemotherapy was significantly higher than that of the first chemotherapy and before treatment (P < 0.05). There was no significant difference between the ADC value after the first chemotherapy and before treatment, compared with before treatment (P > 0.05). There was no significant difference in ADC value between the ineffective patients before treatment and after the first and second chemotherapy (P > 0.05). A total of 8 omics features were extracted based on OSag-T2WI, all of which were wavelet features, including 7 texture features and 1 first-order feature. A total of 10 omics features were extracted based on OAx-T1WI, including 6 wavelet first-order features, 2 gradient first-order features, and 2 wavelet texture features. Based on OAx-T2FS, 6 omics features were extracted, including 3 wavelet texture features, 2 original shape features, and 1 logarithmic first-order feature. Based on OSag-T2WI&OAx-T2FS, 9 histological features were extracted, 4 from OSag-T2WI and 5 from OAx-T2FS. The diagnostic performance of the four arbitrary forest models is indicated in Table 1, and the ROC curve is indicated in Figure 6. The diagnostic performance of the omics model based on OSag-T2WI&OAx-T2FS was the best in both the training set and the test set. The AUC of the training set was 0.991 (95% CI (0.94, 1.00)), and the accuracy rate was 0.925. The AUC of the test set was 0.894 (95% CI (0.75, 1.00)), and the accuracy rate was 0.835. On the other hand, the diagnostic efficiency of the group model based on OAx-T1WI was the worst in both the training set and the test set. The AUC of the training set was 0.713 (95% CI (0.52, 0.92)), and the accuracy rate was 0.71. The AUC of test set is 0.513 (95% CI (0.24, 0.77)), and the accuracy rate was 0.56, which has no practical clinical significance.

Conclusion

A CC diagnosis and prediction model based on machine learning can better distinguish stage IA CC from HSIL in the absence of clear lesions, which is of great significance for reducing invasive examination before surgery, guiding surgical procedures and adjuvant chemotherapy for CC.

Prediction of Overall Survival in Cervical Cancer Patients Using PET/CT Radiomic Features

Background: Radiomics is a field of research medicine and data science in which quantitative imaging features are extracted from medical images and successively analyzed to develop models for providing diagnostic, prognostic, and predictive information. The purpose of this work was to develop a machine learning model to predict the survival probability of 85 cervical cancer patients using PET and CT radiomic features as predictors. Methods: Initially, the patients were divided into two mutually exclusive sets: a training set containing 80% of the data and a testing set containing the remaining 20%. The entire analysis was separately conducted for CT and PET features. Genetic algorithms and LASSO regression were used to perform feature selection on the initial PET and CT feature sets. Two different survival models were employed: the Cox proportional hazard model and random survival forest. The Cox model was built using the subset of features obtained with the feature selection process, while all the available features were used for the random survival forest model. The models were trained on the training set; cross-validation was used to fine-tune the models and to obtain a preliminary measurement of the performance. The models were then validated on the test set, using the concordance index as the metric. In addition, alternative versions of the models were developed using tumor recurrence as an adjunct feature to evaluate its impact on predictive performance. Finally, the selected CT and PET features were combined to build a further Cox model. Results: The genetic algorithm was superior to the LASSO regression for feature selection. The best performing model was the Cox model, which was built using the selected CT features; it achieved a concordance index score of 0.707. With the addition of tumor recurrence as a predictive feature, the Cox CT model reached a concordance index score of 0.776. PET features, however, proved to be inadequate for survival prediction. The CT model performed better than the model with combined PET and CT features. Conclusions: The results showed that radiomic features can be used to successfully predict survival probability in cervical cancer patients. In particular, CT radiomic features proved to be better predictors than PET radiomic features in this specific case.

Multiparametric MRI-based radiomics analysis: differentiation of subtypes of cervical cancer in the early stage

BACKGROUND

There are significant differences in outcomes for different histological subtypes of cervical cancer (CC). Yet, it is difficult to distinguish CC subtypes using non-invasive methods.

PURPOSE

To investigate whether multiparametric magnetic resonance imaging (MRI)-based radiomics analysis can differentiate CC subtypes and explore tumor heterogeneity.

MATERIAL AND METHODS

This study retrospectively analyzed 96 patients with CC (squamous cell carcinoma [SCC] = 50, adenocarcinoma [AC] = 46) who underwent pelvic MRI before surgery. Radiomics features were extracted from the tumor volumes on five sequences (sagittal T2-weighted imaging [T2SAG], transverse T2-weighted imaging [T2TRA], sagittal contrast-enhanced T1-weighted imaging [CESAG], transverse contrast-enhanced T1-weighted imaging [CETRA], and apparent diffusion coefficient [ADC]). Clustering and logistic regression were used to examine the distinguishing capabilities of radiomics features extracted from five different MR sequences.

RESULTS

Among the 105 extracted radiomics features, there were 51, 38, 37, and 2 features that showed intergroup differences for T2SAG, T2TRA, ADC, and CESAG, respectively (all P < 0.05). AC had greater textural heterogeneity than SCC (P < 0.05). Upon unsupervised clustering of significantly different features, T2SAG achieved the highest accuracy (0.844; sensitivity = 0.920; specificity = 0.761). The largest area under the curve (AUC) for classification ability was 0.86 for T2SAG. Hence, the radiomics model from five combined MR sequences (AUC = 0.89; accuracy = 0.81; sensitivity = 0.67; specificity = 0.94) exhibited better differentiation ability than any MR sequence alone.

CONCLUSION

Multiparametric MRI-based radiomics models may be a promising method to differentiate AC and SCC. AC showed more heterogeneous features than SCC.

Radiomics in Oncological PET Imaging: A Systematic Review-Part 2, Infradiaphragmatic Cancers, Blood Malignancies, Melanoma and Musculoskeletal Cancers

The objective of this review was to summarize published radiomics studies dealing with infradiaphragmatic cancers, blood malignancies, melanoma, and musculoskeletal cancers, and assess their quality. PubMed database was searched from January 1990 to February 2022 for articles performing radiomics on PET imaging of at least 1 specified tumor type. Exclusion criteria includd: non-oncological studies; supradiaphragmatic tumors; reviews, comments, cases reports; phantom or animal studies; technical articles without a clinically oriented question; studies including <30 patients in the training cohort. The review database contained PMID, first author, year of publication, cancer type, number of patients, study design, independent validation cohort and objective. This database was completed twice by the same person; discrepant results were resolved by a third reading of the articles. A total of 162 studies met inclusion criteria; 61 (37.7%) studies included >100 patients, 13 (8.0%) were prospective and 61 (37.7%) used an independent validation set. The most represented cancers were esophagus, lymphoma, and cervical cancer (n = 24, n = 24 and n = 19 articles, respectively). Most studies focused on 18F-FDG, and prognostic and response to treatment objectives. Although radiomics and artificial intelligence are technically challenging, new contributions and guidelines help improving research quality over the years and pave the way toward personalized medicine.

MRI-based radiomics model can improve the predictive performance of postlaminar optic nerve invasion in retinoblastoma

OBJECTIVES

To develop an MRI-based radiomics model to predict postlaminar optic nerve invasion (PLONI) in retinoblastoma (RB) and compare its predictive performance with subjective radiologists' assessment.

METHODS

We retrospectively enrolled 124 patients with pathologically proven RB (90 in training set and 34 in validation set) who had MRI scans before surgery. A radiomics model for predicting PLONI was developed by extracting quantitative imaging features from axial T2W images and contrast-enhanced T1W images in the training set. The Kruskal-Wallis test, least absolute shrinkage and selection operator regression, and recursive feature elimination were used for feature selection, where upon a radiomics model was built with a logistic regression (LR) classifier. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve and the accuracy were assessed to evaluate the predictive performance in the training and validation set. The performance of the radiomics model was compared to radiologists' assessment by DeLong test.

RESULTS

The AUC of the radiomics model for the prediction of PLONI was 0.928 in the training set and 0.841 in the validation set. Radiomics model produced better sensitivity than radiologists' assessment (81.1% vs  43.2% in training set, 82.4vs 52.9% in validation set). In all 124 patients, the AUC of the radiomics model was 0.897, while that of radiologists' assessment was 0.674 (p < 0.001, DeLong test).

CONCLUSION

MRI-based radiomics model to predict PLONI in RB patients was shown to be superior to visual assessment with improved sensitivity and AUC, and may serve as a potential tool to guide personalized treatment.

Predicting Disease-Free Survival With Multiparametric MRI-Derived Radiomic Signature in Cervical Cancer Patients Underwent CCRT

Prognostic biomarkers that can reliably predict the disease-free survival (DFS) of locally advanced cervical cancer (LACC) are needed for identifying those patients at high risk for progression, who may benefit from a more aggressive treatment. In the present study, we aimed to construct a multiparametric MRI-derived radiomic signature for predicting DFS of LACC patients who underwent concurrent chemoradiotherapy (CCRT).

Immunotherapy treatment outcome prediction in metastatic melanoma through an automated multi-objective delta-radiomics model

Based on recent studies, immunotherapy led by immune checkpoint inhibitors has significantly improved the patient survival rate and effectively reduced the recurrence risk. However, immunotherapy has different therapeutic effects for different patients, leading to difficulties in predicting the treatment response. Conversely, delta-radiomic features, which measure the difference between pre- and post-treatment through quantitative image features, have proven to be promising descriptors for treatment outcome prediction. Consequently, we developed an effective model termed as the automated multi-objective delta-radiomics (Auto-MODR) model for the prediction of immunotherapy response in metastatic melanoma. In Auto-MODR, delta-radiomic features and traditional radiomic features were used as inputs. Furthermore, a novel automated multi-objective model was developed to obtain more reliable and balanced results between sensitivity and specificity. We conducted extensive comparisons with existing studies on treatment outcome prediction. Our method achieved an area under the curve (AUC) of 0.86 in a cross-validation study and an AUC of 0.73 in an independent study. Compared with the model using conventional radiomic features (pre- and post-treatment) only, better performance can be obtained when conventional radiomic and delta-radiomic features are combined. Furthermore, Auto-MODR outperformed the currently available radiomic strategies.

Radiomic Score as a Potential Imaging Biomarker for Predicting Survival in Patients With Cervical Cancer

OBJECTIVES

Accurate prediction of prognosis will help adjust or optimize the treatment of cervical cancer and benefit the patients. We aimed to investigate the incremental value of radiomics when added to the FIGO stage in predicting overall survival (OS) in patients with cervical cancer.

METHODS

This retrospective study included 106 patients with cervical cancer (FIGO stage IB1-IVa) between October 2017 and May 2019. Patients were randomly divided into a training cohort (n = 74) and validation cohort (n = 32). All patients underwent contrast-enhanced computed tomography (CT) prior to treatment. The ITK-SNAP software was used to delineate the region of interest on pre-treatment standard-of-care CT scans. We extracted 792 two-dimensional radiomic features by the Analysis Kit (AK) software. Pearson correlation coefficient analysis and Relief were used to detect the most discriminatory features. The radiomic signature (i.e., Radscore) was constructed via Adaboost with Leave-one-out cross-validation. Prognostic models were built by Cox regression model using Akaike information criterion (AIC) as the stopping rule. A nomogram was established to individually predict the OS of patients. Patients were then stratified into high- and low-risk groups according to the Youden index. Kaplan-Meier curves were used to compare the survival difference between the high- and low-risk groups.

RESULTS

Six textural features were identified, including one gray-level co-occurrence matrix feature and five gray-level run-length matrix features. Only the FIGO stage and Radscore were independent risk factors associated with OS (p < 0.05). The C-index of the FIGO stage in the training and validation cohorts was 0.703 (95% CI: 0.572-0.834) and 0.700 (95% CI: 0.526-0.874), respectively. Correspondingly, the C-index of Radscore was 0.794 (95% CI: 0.707-0.880) and 0.754 (95% CI: 0.623-0.885). The incorporation of the FIGO stage and Radscore achieved better performance, with a C-index of 0.830 (95% CI: 0.738-0.922) and 0.772 (95% CI: 0.615-0.929), respectively. The nomogram based on the FIGO stage and Radscore could individually predict the OS probability with good discrimination and calibration. The high-risk patients had shorter OS compared with the low-risk patients (p < 0.05).

CONCLUSION

Radiomics has the potential for noninvasive risk stratification and may improve the prediction of OS in patients with cervical cancer when added to the FIGO stage.

Combination of Radiomics and Machine Learning with Diffusion-Weighted MR Imaging for Clinical Outcome Prognostication in Cervical Cancer

Objectives: To explore the potential of Radiomics alone and in combination with a diffusion-weighted derived quantitative parameter, namely the apparent diffusion co-efficient (ADC), using supervised classification algorithms in the prediction of outcomes and prognosis. Materials and Methods: Retrospective evaluation of the imaging was conducted for a study cohort of uterine cervical cancer, candidates for radical treatment with chemo radiation. ADC values were calculated from the darkest part of the tumor, both before (labeled preADC) and post treatment (labeled postADC) with chemo radiation. Post extraction of 851 Radiomics features and feature selection analysis—by taking the union of the features that had Pearson correlation >0.35 for recurrence, >0.49 for lymph node and >0.40 for metastasis—was performed to predict clinical outcomes. Results: The study enrolled 52 patients who presented with variable FIGO stages in the age range of 28–79 (Median = 53 years) with a median follow-up of 26.5 months (range: 7–76 months). Disease recurrence occurred in 12 patients (23%). Metastasis occurred in 15 patients (28%). A model generated with 24 radiomics features and preADC using a monotone multi-layer perceptron neural network to predict the recurrence yields an AUC of 0.80 and a Kappa value of 0.55 and shows that the addition of radiomics features to ADC values improves the statistical metrics by approximately 40% for AUC and approximately 223% for Kappa. Similarly, the neural network model for prediction of metastasis returns an AUC value of 0.84 and a Kappa value of 0.65, thus exceeding performance expectations by approximately 25% for AUC and approximately 140% for Kappa. There was a significant input of GLSZM features (SALGLE and LGLZE) and GLDM features (SDLGLE and DE) in correlation with clinical outcomes of recurrence and metastasis. Conclusions: The study is an effort to bridge the unmet need of translational predictive biomarkers in the stratification of uterine cervical cancer patients based on prognosis.

Prediction of lymphovascular space invasion using a combination of tenascin-C, cox-2, and PET/CT radiomics in patients with early-stage cervical squamous cell carcinoma

BACKGROUND

Lymphovascular space invasion is an independent prognostic factor in early-stage cervical cancer. However, there is a lack of non-invasive methods to detect lymphovascular space invasion. Some researchers found that Tenascin-C and Cyclooxygenase-2 was correlated with lymphovascular space invasion. Radiomics has been studied as an emerging tool for distinguishing tumor pathology stage, evaluating treatment response, and predicting prognosis. This study aimed to establish a machine learning model that combines radiomics based on PET imaging with tenascin-C (TNC) and cyclooxygenase-2 (COX-2) for predicting lymphovascular space invasion (LVSI) in patients with early-stage cervical cancer.

METHODS

One hundred and twelve patients with early-stage cervical squamous cell carcinoma who underwent PET/CT examination were retrospectively analyzed. Four hundred one radiomics features based on PET/CT images were extracted and integrated into radiomics score (Rad-score). Immunohistochemical analysis was performed to evaluate TNC and COX-2 expression. Mann-Whitney U test was used to distinguish differences in the Rad-score, TNC, and COX-2 between LVSI and non-LVSI groups. The correlations of characteristics were tested by Spearman analysis. Machine learning models including radiomics model, protein model and combined model were established by logistic regression algorithm and evaluated by ROC curve. Pairwise comparisons of ROC curves were tested by DeLong test.

RESULTS

The Rad-score of patients with LVSI was significantly higher than those without. A significant correlation was shown between LVSI and Rad-score (r = 0.631, p < 0.001). TNC was correlated to both the Rad-score (r = 0.244, p = 0.024) and COX-2 (r = 0.227, p = 0.036). The radiomics model had the best predictive performance among all models in training and external dataset (AUCs: 0.914, 0.806, respectively, p < 0.001). However, in testing dataset, the combined model had better efficiency for predicting LVSI than other models (AUCs: 0.801 vs. 0.756 and 0.801 vs. 0.631, respectively).

CONCLUSION

The machine learning model of the combination of PET radiomics with COX-2 and TNC provides a new tool for detecting LVSI in patients with early-stage cervical cancer. In the future, multicentric studies on larger sample of patients will be used to test the model.

TRIAL REGISTRATION

This is a retrospective study and there is no experimental intervention on human participants. The Ethics Committee has confirmed that retrospectively registered is not required.

Radiomics in cervical and endometrial cancer

Radiomics is an emerging field of research that aims to find associations between quantitative information extracted from imaging examinations and clinical data to support the best clinical decision. In the last few years, some papers have been evaluating the role of radiomics in gynecological malignancies, mainly focusing on ovarian cancer. Nonetheless, cervical cancer is the most frequent gynecological malignancy in developing countries and endometrial cancer is the most common in western countries. The purpose of this narrative review is to give an overview of the latest published papers evaluating the role of radiomics in cervical and endometrial cancer, mostly evaluating association with tumor prognostic factors, with response to therapy and with prediction of recurrence and distant metastasis.

A Systematic Review of PET Textural Analysis and Radiomics in Cancer

Background: Although many works have supported the utility of PET radiomics, several authors have raised concerns over the robustness and replicability of the results. This study aimed to perform a systematic review on the topic of PET radiomics and the used methodologies. Methods: PubMed was searched up to 15 October 2020. Original research articles based on human data specifying at least one tumor type and PET image were included, excluding those that apply only first-order statistics and those including fewer than 20 patients. Each publication, cancer type, objective and several methodological parameters (number of patients and features, validation approach, among other things) were extracted. Results: A total of 290 studies were included. Lung (28%) and head and neck (24%) were the most studied cancers. The most common objective was prognosis/treatment response (46%), followed by diagnosis/staging (21%), tumor characterization (18%) and technical evaluations (15%). The average number of patients included was 114 (median = 71; range 20–1419), and the average number of high-order features calculated per study was 31 (median = 26, range 1–286). Conclusions: PET radiomics is a promising field, but the number of patients in most publications is insufficient, and very few papers perform in-depth validations. The role of standardization initiatives will be crucial in the upcoming years.

PET-CT radiomics by integrating primary tumor and peritumoral areas predicts E-cadherin expression and correlates with pelvic lymph node metastasis in early-stage cervical cancer

OBJECTIVES

To explore the role of radiomics in integrating primary tumor and peritumoral areas based on PET-CT scans for predicting E-cadherin (E-cad) expression in early-stage cervical cancer (ESCC) and its correlation with pelvic lymph node metastasis (PLNM).

METHODS

Ninety-seven ESCC patients who had undergone PET-CT scans were retrospectively analyzed. The ROI of primary tumors, peritumoral areas, and plus tumors were semi-automatically segmented on PET-CT images. A total of 1188 radiomics features were extracted, selected, and eventually integrated into radiomics score (rad-score). The rad-score difference between patients with E-cad expression of high and low was analyzed using Mann-Whitney tests. Characteristic correlation was tested using a Spearman analysis. Four models were established using logistic regression algorithms and evaluated using ROC and calibration curves. A DeLong test was used to perform pairwise comparisons of AUCs.

RESULTS

The rad-score of patients with low E-cad expression was higher than that of patients with high E-cad expression in both training and testing cohorts (p < 0.001 and p = 0.027, respectively). A significant correlation was observed between the rad-score and E-cad (p < 0.001). PLNM correlated slightly with rad-score and E-cad values (p = 0.01 and p < 0.001, respectively). The ROC curve and calibration curve of the rad-score model performed best in both training and testing cohorts (AUC = 0.915, 0.844, p < 0.001, respectively).

CONCLUSIONS

The radiomics of integrating primary tumor and peritumoral areas based on PET-CT showed correlations with PLNM. It was also able to predict E-cad expression in ESCC patients, allowing for evaluation of those patients' prognosis and more individualized medical treatment.

KEY POINTS

• By integrating the primary tumor and peritumoral area based on PET-CT, radiomics was feasible. • The rad-score was associated with E-cad expression and PLNM in patients with ESCC. • Radiomics that integrated the primary tumor and peritumoral areas based on PET-CT could predict E-cad expression in patients with ESCC.

A Machine Learning Tool to Predict the Response to Neoadjuvant Chemotherapy in Patients with Locally Advanced Cervical Cancer

Despite several studies having identified factors associated with successful treatment outcomes in locally advanced cervical cancer, there is the lack of accurate predictive modeling for progression-free survival (PFS) in patients who undergo radical hysterectomy after neoadjuvant chemotherapy (NACT). Here we investigated whether machine learning (ML) may have the potential to provide a tool to predict neoadjuvant treatment response as PFS. In this retrospective observational study, we analyzed patients with locally advanced cervical cancer (FIGO stages IB2, IB3, IIA1, IIA2, IIB, and IIIC1) who were followed in a tertiary center from 2010 to 2018. Demographic and clinical characteristics were collected at either treatment baseline or at 24-month follow-up. Furthermore, we recorded data about magnetic resonance imaging (MRI) examinations and post-surgery histopathology. Proper feature selection was used to determine an attribute core set. Three different machine learning algorithms, namely Logistic Regression (LR), Random Forest (RFF), and K-nearest neighbors (KNN), were then trained and validated with 10-fold cross-validation to predict 24-month PFS. Our analysis included n. 92 patients. The attribute core set used to train machine learning algorithms included the presence/absence of fornix infiltration at pre-treatment MRI as well as of either parametrium invasion and lymph nodes involvement at post-surgery histopathology. RFF showed the best performance (accuracy 82.4%, precision 83.4%, recall 96.2%, area under receiver operating characteristic curve (AUROC) 0.82). We developed an accurate ML model to predict 24-month PFS.

A Phantom Study to Investigate Robustness and Reproducibility of Grey Level Co-Occurrence Matrix (GLCM)-Based Radiomics Features for PET

Quantification and classification of heterogeneous radiotracer uptake in Positron Emission Tomography (PET) using textural features (termed as radiomics) and artificial intelligence (AI) has the potential to be used as a biomarker of diagnosis and prognosis. However, textural features have been predicted to be strongly correlated with volume, segmentation and quantization, while the impact of image contrast and noise has not been assessed systematically. Further continuous investigations are required to update the existing standardization initiatives. This study aimed to investigate the relationships between textural features and these factors with 18F filled torso NEMA phantom to yield different contrasts and reconstructed with different durations to represent varying levels of noise. The phantom was also scanned with heterogeneous spherical inserts fabricated with 3D printing technology. All spheres were delineated using: (1) the exact boundaries based on their known diameters; (2) 40% fixed; and (3) adaptive threshold. Six textural features were derived from the gray level co-occurrence matrix (GLCM) using different quantization levels. The results indicate that homogeneity and dissimilarity are the most suitable for measuring PET tumor heterogeneity with quantization 64 provided that the segmentation method is robust to noise and contrast variations. To use these textural features as prognostic biomarkers, changes in textural features between baseline and treatment scans should always be reported along with the changes in volumes.

Preoperative Predicting the WHO/ISUP Nuclear Grade of Clear Cell Renal Cell Carcinoma by Computed Tomography-Based Radiomics Features

Nuclear grade is important for treatment selection and prognosis in patients with clear cell renal cell carcinoma (ccRCC). This study aimed to determine the ability of preoperative four-phase multiphasic multidetector computed tomography (MDCT)-based radiomics features to predict the WHO/ISUP nuclear grade. In all 102 patients with histologically confirmed ccRCC, the training set (n = 62) and validation set (n = 40) were randomly assigned. In both datasets, patients were categorized according to the WHO/ISUP grading system into low-grade ccRCC (grades 1 and 2) and high-grade ccRCC (grades 3 and 4). The feature selection process consisted of three steps, including least absolute shrinkage and selection operator (LASSO) regression analysis, and the radiomics scores were developed using 48 radiomics features (10 in the unenhanced phase, 17 in the corticomedullary (CM) phase, 14 in the nephrographic (NP) phase, and 7 in the excretory phase). The radiomics score (Rad-Score) derived from the CM phase achieved the best predictive ability, with a sensitivity, specificity, and an area under the curve (AUC) of 90.91%, 95.00%, and 0.97 in the training set. In the validation set, the Rad-Score derived from the NP phase achieved the best predictive ability, with a sensitivity, specificity, and an AUC of 72.73%, 85.30%, and 0.84. We constructed a complex model, adding the radiomics score for each of the phases to the clinicoradiological characteristics, and found significantly better performance in the discrimination of the nuclear grades of ccRCCs in all MDCT phases. The highest AUC of 0.99 (95% CI, 0.92-1.00, p < 0.0001) was demonstrated for the CM phase. Our results showed that the MDCT radiomics features may play a role as potential imaging biomarkers to preoperatively predict the WHO/ISUP grade of ccRCCs.

18F-FDG PET/CT Habitat Radiomics Predicts Outcome of Patients with Cervical Cancer Treated with Chemoradiotherapy

PURPOSE

To determine if quantitative features extracted from pretherapy fluorine 18 fluorodeoxyglucose (18F-FDG) PET/CT estimate prognosis in patients with locally advanced cervical cancer treated with chemoradiotherapy.

MATERIALS AND METHODS

In this retrospective study, PET/CT images and outcomes were curated from 154 patients with locally advanced cervical cancer, who underwent chemoradiotherapy from two institutions between March 2008 and June 2016, separated into independent training (n = 78; mean age, 51 years ± 13 [standard deviation]) and testing (n = 76; mean age, 50 years ± 10) cohorts. Radiomic features were extracted from PET, CT, and habitat (subregions with different metabolic characteristics) images that were derived by fusing PET and CT images. Parsimonious sets of these features were identified by the least absolute shrinkage and selection operator analysis and used to generate predictive radiomics signatures for progression-free survival (PFS) and overall survival (OS) estimation. Prognostic validation of the radiomic signatures as independent prognostic markers was performed using multivariable Cox regression, which was expressed as nomograms, together with other clinical risk factors.

RESULTS

The radiomics nomograms constructed with T stage, lymph node status, and radiomics signatures resulted in significantly better performance for the estimation of PFS (Harrell concordance index [C-index], 0.85 for training and 0.82 for test) and OS (C-index, 0.86 for training and 0.82 for test) compared with International Federation of Gynecology and Obstetrics staging system (C-index for PFS, 0.70 for training [P = .001] and 0.70 for test [P = .002]; C-index for OS, 0.73 for training [P < .001] and 0.70 for test [P < .001]), respectively.

CONCLUSION

Prognostic models were generated and validated from quantitative analysis of 18F-FDG PET/CT habitat images and clinical data, and may have the potential to identify the patients who need more aggressive treatment in clinical practice, pending further validation with larger prospective cohorts.Supplemental material is available for this article.© RSNA, 2020.

Noninvasive prediction of lymph node status for patients with early-stage cervical cancer based on radiomics features from ultrasound images

OBJECTIVE

To investigate the feasibility of a noninvasive detection of lymph node metastasis (LNM) for early-stage cervical cancer (ECC) patients with radiomics methods based on the textural features from ultrasound images.

METHODS

One hundred seventy-two ECC patients between January 2014 and September 2018 with pathologically confirmed lymph node status (LNS) and preoperative ultrasound images were retrospectively reviewed. Regions of interest (ROIs) were delineated by a senior radiologist in the ultrasound images. LIFEx was applied to extract textural features for radiomics study. Least absolute shrinkage and selection operator (LASSO) regression was applied for dimension reduction and for selection of key features. A multivariable logistic regression analysis was adopted to build the radiomics signature. The Mann-Whitney U test was applied to investigate the correlation between radiomics and LNS for both training and validation cohorts. Receiver operating characteristic (ROC) curves were applied to evaluate the accuracy of the radiomics prediction models.

RESULTS

A total of 152 radiomics features were extracted from ultrasound images, in which 6 features were significantly associated with LNS (p < 0.05). The radiomics signatures demonstrated a good discrimination between patients with LNM and non-LNM groups. The best radiomics performance model achieved an area under the curve (AUC) of 0.79 (95% confidence interval (CI), 0.71-0.88) in the training cohort and 0.77 (95% CI, 0.65-0.88) in the validation cohort.

CONCLUSIONS

The feasibility of radiomics features from ultrasound images for the prediction of LNM in ECC was investigated. This noninvasive prediction method may be used to facilitate preoperative identification of LNS in patients with ECC.

KEY POINTS

• Few studied had investigated the feasibility of radiomics based on ultrasound images for cervical cancer, even though it is the most common practice for gynecological cancer diagnosis and treatment. • The radiomics signatures based on ultrasound images demonstrated a good discrimination between patients with and without lymph node metastasis with an area under the curve (AUC) of 0.79 and 0.77 in the training and validation cohorts, respectively. • The radiomics model based on preoperative ultrasound images has the potential ability to predict lymph node status noninvasively in patients with early-state cervical cancer, so as to reduce the impact of invasive examination and to optimize the treatment choices.

18F-FDG PET/CT Quantitative Parameters and Texture Analysis Effectively Differentiate Endometrial Precancerous Lesion and Early-Stage Carcinoma

Objective:

This study evaluated the metabolic parameters and texture features of fluorodeoxyglucose positron emission tomography–computed tomography (PET/CT) for the diagnosis and differentiation of endometrial atypical hyperplasia (EAH), EAH with field cancerization (FC), and stage 1A endometrial carcinoma (EC 1a).

Materials and Methods:

We retrospectively analyzed the metabolic parameters of PET/CT in 170 patients with diagnoses confirmed by pathology, including 57 cases of EAH (57/170, 33.53%), 45 cases of FC (45/170, 26.47%), and 68 cases of EC 1a (68/170, 40.0%). Then, the texture features of each tumor were extracted and compared with the metabolic parameters and pathological results using nonparametric tests and linear regression analysis. The diagnostic performance was assessed by the area under the curve (AUC) values obtained from receiver operating characteristic analysis.

Results:

There were moderate positive correlations between the PET standardized uptake values (SUVpeak, SUVmax, and SUVmean) and postoperative pathological features with correlation coefficients (r_s) of 0.663, 0.651, and 0.651, respectively (P < .001). Total lesion glycolysis showed relatively low correlation with pathological characteristics (r_s = 0.476), whereas metabolic tumor volume and age showed the weakest correlations (r_s = 0.186 and 0.232, respectively). To differentiate between the diagnosis of EAH and FC, SUVmax displayed the largest AUC of 0.857 (sensitivity, 82.2%; specificity, 84.2%). Five texture features were screened out as Percentile 40, Percentile 45, InverseDifferenceMoment_AllDirection_offset 1, InverseDifferenceMoment_angle 45_offset 4, and ClusterProminence_angle 135_offset 7 (P < .001) by linear model of texture analysis (AUC = 0.851; specificity = 0.692; sensitivity = 0.871). To differentiate between the diagnoses of FC and EC 1a, SUVpeak displayed the largest AUC of 0.715 (sensitivity, 67.6%; specificity, 77.8%), and 2 texture features were identified as Percentile 10 and CP_angle 135_offset 7 (AUC = 0.819; specificity = 0.871; sensitivity = 0.766; P < .001).

Conclusions:

SUVmax and SUVpeak had the highest diagnostic values for EAH, FC, and EC 1a compared with the other tested parameters. SUVmax, Percentile 40, Percentile 45, InverseDifferenceMoment_AllDirection_offset 1, InverseDifferenceMoment_angle 45_offset 4, and ClusterProminence_angle 135_offset 7 distinguished EAH from FC. SUVpeak, Percentile 10, and ClusterProminence_angle 135_offset 7 distinguished FC from EC 1a. This study showed that the addition of texture features provides valuable information for differentiating EAH, FC, and EC 1a diagnoses.

Diagnosis of Cervical Cancer With Parametrial Invasion on Whole-Tumor Dynamic Contrast-Enhanced Magnetic Resonance Imaging Combined With Whole-Lesion Texture Analysis Based on T2- Weighted Images

Purpose: To evaluate the diagnostic value of the combination of whole-tumor dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and whole-lesion texture features based on T2-weighted images for cervical cancer with parametrial invasion. Materials and Methods: Sixty-two patients with cervical cancer (27 with parametrial invasion and 35 without invasion) preoperatively underwent routine MRI and DCE-MRI examinations. DCE-MRI parameters (K^trans, K_ep, and V_e) and texture features (mean, skewness, kurtosis, uniformity, energy, and entropy) based on T2-weighted images were acquired by two observers. All parameters of parametrial invasion and non-invasion were analyzed by one-way analysis of variance. The diagnostic efficiency of significant variables was assessed using receiver operating characteristic analysis. Results: The invasion group of cervical cancer demonstrated significantly higher K^trans (0.335 ± 0.050 vs. 0.269 ± 0.079; p < 0.001), lower energy values (0.503 ± 0.093 vs. 0.602 ± 0.087; p < 0.001), and higher entropy values (1.391 ± 0.193 vs. 1.24 ± 0.129; p < 0.001) than those in the non-invasion group. Optimal diagnostic performance [area under curve [AUC], 0.925; sensitivity, 0.935; specificity, 0.829] could be obtained by the combination of K^trans, energy, and entropy values. The AUC values of K^trans (0.788), energy (0.761), entropy (0.749), the combination of K^trans and energy (0.814), the combination of K^trans and entropy (0.727), and the combination of energy and entropy (0.619) were lower than those of the combination of K^trans, energy, and entropy values. Conclusion: The combination of DCE-MRI and texture analysis is a promising method for diagnosis cervical cancer with parametrial infiltration. Moreover, the combination of K^trans, energy, and entropy is more valuable than any one alone, especially in improving diagnostic sensitivity.

MRI texture features differentiate clinicopathological characteristics of cervical carcinoma

OBJECTIVES

To evaluate MRI texture analysis in differentiating clinicopathological characteristics of cervical carcinoma (CC).

METHODS

Patients with newly diagnosed CC who underwent pre-treatment MRI were retrospectively reviewed. Texture analysis was performed using commercial software (TexRAD). Largest single-slice ROIs were manually drawn around the tumour on T2-weighted (T2W) images, apparent diffusion coefficient (ADC) maps and contrast-enhanced T1-weighted (T1c) images. First-order texture features were calculated and compared among histological subtypes, tumour grades, FIGO stages and nodal status using the Mann-Whitney U test. Feature selection was achieved by elastic net. Selected features from different sequences were used to build the multivariable support vector machine (SVM) models and the performances were assessed by ROC curves and AUC.

RESULTS

Ninety-five patients with FIGO stage IB~IVB were evaluated. A number of texture features from multiple sequences were significantly different among all the clinicopathological subgroups (p < 0.05). Texture features from different sequences were selected to build the SVM models. The AUCs of SVM models for discriminating histological subtypes, tumour grades, FIGO stages and nodal status were 0.841, 0.850, 0.898 and 0.879, respectively.

CONCLUSIONS

Texture features derived from multiple sequences were helpful in differentiating the clinicopathological signatures of CC. The SVM models with selected features from different sequences offered excellent diagnostic discrimination of the tumour characteristics in CC.

KEY POINTS

• First-order texture features are able to differentiate clinicopathological signatures of cervical carcinoma. • Combined texture features from different sequences can offer excellent diagnostic discrimination of the tumour characteristics in cervical carcinoma.

Prediction of different stages of rectal cancer: Texture analysis based on diffusion-weighted images and apparent diffusion coefficient maps

BACKGROUND

It is evident that an accurate evaluation of T and N stage rectal cancer is essential for treatment planning. It has not been extensively investigated whether texture features derived from diffusion-weighted imaging (DWI) images and apparent diffusion coefficient (ADC) maps are associated with the extent of local invasion (pathological stage T1-2 vs T3-4) and nodal involvement (pathological stage N0 vs N1-2) in rectal cancer.

AIM

To predict different stages of rectal cancer using texture analysis based on DWI images and ADC maps.

METHODS

One hundred and fifteen patients with pathologically proven rectal cancer, who underwent preoperative magnetic resonance imaging, including DWI, were enrolled, retrospectively. The ADC measurements (ADC_mean, ADC_min, ADC_max) as well as texture features, including the gray level co-occurrence matrix parameters, the gray level run-length matrix parameters and wavelet parameters were calculated based on DWI (b = 0 and b = 1000) images and the ADC maps. Independent sample t-tests or Mann-Whitney U tests were used for statistical analysis. Multivariate logistic regression analysis was conducted to establish the models. The predictive performance was validated by receiver operating characteristic curve analysis.

 

RESULTS

Dissimilarity, sum average, information correlation and run-length nonuniformity from DWI_b=0 images, gray level nonuniformity, run percentage and run-length nonuniformity from DWI_b=1000 images, and dissimilarity and run percentage from ADC maps were found to be independent predictors of local invasion (stage T3-4). The area under the operating characteristic curve of the model reached 0.793 with a sensitivity of 78.57% and a specificity of 74.19%. Sum average, gray level nonuniformity and the horizontal components of symlet transform (SymletH) from DWI_b=0 images, sum average, information correlation, long run low gray level emphasis and SymletH from DWI_b=1000 images, and ADC_max, ADC_mean and information correlation from ADC maps were identified as independent predictors of nodal involvement. The area under the operating characteristic curve of the model reached 0.802 with a sensitivity of 80.77% and a specificity of 68.25%.

CONCLUSION

Texture features extracted from DWI images and ADC maps are useful clues for predicting pathological T and N stages in rectal cancer.

Preoperative CT texture features predict prognosis after curative resection in pancreatic cancer

Pancreatic cancer is a lethal disease, and resistance to chemotherapy is a critical factor influencing the postoperative prognosis. Tumour heterogeneity is an important indicator of chemoresistance. Therefore, we analysed tumour heterogeneity in preoperative computed tomography scans by performing texture analysis using the grey-level run-length matrix and analysed the correlation of survival with the value obtained in these analyses. We analysed 116 consecutive patients who underwent curative resection and had preoperative contrast-enhanced computed tomography data available for analysis. A region of interest was drawn on all slices with a visible tumour and normal pancreas on the arterial phase computed tomography scans; the correlation of pathological characteristics with grey-level run-length matrix features was analysed. We then performed Kaplan–Meier survival curve analysis among pancreatic cancer patients. The grey-level non-uniformity values in grey-level run-length matrix features for tumours were higher than those for normal pancreas. High grey-level non-uniformity values represent a non-uniform texture, i.e., heterogeneity. Grey-level run-length matrix features showed that recurrence-free survival was shorter in the group with high grey-level non-uniformity 135 values (p = 0.025). Our analyses of the correlation between pathological outcomes and grey-level run-length matrix features in pancreatic cancer patients showed that grey-level non-uniformity values were powerful prognostic indicators.

Assessment of Stability and Discrimination Capacity of Radiomic Features on Apparent Diffusion Coefficient Images

The objectives of the study are to develop a new way to assess stability and discrimination capacity of radiomic features without the need of test-retest or multiple delineations and to use information obtained to perform a preliminary feature selection. Apparent diffusion coefficient (ADC) maps were computed from diffusion-weighted magnetic resonance images (DW-MRI) of two groups of patients: 18 with soft tissue sarcomas (STS) and 18 with oropharyngeal cancers (OPC). Sixty-nine radiomic features were computed, using three different histogram discretizations (16, 32, and 64 bins). Geometrical transformations (translations) of increasing entity were applied to the regions of interest (ROIs), and the intra-class correlation coefficient (ICC) was used to compare the features computed on the original and modified ROIs. The distribution of ICC values for minimal and maximal entity translations (ICC₁₀ and ICC₁₀₀, respectively) was used to adjust thresholds of ICC (ICC_min and ICC_max) used to discriminate between good, unstable (ICC₁₀ < ICC_min), and non-discriminative features (ICC₁₀₀ > ICC_max). Fifty-four and 59 radiomic features passed the stability-based selection for all the three histogram discretizations for the OPC and STS datasets, respectively. The excluded features were similar across the different histogram discretizations (Jaccard’s index 0.77 ± 0.13 and 0.9 ± 0.1 for OPC and STS, respectively) but different between datasets (Jaccard’s index 0.19 ± 0.02). The results suggest that the observed radiomic features are mainly stable and discriminative, but the stability depends on the region of the body under observation. The method provides a way to assess stability without the need of test-retest or multiple delineations.

Radiomics features on radiotherapy treatment planning CT can predict patient survival in locally advanced rectal cancer patients

This retrospective study was to investigate whether radiomics feature come from radiotherapy treatment planning CT can predict prognosis in locally advanced rectal cancer patients treated with neoadjuvant chemoradiation followed by surgery. Four-hundred-eleven locally advanced rectal cancer patients which were treated with neoadjuvant chemoradiation enrolled in this study. All patients’ radiotherapy treatment planning CTs were collected. Tumor was delineated on these CTs by physicians. An in-house radiomics software was used to calculate 271 radiomics features. The results of test-retest and contour-recontour studies were used to filter stable radiomics (Spearman correlation coefficient > 0.7). Twenty-one radiomics features were final enrolled. The performance of prediction model with the radiomics or clinical features were calculated. The clinical outcomes include local control, distant control, disease-free survival (DFS) and overall survival (OS). Model performance C-index was evaluated by C-index. Patients are divided into two groups by cluster results. The results of chi-square test revealed that the radiomics feature cluster is independent of clinical features. Patients have significant differences in OS (p = 0.032, log rank test) for these two groups. By supervised modeling, radiomics features can improve the prediction power of OS from 0.672 [0.617 0.728] with clinical features only to 0.730 [0.658 0.801]. In conclusion, the radiomics features from radiotherapy CT can potentially predict OS for locally advanced rectal cancer patients with neoadjuvant chemoradiation treatment.

Radiomics in stratification of pancreatic cystic lesions: Machine learning in action

Pancreatic cystic lesions (PCLs) are well-known precursors of pancreatic cancer. Their diagnosis can be challenging as their behavior varies from benign to malignant disease. Precise and timely management of malignant pancreatic cysts might prevent transformation to pancreatic cancer. However, the current consensus guidelines, which rely on standard imaging features to predict cyst malignancy potential, are conflicting and unclear. This has led to an increased interest in radiomics, a high-throughput extraction of comprehensible data from standard of care images. Radiomics can be used as a diagnostic and prognostic tool in personalized medicine. It utilizes quantitative image analysis to extract features in conjunction with machine learning and artificial intelligence (AI) methods like support vector machines, random forest, and convolutional neural network for feature selection and classification. Selected features can then serve as imaging biomarkers to predict high-risk PCLs. Radiomics studies conducted heretofore on PCLs have shown promising results. This cost-effective approach would help us to differentiate benign PCLs from malignant ones and potentially guide clinical decision-making leading to better utilization of healthcare resources. In this review, we discuss the process of radiomics, its myriad applications such as diagnosis, prognosis, and prediction of therapy response. We also discuss the outcomes of studies involving radiomic analysis of PCLs and pancreatic cancer, and challenges associated with this novel field along with possible solutions. Although these studies highlight the potential benefit of radiomics in the prevention and optimal treatment of pancreatic cancer, further studies are warranted before incorporating radiomics into the clinical decision support system.

Prediction of Immunohistochemistry of Suspected Thyroid Nodules by Use of Machine Learning-Based Radiomics

OBJECTIVE. The purpose of this study was to develop and validate a radiomics model for evaluating immunohistochemical characteristics in patients with suspected thyroid nodules. MATERIALS AND METHODS. A total of 103 patients (training cohort-to-validation cohort ratio, ≈ 3:1) with suspected thyroid nodules who had undergone thyroidectomy and immunohistochemical analysis were enrolled. The immunohistochemical markers were cytokeratin 19, galectin 3, thyroperoxidase, and high-molecular-weight cytokeratin. All patients underwent CT before surgery, and a 3D slicer was used to analyze images of the surgical specimen. Test-retest and Spearman correlation coefficient (ρ) were used to select reproducible and nonredundant features. The Kruskal-Wallis test (p < 0.05) was used for feature selection, and a feature-based model was built by support vector machine methods. The performance of the radiomic models was assessed with respect to accuracy, sensitivity, specificity, corresponding AUC, and independent validation. RESULTS. Eighty-six reproducible and nonredundant features selected from the 828 features were used to build the model. The best performance of the cytokeratin 19 model yielded accuracy of 84.4% in the training cohort and 80.0% in the validation cohort. The thyroperoxidase and galectin 3 predictive models yielded accuracies of 81.4% and 82.5% in the training cohort and 84.2% and 85.0% in the validation cohort. The performance of the high-molecular-weight cytokeratin predictive model was not good (accuracy, 65.7%) and could not be validated. CONCLUSION. A radiomics model with excellent performance was developed for individualized noninvasive prediction of the presence of cytokeratin 19, galectin 3, and thyroperoxidase based on CT images. This model may be used to identify benign and malignant thyroid nodules.

Radiomics in nuclear medicine: robustness, reproducibility, standardization, and how to avoid data analysis traps and replication crisis

Radiomics in nuclear medicine is rapidly expanding. Reproducibility of radiomics studies in multicentre settings is an important criterion for clinical translation. We therefore performed a meta-analysis to investigate reproducibility of radiomics biomarkers in PET imaging and to obtain quantitative information regarding their sensitivity to variations in various imaging and radiomics-related factors as well as their inherent sensitivity. Additionally, we identify and describe data analysis pitfalls that affect the reproducibility and generalizability of radiomics studies. After a systematic literature search, 42 studies were included in the qualitative synthesis, and data from 21 were used for the quantitative meta-analysis. Data concerning measurement agreement and reliability were collected for 21 of 38 different factors associated with image acquisition, reconstruction, segmentation and radiomics-specific processing steps. Variations in voxel size, segmentation and several reconstruction parameters strongly affected reproducibility, but the level of evidence remained weak. Based on the meta-analysis, we also assessed inherent sensitivity to variations of 110 PET image biomarkers. SUV_mean and SUV_max were found to be reliable, whereas image biomarkers based on the neighbourhood grey tone difference matrix and most biomarkers based on the size zone matrix were found to be highly sensitive to variations, and should be used with care in multicentre settings. Lastly, we identify 11 data analysis pitfalls. These pitfalls concern model validation and information leakage during model development, but also relate to reporting and the software used for data analysis. Avoiding such pitfalls is essential for minimizing bias in the results and to enable reproduction and validation of radiomics studies.

A machine-learning-based prediction model of fistula formation after interstitial brachytherapy for locally advanced gynecological malignancies

PURPOSE

External beam radiotherapy combined with interstitial brachytherapy is commonly used to treat patients with bulky, advanced gynecologic cancer. However, the high radiation dose needed to control the tumor may result in fistula development. There is a clinical need to identify patients at high risk for fistula formation such that treatment may be managed to prevent this toxic side effect. This work aims to develop a fistula prediction model framework using machine learning based on patient, tumor, and treatment features.

METHODS AND MATERIALS

This retrospective study included 35 patients treated at our institution using interstitial brachytherapy for various gynecological malignancies. Five patients developed rectovaginal fistula and two developed both rectovaginal and vesicovaginal fistula. For each patient, 31 clinical features of multiple data types were collected to develop a fistula prediction framework. A nonlinear support vector machine was used to build the prediction model. Sequential backward feature selection and sequential floating backward feature selection methods were used to determine optimal feature sets. To overcome data imbalance issues, the synthetic minority oversampling technique was used to generate synthetic fistula cases for model training.

RESULTS

Seven mixed data features were selected by both sequential backward selection and sequential floating backward selection methods. Our prediction model using these features achieved a high prediction accuracy, that is, 0.904 area under the curve, 97.1% sensitivity, and 88.5% specificity.

CONCLUSIONS

A machine-learning-based prediction model of fistula formation has been developed for patients with advanced gynecological malignancies treated using interstitial brachytherapy. This model may be clinically impactful pending refinement and validation in a larger series.

Survival outcome prediction in cervical cancer: Cox models vs deep-learning model

BACKGROUND

Historically, the Cox proportional hazard regression model has been the mainstay for survival analyses in oncologic research. The Cox proportional hazard regression model generally is used based on an assumption of linear association. However, it is likely that, in reality, there are many clinicopathologic features that exhibit a nonlinear association in biomedicine.

OBJECTIVE

The purpose of this study was to compare the deep-learning neural network model and the Cox proportional hazard regression model in the prediction of survival in women with cervical cancer.

STUDY DESIGN

This was a retrospective pilot study of consecutive cases of newly diagnosed stage I-IV cervical cancer from 2000-2014. A total of 40 features that included patient demographics, vital signs, laboratory test results, tumor characteristics, and treatment types were assessed for analysis and grouped into 3 feature sets. The deep-learning neural network model was compared with the Cox proportional hazard regression model and 3 other survival analysis models for progression-free survival and overall survival. Mean absolute error and concordance index were used to assess the performance of these 5 models.

RESULTS

There were 768 women included in the analysis. The median age was 49 years, and the majority were Hispanic (71.7%). The majority of tumors were squamous (75.3%) and stage I (48.7%). The median follow-up time was 40.2 months; there were 241 events for recurrence and progression and 170 deaths during the follow-up period. The deep-learning model showed promising results in the prediction of progression-free survival when compared with the Cox proportional hazard regression model (mean absolute error, 29.3 vs 316.2). The deep-learning model also outperformed all the other models, including the Cox proportional hazard regression model, for overall survival (mean absolute error, Cox proportional hazard regression vs deep-learning, 43.6 vs 30.7). The performance of the deep-learning model further improved when more features were included (concordance index for progression-free survival: 0.695 for 20 features, 0.787 for 36 features, and 0.795 for 40 features). There were 10 features for progression-free survival and 3 features for overall survival that demonstrated significance only in the deep-learning model, but not in the Cox proportional hazard regression model. There were no features for progression-free survival and 3 features for overall survival that demonstrated significance only in the Cox proportional hazard regression model, but not in the deep-learning model.

CONCLUSION

Our study suggests that the deep-learning neural network model may be a useful analytic tool for survival prediction in women with cervical cancer because it exhibited superior performance compared with the Cox proportional hazard regression model. This novel analytic approach may provide clinicians with meaningful survival information that potentially could be integrated into treatment decision-making and planning. Further validation studies are necessary to support this pilot study.

Reproducibility and non-redundancy of radiomic features extracted from arterial phase CT scans in hepatocellular carcinoma patients: impact of tumor segmentation variability

Background

The reproducibility and non-redundancy of radiomic features are challenges in accelerating the clinical translation of radiomics. In this study, we focused on the robustness and non-redundancy of radiomic features extracted from computed tomography (CT) scans in hepatocellular carcinoma (HCC) patients with respect to different tumor segmentation methods.

Methods

Arterial enhanced CT images were retrospectively randomly obtained from 106 patients. As a training data set, 26 HCC patients were used to calculate the features' reproducibility and redundancy. Another data set (55 HCC patients and 25 healthy volunteers) was used for classification. The GrowCut and GraphCut semiautomatic segmentation methods were implemented in 3D Slicer software by two independent observers, and manual delineation was performed by five abdominal radiation oncologists to acquire the gross tumor volume (GTV). Seventy-one radiomic features were extracted from GTVs using Imaging Biomarker Explorer (IBEX) software, including 17 tumor intensity statistical features, 16 shape features and 38 textural features. For each radiomic feature, intraclass correlation coefficient (ICC) and hierarchical clustering were used to quantify its reproducibility and redundancy. Features with ICC values greater than 0.75 were considered reproducible. To generate the number of non-redundancy feature subgroups, the R² statistic method was used. Then, a classification model was built using a support vector machine (SVM) algorithm with 10-fold cross validation, and area under ROC curve (AUC) was used to evaluate the utility of non-redundant feature extraction by hierarchical clustering.

Results

The percentages of excellent reproducible features in the manual delineation group, GraphCut and GrowCut segmentation group were 69% [49], 73% [52] and 79% [56], respectively. Sixty-five percent [46] of the features showed strong robustness for all segmentation methods. The optimal number of cluster subgroup were 9, 13 and 11 for manual delineation, GraphCut and GrowCut segmentation, respectively. The optimal cluster subgroup number was 6 for all groups when the collectively high reproducibility features were selected for clustering. The receiver operating characteristic (ROC) analysis of radiomics classification model with and without feature reduction for healthy liver and HCC had an AUC value of 0.857 and 0.721 respectively.

Conclusions

Our study demonstrates that variations exist in the reproducibility of quantitative imaging features extracted from tumor regions segmented using different methods. The reproducibility and non-redundancy of the radiomic features rely greatly on the tumor segmentation in HCC CT images. We recommend that the most reliable and uniform radiomic features should be selected in the clinical use of radiomics. Classification experiments with feature reduction showed that radiomic features were effective in identifying healthy liver and HCC.

Quantitative Identification of Nonmuscle-Invasive and Muscle-Invasive Bladder Carcinomas: A Multiparametric MRI Radiomics Analysis

BACKGROUND

Preoperative discrimination between nonmuscle-invasive bladder carcinomas (NMIBC) and the muscle-invasive ones (MIBC) is very crucial in the management of patients with bladder cancer (BC).

PURPOSE

To evaluate the discriminative performance of multiparametric MRI radiomics features for precise differentiation of NMIBC from MIBC, preoperatively.

STUDY TYPE

Retrospective, radiomics.

POPULATION

Fifty-four patients with postoperative pathologically proven BC lesions (24 in NMIBC and 30 in MIBC groups) were included.

FIELD STRENGTH/SEQUENCE

3.0T MRI/T₂ -weighted (T₂ W) and multi-b-value diffusion-weighted (DW) sequences.

ASSESSMENT

A total of 1104 radiomics features were extracted from carcinomatous regions of interest on T₂ W and DW images, and the apparent diffusion coefficient maps. Support vector machine with recursive feature elimination (SVM-RFE) and synthetic minority oversampling technique (SMOTE) were used to construct an optimal discriminative model, and its performance was evaluated and compared with that of using visual diagnoses by experts.

STATISTICAL TESTS

Chi-square test and Student's t-test were applied on clinical characteristics to analyze the significant differences between patient groups.

RESULTS

Of the 1104 features, an optimal subset involving 19 features was selected from T₂ W and DW sequences, which outperformed the other two subsets selected from T₂ W or DW sequence in muscle invasion discrimination. The best performance for the differentiation task was achieved by the SVM-RFE+SMOTE classifier, with averaged sensitivity, specificity, accuracy, and area under the curve of receiver operating characteristic of 92.60%, 100%, 96.30%, and 0.9857, respectively, which outperformed the diagnostic accuracy by experts.

DATA CONCLUSION

The proposed radiomics approach has potential for the accurate differentiation of muscle invasion in BC, preoperatively. The optimal feature subset selected from multiparametric MR images demonstrated better performance in identifying muscle invasiveness when compared with that from T₂ W sequence or DW sequence only.

LEVEL OF EVIDENCE

3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;49:1489-1498.

Value of [18F]FDG PET radiomic features and VEGF expression in predicting pelvic lymphatic metastasis and their potential relationship in early-stage cervical squamous cell carcinoma

PURPOSE

The purpose of this study was to explore the value of 2-deoxy-2-[18F]fluoro-d-glucose positron emission tomography/computed tomography ([18F]FDG PET/CT) radiomic features combined with vascular endothelial growth factor (VEGF) expression in predicting pelvic lymphatic metastasis in patients with early-stage cervical squamous cell carcinoma and the added value of radiomic features in predicting VEGF expression.

MATERIALS AND METHODS

Ninety-four newly diagnosed cervical squamous cell carcinoma patients (training dataset: n = 64, validation cohort: n = 30) in stage Ia to IIa, according to the International Federation of Gynecology and Obstetrics (FIGO) staging system, who underwent [¹⁸F]FDG PET/CT were retrospectively analyzed. Radiomic features of the [¹⁸F]FDG PET scans were extracted, and the value of the lymph node sizes, metabolic parameters (both tumor and lymph nodes), radiomic features and VEGF expression level in predicting lymphatic metastasis were evaluated by receiver operating characteristics curves (ROC) and were compared using DeLong test. Moreover, we studied the associations between the [¹⁸F]FDG PET radiomic features and VEGF expression.

RESULTS

Total lesion glycolysis (TLG) and the expression of VEGF were significantly higher in subjects with lymphatic metastasis than in those without. The homogeneity feature derived from the histogram, the skewness, had a certain value in predicting lymphatic metastasis (AUC = 0.803 in training dataset, P < 0.05, 95% CI 0.684, 0.892; AUC = 0.757 in validation dataset, P < 0.05, 95% CI 0.545, 0.904). Additionally, the combination of this radiomic feature and VEGF expression had a significantly superior predictive value (AUC = 0.878, P < 0.05, 95% CI 0.772- 0.947), compared to that of the conventional parameters. Moreover, 26 radiomic features derived from the histogram and GLCM features correlated with VEGF expression.

CONCLUSIONS

In patients with early-stage cervical squamous cell carcinoma, PLN metastasis can be predicted by TLG and the textural feature of homogeneity. Radiomic features in combination with the VEGF expression level improved the prediction accuracy. In addition, some features derived from the histogram and gray-level co-occurrence matrices (GLCM) may have a certain value in predicting the VEGF expression level.

Radiomics in Oncological PET/CT: Clinical Applications

18F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) is widely used for staging, evaluating treatment response, and predicting prognosis in malignant diseases. FDG uptake and volumetric PET parameters such as metabolic tumor volume have been used and are still used as conventional PET parameters to assess biological characteristics of tumors. However, in recent years, additional features derived from PET images by computational processing have been found to reflect intratumoral heterogeneity, which is related to biological tumor features, and to provide additional predictive and prognostic information, which leads to the concept of radiomics. In this review, we focus on recent clinical studies of malignant diseases that investigated intratumoral heterogeneity on PET/CT, and we discuss its clinical role in various cancers.

Prediction of cervical cancer recurrence using textural features extracted from 18F-FDG PET images acquired with different scanners

Objectives

To identify an imaging signature predicting local recurrence for locally advanced cervical cancer (LACC) treated by chemoradiation and brachytherapy from baseline ¹⁸F-FDG PET images, and to evaluate the possibility of gathering images from two different PET scanners in a radiomic study.

Methods

118 patients were included retrospectively. Two groups (G1, G2) were defined according to the PET scanner used for image acquisition. Eleven radiomic features were extracted from delineated cervical tumors to evaluate: (i) the predictive value of features for local recurrence of LACC, (ii) their reproducibility as a function of the scanner within a hepatic reference volume, (iii) the impact of voxel size on feature values.

Results

Eight features were statistically significant predictors of local recurrence in G1 (p < 0.05). The multivariate signature trained in G2 was validated in G1 (AUC=0.76, p<0.001) and identified local recurrence more accurately than SUV_max (p=0.022). Four features were significantly different between G1 and G2 in the liver. Spatial resampling was not sufficient to explain the stratification effect.

Conclusion

This study showed that radiomic features could predict local recurrence of LACC better than SUV_max. Further investigation is needed before applying a model designed using data from one PET scanner to another.