Challenges and caveats of a multi-center retrospective radiomics study: an example of early treatment response assessment for NSCLC patients using FDG-PET/CT radiomics (original) (raw)
Related papers
2021
Background Radiomics is a promising tool for identifying imaging-based biomarkers. Radiomics-based models are often trained on single-institution datasets; however, multi-centre imaging datasets are preferred for external generalizability owing to the influence of inter-institutional scanning differences and acquisition settings. The study aim was to determine the value of preselection of robust radiomic features in routine clinical positron emission tomography (PET) images to predict clinical outcomes in locally advanced non-small cell lung cancer (NSCLC). Methods A total of 1404 primary tumour radiomic features were extracted from pre-treatment [ 18 F]fluorodeoxyglucose (FDG)-PET scans of stage IIIA/N2 or IIIB NSCLC patients using a training cohort ( n = 79; prospective Swiss multi-centre randomized phase III trial SAKK 16/00; 16 centres) and an internal validation cohort ( n = 31; single centre). Robustness studies investigating delineation variation, attenuation correction and...
OncoImmunology, 2022
To develop a short-term follow-up CT-based radiomics approach to predict response to immunotherapy in advanced non-small-cell lung cancer (NSCLC) and investigate the prognostic value of radiomics features in predicting progression-free survival (PFS) and overall survival (OS). We first retrospectively collected 224 advanced NSCLC patients from two centers, and divided them into a primary cohort and two validation cohorts respectively. Then, we processed CT scans with a series of image preprocessing techniques namely, tumor segmentation, image resampling, feature extraction and normalization. To select the optimal features, we applied the feature ranking with recursive feature elimination method. After resampling the training dataset with a synthetic minority oversampling technique, we applied the support vector machine classifier to build a machine-learning-based classification model to predict response to immunotherapy. Finally, we used Kaplan-Meier (KM) survival analysis method to evaluate prognostic value of rad-score generated by CT-radiomics model. In two validation cohorts, the delta-radiomics model significantly improved the area under receiver operating characteristic curve from 0.64 and 0.52 to 0.82 and 0.87, respectively (P < .05). In subgroup analysis, pre-and delta-radiomics model yielded higher performance for adenocarcinoma (ADC) patients than squamous cell carcinoma (SCC) patients. Through the KM survival analysis, the rad-score of delta-radiomics model had a significant prognostic for PFS and OS in validation cohorts (P < .05). Our results demonstrated that (1) delta-radiomics model could improve the prediction performance, (2) radiomics model performed better on ADC patients than SCC patients, (3) delta-radiomics model had prognostic values in predicting PFS and OS of NSCLC patients.
Journal of nuclear medicine : official publication, Society of Nuclear Medicine, 2018
The optimal methodology for defining response with F-fluorodeoxyglucose positron emission tomography (FDG-PET) after curative-intent chemoradiation for non-small cell lung cancer (NSCLC) is unknown. We compared survival outcomes according to European Organization for Research and Treatment of Cancer (EORTC), Positron Emission tomography Response Criteria In Solid Tumors 1.0 (PERCIST), Peter Mac Metabolic Visual and Deauville criteria respectively. Three prospective trials of chemoradiation for NSCLC, involving baseline and post-treatment FDG-PET/Computer Tomography (CT) imaging, were conducted between 2004 and 2016. Responses were categorized as complete metabolic response (CMR), partial metabolic response, stable metabolic disease or progressive metabolic disease. Cox proportional hazard models and logrank tests assessed the impact of each response on overall survival (OS). Eighty-seven patients underwent FDG-PET/CT before and after radical chemoradiation for NSCLC. Follow-up FDG-P...
Cancers
The aim of this study is to identify clinically relevant image feature (IF) changes during chemoradiation and evaluate their efficacy in predicting treatment response. Patients with non-small-cell lung cancer (NSCLC) were enrolled in two prospective trials (STRIPE, PET-Plan). We evaluated 48 patients who underwent static (3D) and retrospectively-respiratory-gated 4D PET/CT scans before treatment and a 3D scan during or after treatment. Our proposed method rejects IF changes due to intrinsic variability. The IF variability observed across 4D PET is employed as a patient individualized normalization factor to emphasize statistically relevant IF changes during treatment. Predictions of overall survival (OS), local recurrence (LR) and distant metastasis (DM) were evaluated. From 135 IFs, only 17 satisfied the required criteria of being normally distributed across 4D PET and robust between 3D and 4D images. Changes during treatment in the area-under-the-curve of the cumulative standard-u...
Journal of Clinical Oncology, 2005
The objective of this study was to determine the accuracy of (early) response measurements using [18F]-2-fluoro-2-deoxy-D-glucose positron emission tomography (18FDG PET) with respect to survival of patients with stage IIIA-N2 non-small-cell lung cancer (NSCLC) undergoing induction chemotherapy (IC), with a comparative analysis of PET methods. In a prospective multicenter study, PET was performed in patients before IC and after one and three cycles. Computed tomography (CT) was performed before and after IC. Glucose consumption (metabolic rate of glucose [MRglu]) was measured using Patlak graphical analysis and correlated with simplified methods. Mediastinal lymph node (MLN) status was assessed visually. Cox proportional hazards analysis was used to determine the prognostic relevance of CT and PET measures of response with respect to survival. Complete PET data sets were available in 47 patients. Median survival was 21 months. MLN status after IC by PET predicted survival (hazard ratio [HR], 2.33; 95% CI, 1.04 to 5.22; P = .04) in contrast with CT (HR, 1.87; 95% CI, 0.81 to 4.30; P = .14). Residual MRglu after IC proved to be the best prognostic factor (HR, 1.95; 95% CI, 1.28 to 2.97; P = .002). Multivariate stepwise analysis showed that PET identified prognostically different strata in patients considered responsive according to CT. Residual MRglu after one cycle selected patients with different outcomes (HR, 2.04; 95% CI, 1.18 to 3.52; P = .01). Simplified quantitative 18FDG PET methods were correlated with Patlak graphical analysis during and after therapy (r &amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;gt; or = 0.90). 18FDG PET has additional value over CT in monitoring response to IC in patients with stage IIIA-N2 NSCLC, and it seems feasible to predict survival early during IC. Simple semiquantitative and complex PET methods perform equally well.
Cancers
Purpose: We aimed to assess the ability of radiomics features extracted from baseline (PET/CT0) and follow-up PET/CT scans, as well as their evolution (delta-radiomics), to predict clinical outcome (durable clinical benefit (DCB), progression, response to therapy, OS and PFS) in non-small cell lung cancer (NSCLC) patients treated with immunotherapy. Methods: 83 NSCLC patients treated with immunotherapy who underwent a baseline PET/CT were retrospectively included. Response was assessed at 6–8 weeks (PET/CT1) using PERCIST criteria and at 3 months with iPERCIST (PET/CT2) or RECIST 1.1 criteria using CT. The predictive performance of clinical parameters (CP), standard PET metrics (SUV, Metabolic Tumor volume, Total Lesion Glycolysis), delta-radiomics and PET and CT radiomics features extracted at baseline and during follow-up were studied. Seven multivariate models with different combinations of CP and radiomics were trained on a subset of patients (75%) using least absolute shrinkage...
Medicine, 2014
The aim of this study is to determine the prognostic role and the timing of metabolic response to chemotherapy, based on 18 F-fluorodeoxyglucose positron emission tomography (18 F-FDG-PET), in patients with metastatic non-small-cell lung cancer (NSCLC). The study included 55 patients with metastatic NSCLC that were analyzed in terms of prognostic factors and survival. 18 F-FDG-PET/CT findings were evaluated in patients separated into 3 groups, before and after 1st, 2nd, 3rd cycle of the first line chemotherapy. Metabolic response was assessed according to PET Response Criteria in Solid Tumors (PERCIST 1.0). Among the 55 patients, 34 (62%) died, and 21 (38%) remained alive during a mean follow-up of 13.5 months. Median overall survival (OS) was 11.69 months (range 2-26.80 months) and median progressionfree survival (PFS) was 6.27 months (range 1.37-20.43 months). Univariate analysis showed that the only favorable prognostic factor for OS in all the patients was the achievement of metabolic response. Metabolic response according to PERCIST, and weight lose 5% were also independent favorable prognostic factors predictive of survival in all patients based on multivariet analysis (metabolic response: P ¼ 0.002, OR; 1.90, 95% CI 1.26-2.89, and weight lose 5%: P ¼ 0.022, OR; 2.24, 95% CI 1.12-4.47). Median OS in all patients with partial response (PR)-according to the PERCIST 1.0-was significantly longer than in those with progressive disease (PD) (16.36 months vs 8.14 months, P ¼ 0.008). Median OS in the patients with PR was significantly longer than in those with PD based on PET/CT performed after 2nd and 3rd cycles of chemotherapy (18.35 months vs 7.54 months, P ¼ 0.012 and 18.04 months vs 7.43 months, P < 0.001, respectively), whereas, median OS did not differ significantly between patients with PR and those with PD based on PET/CT performed after the 1st cycle of chemotherapy (8.01 months vs 5.08 months, P ¼ 0.290). Metabolic response according to PERCIST and weight loss are independent factors predictive of OS. PET/CT performed after second cycle of chemotherapy may be the earliest predictor of treatment response in patients with advanced stage NSCLC. (Medicine 93(28):e299) Abbreviations: CR = complete metabolic response, ECOG PS = Eastern Cooperative Group Scale performance status, 18 F-FDG-PET = 18 F-fluorodeoxyglucose positron emission tomography, NSCLC = non-small-cell lung cancer, OS = overall survival, PR = partial metabolic response, PERCIST = PET response criteria in solid tumors, PFS = progression-free survival, PD = progressive metabolic disease, RECIST = response evaluation criteria in solid tumors, SD = stable metabolic disease.
European journal of nuclear medicine and molecular imaging, 2018
The aim of this multi-center study was to discover and validate radiomics classifiers as image-derived biomarkers for risk stratification of non-small-cell lung cancer (NSCLC). Pre-therapy PET scans from a total of 358 Stage I-III NSCLC patients scheduled for radiotherapy/chemo-radiotherapy acquired between October 2008 and December 2013 were included in this seven-institution study. A semi-automatic threshold method was used to segment the primary tumors. Radiomics predictive classifiers were derived from a training set of 133 scans using TexLAB v2. Least absolute shrinkage and selection operator (LASSO) regression analysis was used for data dimension reduction and radiomics feature vector (FV) discovery. Multivariable analysis was performed to establish the relationship between FV, stage and overall survival (OS). Performance of the optimal FV was tested in an independent validation set of 204 patients, and a further independent set of 21 (TESTI) patients. Of 358 patients, 249 die...
Lung Cancer, 2012
It is obvious that FDG-PET has added value to CT, but there is still insufficient data to define the role of FDG-PET/CT in clinical practice. Usually data are gathered from multiple sources and in consequence the information obtained is heterogeneous and not always comparable between patients. To alleviate this lack of data, we attempted to investigate the differences in staging and therapeutic intent as compared with conventional staging in non small cell lung cancer (NSCLC) patients scheduled for RT after adding FDG-PET/CT to conventional staging in 104 included subjects. In contrary to the multicentric studies relying on patients medical records from outside institutions, these data were generated entirely with the institution's PET/CT unit. Significant modifications of both, M-stage and clinical stage were detected after inclusion of FDG-PET/CT data (p < 0.001), while there was no statistically significant T-and N-stage modification. Overall implenting FDG-PET/CT revised RT intention decision in 34% of patients. FDG-PET/CT provides enhanced staging capabilities compared to conventional CT in staging of non small cell lung carcinoma and allows improved selection of patients suitable for curative intention, while avoiding unnecessary irradiation and costs in patients eligible to palliative intention.
International Journal of Radiation Oncology*Biology*Physics, 2019
Purpose/Objective(s): A secondary analysis of the ACRIN6668/ RTOG0235 study identified SumMean, a textural feature calculated from the grey level co-occurrence matrix on 18 F-FDG PET, as a potential prognostic factor for patients with locally advanced non-small cell lung cancer (LA-NSCLC) who have large primary tumors and receive definitive, concurrent chemoradiotherapy (CRT). Here we attempted to validate that finding using an independent patient cohort. Materials/Methods: We identified LA-NSCLC patients at our institution who underwent staging 18 F-FDG PET and then received definitive CRT between 2007 and 2018. Primary tumors were segmented using a semiautomatic thresholding algorithm. Based on the ACRIN6668/RTOG0235 analysis, SumMean was calculated for tumors larger than 93.3 cm 3 , and a cutoff of 0.018 was utilized to categorize SumMean as high or low. Rates of progression-free survival (PFS) and overall survival (OS) after CRT initiation for patients with high and low SumMean values were estimated using the Kaplan-Meier method. SumMean was tested as a predictor of PFS and OS using Cox proportional hazards models. Results: One hundred fifty-five LA-NSCLC patients underwent staging PET followed by CRT in the study period. Thirty-four patients had tumors larger than 93.3 cm 3. Twelve (35%) tumors had high SumMean, and 22 (65%) had low SumMean. SumMean category was not statistically significantly associated with age, performance status, clinical stage, metabolic tumor volume (MTV), or histologic subtype. Median follow-up duration for living patients is 26.5 months. The 2-year PFS for patients with large tumors and high SumMean is 50%, compared to 14% for patients with large tumors and low SumMean (HR for high SumMeanZ0.28, 95% CI 0.11 to 0.77, pZ0.014). The 2-year OS for patients with large tumors and high SumMean is 90%, compared to 38% for patients with large tumors and low SumMean (HR for high SumMeanZ0.26, 95% CI 0.08 to 0.91, pZ0.034). Conclusion: Using an independent patient cohort, we have validated the finding that SumMean, assessed on staging 18 F-FDG PET for LA-NSCLC patients with large tumors, predicts clinical outcomes following treatment with definitive CRT. Notably, the prognostic value of SumMean appears to be robust across multiple PET scanners and the use of different reconstruction algorithms. Future studies to elucidate the biological correlates of the Sum-Mean feature are required, providing a data driven pathway toward treatment individualization and a rationale for exploring alternative treatment options for patients with large LA-NSCLC tumors and low SumMean score.