Development and validation of multiparametric models incorporating 18F-FDG PET/CT dissemination characteristic for predicting outcomes of small cell lung cancer.

Aim/Introduction[18F]FDOPA PET/CT is one of the most frequently used functional imaging modalities for the diagnosis of pheochromocytoma and paraganglioma (PPGL). The biochemical secretion type of PPGL is crucial for patient management, being linked to varying aggressiveness and metastatic risk. The aim of this study is to compare the biochemical phenotype and secretion with uptake intensity and tumour volume on [18F]FDOPA PET/CT.MethodsAll patients with sympathetic PPGL undergoing [18F]FDOPA PET/CT and laboratory analysis at first diagnosis at LMU University Hospital between 03/2012 and 11/2023 were included. Metabolic tumour volume (MTV), SUVmax, SUVmean and total lesion uptake (TLU) were compared to biochemical secretion using Rho-Spearman’s correlation and linear regression correcting for age, gender and secretion type. Biochemical phenotypes were compared with Mann-Whitney-U-test, ROC analysis was used to test the diagnostic discriminative power of radioligand uptake.Results71 of 74 PPGL were [18F]DOPA-positive. TLU and MTV showed a moderate to strong correlation with plasma and urinary normetanephrines (R = 0.68–0.82, p < 0.001), plasma 3-methoxytyramine (R = 0.50, p = 0.003) and urinary metanephrines (R = 0.69–0.80, p < 0.001). Regression analysis revealed a significant relationship between biochemical secretion and TLU (r2 = 0.45–0.52, p < 0.001). Compared to adrenergic PPGL, noradrenergic PPGL demonstrated an increased radioligand uptake (p < 0.001). ROC analysis identified thresholds for SUVmax ( > 12.1) and SUVmean ( > 6.95) that moderately distinguished both phenotypes (AUC = 0.75-0.76).ConclusionRadioligand uptake on [18F ]FDOPA PET/CT is associated with the biochemical phenotype of PPGL. This finding may facilitate the metabolic profiling of patients with suspected impaired or delayed laboratory results. Normetanephrine concentrations in plasma and 24-hour urine may be employed as predictive markers of MTV and TLU.

Abstract Glioblastoma response assessment relies on MRI-based RANO 2.0 criteria, where the presence of measurable disease often is required for clinical trial eligibility. Key enrollment timepoints include the post-radiotherapy period and first recurrence. The recently introduced PET RANO 1.0 criteria incorporate [¹⁸F]FET-PET and metabolic tumor volume as novel assessment parameters. This study compares PET RANO 1.0 criteria and MRI-based RANO 2.0 criteria in detecting measurable disease at these two critical timepoints. In this retrospective, single-center study, we included patients with IDH-wildtype glioblastoma who underwent both [¹⁸F]FET-PET and MRI, either after standard first-line radiotherapy (within 35 days post-radiation, per RANO 2.0) or at first recurrence (defined by histological confirmation or a treatment change due to radiologically confirmed progression). Two independent raters evaluated measurable disease using both PET RANO 1.0 and RANO 2.0 criteria. Lesion size, tracer uptake metrics (TBRmax, TBRmean), and their correlation with Karnofsky Performance Status (KPS) at the timepoint of PET imaging were also analyzed. We assessed 139 patients at the timepoint after post-radiotherapy (median age 59 years, IQR 53-67) and 170 patients at first recurrence (median age 59, IQR 54-68). MRI-based RANO 2.0 identified measurable disease in 69/139 patients (49.6%) at baseline and 111/170 patients (65.3%) at recurrence, with a median sum of cross-sectional diameters of 23.5mm and 24.4mm, respectively. In contrast, PET RANO 1.0 identified significantly more cases with measurable disease: 125/139 (89.9%) at baseline and 161/170 (94.7%) at recurrence, with median metabolic tumor volumes of 8.6cm³ and 14.8cm³ (p&amp;lt;0.001), respectively. Notably, higher post-radiation TBRmax was associated with poorer KPS (p=0.0248). PET RANO criteria detect a higher proportion of patients with measurable disease compared to MRI-based RANO 2.0, potentially broadening clinical trial eligibility and improving tumor burden assessment. These results support further prospective validation of PET-based criteria for trial enrollment and patient stratification in glioblastoma.

ObjectiveTo investigate the value of Fluorine-18 fluorodeoxyglucose-positron emission tomography/computed tomography ([18F]-FDG PET/CT) in the diagnostic of patients having gastrointestinal (GI) cancers with second primary malignancies (SPMs).MethodsFifty-seven patients (57/1384, 4.1%) diagnosed with SPMs were retrospectively enrolled. The analysis included the following factors: clinical information (sex, age, smoking and drinking history, BMI), site of the second primary tumor, the interval between the diagnoses of the first GI cancer and the SPMs, and histopathology. According to the incidence of SPMs, the patients were divided into lung cancer and non-lung cancer groups. The two groups were compared smoking and drinking history, interval time, distant metastasis, and [18F]-FDG PET/CT-related parameters (maximum standardized uptake value [SUVmax], metabolic tumor volume [MTV], and total lesion glycolysis [TLG]). Twenty-two patients with synchronous cancers were included to compare the Ki-67 index and [18F]-FDG PET/CT-related parameters (SUVmax, MTV, and TLG) between the first and second primary tumors.ResultsThe most common SPMs of the GI system was lung cancer (47.4%, 27/57). Genetic testing revealed abnormalities in three patients, and the pathological type was adenocarcinoma in all three cases. Among the 57 patients diagnosed with multiple cancers (27 synchronous, 30 metachronous), the lung was the most frequent site for both synchronous and metachronous tumors. Between the lung cancer group and the non-lung cancer group differences in the interval between the first GI, age, distant metastasis rate, SUVmax, MTV, and TLG were not statistically significant (P = 0.09, P = 0.288 and P = 0.57). Chest CT and PET/CT were performed preoperatively in all 27 patients in the lung cancer group, and the diagnostic accuracy of PET/CT for the second primary tumor in the group was 100% and chest CT diagnostic accuracy was 77.8%. We found significant differences in the Ki-67 indices between the synchronous cancers (64.5 ± 24.0 vs. 35.1 ± 22.7, p < 0.000). Furthermore, the Ki-67 index was highly expressed in patients with lymph node and distant metastases, but the SUVmax, MTV, and TLG were not significantly different between the groups with lymph node and distant metastases (P = 0.366, P = 0.565 and P = 0.869).ConclusionThe most common SPMs of the GI system was lung cancer. We found that [18F]-FDG PET/CT in patients with GI cancers can help identify primary and metastatic lesions and detect the presence of SPMs at an early stage. Patients with SPMs may have unique characteristics, can be beneficial in helping high-risk patients with early intervention.

Background: The aim was to evaluate whether an artificial intelligence (AI)-based tool for the automated quantification of the total metabolic tumour volume (tMTV) in patients with Hodgkin lymphoma (HL) could support nuclear medicine specialists in lesion segmentation and thereby enhance inter-observer agreement. Methods: Forty-eight consecutive patients who underwent staging with [18F]FDG PET/CT were included. Eight invited specialists from different hospitals were asked to manually segment lesions for tMTV calculations in 12 cases without AI advice, and to use automated AI segmentation in a further 12 cases, with editing as required, i.e., segmenting/adjusting 24 cases each. Each case was segmented by two specialists manually and by two different specialists using the AI tool, allowing for the pairwise comparison of inter-observer variability. Results: The median difference between two specialists performing manual tMTV segmentations was 26 cm3 (IQR 10–86 cm3) corresponding to 23% (IQR 7–50%) of the median tMTV in the dataset, while the median difference between two specialists tMTV adjustments using AI segmentations was 12 cm3 (IQR 4–39 cm3) corresponding to 9% (IQR 2–21%) (p = 0.023). The median difference in tMTV between measurements with and without AI was 3.3 cm3, corresponding to 2.3% of the median tMTV. Conclusions: An automated AI-based tool can significantly increase agreement among specialists quantifying tMTV in HL patients staged with [18F]FDG PET/CT, without markedly changing the measurements.

This study aimed to assess the prognostic significance of novel and established intratumoral heterogeneity indices (HIs) derived from 18F-fluorodeoxyglucose positron emission tomography/computed tomography ([18F]FDG PET/CT) and multiparametric composite risk scores (CRS) combining these indices with PET/CT-derived metrics, clinical parameters, and metastatic variables in breast cancer (BC) patients. We retrospectively evaluated 135 BC patients who underwent [18F]FDG PET/CT for pretreatment staging. Metabolic and volumetric data of primary tumors obtained from [18F]FDG PET/CT images, such as the maximum, mean, peak, and minimum standardized uptake values (SUVmax, SUVmean, SUVpeak, and SUVmin), the SUV corrected for lean body mass (LBM) calculated by James's and Janmahasatian's methods (SULmax, SULmean, SULpeak, and SULmin), metabolic tumor volume (MTV), and total lesion glycolysis (TLG), and two predefined and seven novel HIs were compared between molecular subtypes via the Kruskal-Wallis (KW) test. All relevant HI, PET/CT-derived metrics, and clinical, pathological, and metastatic variables were included in the cross-validated LASSO regression models to estimate the overall survival (OS) endpoints for 1, 2, 3, 4, and 5 years. Significant differences were observed between molecular subtypes for SUVmax, SUVpeak, TLG_40, and HI5 (Janma/James) (p < 0.05), with the highest values in the HER2-enriched and triple-negative (TNBC) subtypes. CRS, which combines clinical factors, metastatic status, PET/CT-derived metrics, and HI, demonstrated robust discrimination of OS (area under the curve [AUC]: 0.79-0.91) and outperformed single-parameter models. Among the heterogeneity indices, HI2 and HI4 showed the strongest independent predictions of OS at multiple time points, although a combination of multiple parameters was required for optimal prognostic accuracy. CRS, which integrates imaging-derived heterogeneity and metabolic and clinical data, offers improved OS prediction and individualized risk stratification in BC patients.

Baseline PET-CT in predicting CRS and survival outcomes of R/R MM patients following CAR T-cell therapy

Prognostic role of FDG-PET in outcomes of relapsed/refractory large B-cell lymphoma treated with CD3-CD20 directed bispecific antibodies

Total metabolic tumor volume as a prognostic biomarker in high-burden follicular lymphoma: A systematic review and meta-analysis

Assessment of the prognostic value of FDG PET-derived markers and responses in POLARIX

Prognostic value of combining beta-2-microglobulin and total metabolic tumor volume in patients with follicular lymphoma: A post-hoc analysis of the relevance trial

An immunometabolic companion biomarker to enhance FDG-PET interpretation and guide frontline therapy in follicular lymphoma

Clinical contribution of 18F-FDG PET/CT in patients with pediatric bone tissue and soft tissue sarcoma; a retrospective study.

To develop and validate an AI method for automated quantification of whole-skeleton bone marrow (BM) metabolic activity using Carbon 11 (11C)-methionine (MET) PET/CT and to evaluate its prognostic value compared with Fluorine 18 (18F)-fluorodeoxyglucose (FDG) PET/CT in patients with newly diagnosed MM. This prospective study included 49 patients (median age, 68 years; 29 males) with newly diagnosed MM. All patients underwent both 11C-MET and 18F-FDG PET/CT. An AI algorithm initially segments the skeleton on CT images, then propagates the resulting mask to the standardized uptake value (SUV) PET images for automated PET/CT segmentation and quantitative volumetric assessment of BM metabolism. By applying a series of SUV thresholds, the algorithm calculates 11C-MET metabolic tumor volume (MTV) and total lesion methionine uptake (TLMU). Associations with clinical markers (bone marrow plasma cell [BMPC] percentage, serum β₂-microglobulin, International Staging System [ISS]/Revised ISS [R-ISS] stage) and progression-free survival (PFS) were assessed. AI-quantified 11C-MET MTV and TLMU showed significant correlations with BMPC percentage (MTV: r = 0.32, p = 0.02; TLMU: r = 0.31, p = 0.03), serum β₂-microglobulin (MTV: r = 0.29, p = 0.05; TLMU: r = 0.29, p = 0.05), ISS stage (MTV: r = 0.31, p = 0.03; TLMU: r = 0.32, p = 0.03), and R-ISS stage (MTV: r = 0.40, p = 0.02; TLMU: r = 0.37, p = 0.03). In multivariable Cox analysis, both ¹¹C-MET MTV (HR = 1.0023; [95% CI: 1.0004-1.0042]; p = 0.02) and TLMU (HR = 1.0003; [95% CI: 1.0001-1.0005]; p = 0.01) independently predicted PFS. For PFS prediction, 11C-MET MTV (Area Under the Receiver Operating Characteristic Curve [AUC] = 0.743; [95% CI: 0.563-0.903]; p < 0.01) and TLMU (AUC = 0.749; [95% CI: 0.576-0.904]; p < 0.01) outperformed ¹⁸F-FDG PET/CT total lesion glycolysis (TLG) (AUC = 0.713, p < 0.01) and MTV (AUC = 0.719, p < 0.01) using the proposed thresholds. AI-quantified 11C-MET MTV and TLMU act as objective biomarkers of disease burden, they independently predict MM prognosis more effectively than 18F-FDG parameters and may enhance risk stratification.

ObjectivesTo evaluate the potential value of 18F-FDG positron emission tomography (PET) and multiparametric MRI (intravoxel incoherent motion, IVIM, and diffusion kurtosis imaging, DKI) in the prediction of lymphovascular invasion (LVI) in non-small cell lung cancer (NSCLC).Materials and methodsA total of 73 patients with NSCLC who underwent integrated 18F-FDG PET/MRI were included. IVIM, DKI, and PET parameters with or without LVI of NSCLC were measured and compared, and the area under the receiver operating characteristic curve (AUC) was used to evaluate the diagnostic efficacy of each parameter. Univariate and multivariate logistic regression models were used to study the optimal combination of PET/MRI parameters for predicting LVI.ResultsPET-derived parameters (SUVmax, MTV, TLG) and IVIM, DKI MRI-derived parameters (ADCstand, D, MK, MD) were significantly different between patients with and without LVI (p < 0.05). Multivariate logistic regression analysis showed that MTV and D were independent predictors of LVI, and the combined prediction model of the two parameters had the highest predictive value for the diagnosis of LVI (AUC = 0.841; sensitivity = 63.83%; specificity = 92.31%).ConclusionThe present study demonstrates that IVIM, DKI, and PET can be utilized to evaluate LVI status in NSCLC, with the combined diagnostic approach of MTV and D showing the highest diagnostic performance, which may provide a novel reference for clinical management.Critical relevance statementThe performance of metabolic parameters and diffusion parameters in the identification of lymphovascular invasion (LVI) in non-small cell lung cancer (NSCLC) is similar, but the combination of metabolic tumor volume (MTV) and true diffusion coefficient (D) may improve the diagnostic efficacy.Key PointsA multimodal PET-MRI model evaluates lymphovascular invasion (LVI) in patients with non-small cell lung cancer (NSCLC).Metabolic and diffusion parameters have similar efficacy in predicting LVI in NSCLC.The combined metabolic tumor volume and true diffusion coefficient prediction model is the most valuable.Graphical

Background/AimPositron emission tomography/computed tomography with 18F-fluorodeoxyglucose (18F-FDG PET/CT) is frequently used to differentiate schwannomas from malignant peripheral nerve sheath tumors. Schwannomas exhibit pathological heterogeneity, with highly cellular (Antoni A) and hypocellular (Antoni B) areas, but current PET/CT methods do not adequately reflect this heterogeneity. This study aimed to compare imaging characteristics of schwannomas in the trunk versus the extremities, with emphasis on metabolic heterogeneity.Patients and MethodsThis retrospective study included patients with solitary schwannomas who underwent MRI and 18F-FDG PET/CT before surgical excision (June 2013-September 2023). Exclusion criteria were plexiform, multiple, biopsy-only lesions, and tumors originating from internal organs. Tumors were classified as trunk or extremity lesions. MRI was used to determine size and volume, while PET/CT measured SUVmax, SUVmean, metabolic tumor volume (MTV), and total lesion glycolysis (TLG). Heterogeneity was assessed using three indices: MTV-to-volume ratio (MTV/volume), SUV-based heterogeneity index (HISUV), and metabolic region-adjusted SUV-based heterogeneity index (MRA-HISUV).ResultsFifty-six patients were included. Trunk schwannomas were larger than extremity tumors in diameter (4.33 cm vs. 2.77 cm; p<0.05) and volume (27.71 cm3vs. 6.25 cm3; p<0.05). SUVmax (4.09 vs. 3.71) and SUVmean (2.47 vs. 2.22) did not differ significantly. MTV (18.43 cm3vs. 6.19 cm3, p<0.05) and TLG (58.41 vs. 14.40, p<0.05) were higher in trunk tumors. MTV/volume ratio was lower (0.77 vs. 1.12, p<0.05), while HISUV and MRA-HISUV were higher in trunk schwannomas (1.79 vs. 1.65 and 2.36 vs. 1.49, p<0.05).ConclusionTrunk schwannomas were larger and exhibited higher metabolic activity and heterogeneity. Novel parameters such as MTV/volume and MRA-HISUV may enhance the characterization of schwannoma heterogeneity.

Ultra-short immunotherapy may spare cutaneous squamous cell carcinoma (CSCC) patients from mutilating surgery, but early identification of (non-)response is needed to safely guide treatment adaptation. This study evaluated the feasibility of sequential [18F]FDG-PET/CT (FDG-PET) as a response biomarker in resectable CSCC patients. In the MATISSE, a randomized phase-II trial, 50 CSCC patients received two courses of neoadjuvant nivolumab (weeks 0 and 2) with or without low-dose ipilimumab (week 0) before surgery (week 4). FDG-PET scans were obtained pre-treatment and shortly prior to surgery to assess the change (Δ) in maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) at the primary tumour and largest (= index) lymph node metastasis (ILN). ΔMTV50%/4.0 and ΔTLG50%/4.0 were calculated using thresholds of 50% SUVmax and SUV ≥ 4.0. In 42 evaluable patients, 31 (74%) patients showed major or partial responses to immunotherapy. EORTC-criteria underestimated response but accurately identified non-responders (70% sensitivity, 100% specificity). In 28 primary tumours and 22 ILNs, a significant reduction in median SUVmax, MTV50%, MTV4.0, TLG50%, and TLG4.0 was observed in responders versus non-responders (overall, p ≤ 0.004, and p ≤ 0.03, respectively). ΔTLG50% and ΔTLG4.0 correlated strongly with response (primary: 92% and 96% accuracy; ILN: 91% and 89% accuracy). Quantitative FDG-PET-response assessment allows early identification of (non-)responders upon neoadjuvant immunotherapy prior to surgery in locoregionally advanced CSCC patients. Early changes in FDG-PET's TLG can support future trials aiming at safe de-escalation of current standard of care surgery with or without adjuvant radiotherapy. EudraCT 2020-001074-30. Registered 9 March 2020, https://www.clinicaltrialsregister.eu/ctr-search/search?query=2020-001074-30.

68Ga-DOTATATE binds to somatostatin receptors (SSTR) and is used for PET/CT imaging for diagnosing meningioma and guiding postoperative radiotherapy. However, the relationship between the clinicopathological characteristics and the imaging features of 68Ga-DOTATATE PET/CT in meningioma remains undetermined. We conducted a retrospective study at a single neurosurgical center. Semiquantitative indices of 68Ga-DOTATATE PET/CT, including maximum standardized uptake value (SUV-max), median standardized uptake value (SUV-median), mean standardized uptake value (SUV-mean), and metabolic tumor volume (MTV), were measured. The correlations between these parameters and clinicopathological characteristics were analyzed. Eighty patients were retrospectively analyzed, including 45 with WHO grade 1 meningiomas, 33 with grade 2, and 2 with grade 3. The median SUV-max value was significantly higher in meningiomas with high WHO grade (P = 0.017), positive SSTR2a expression (P = 0.023), and in male patients (P = 0.002). Bulk RNA data was available in forty patients. RNA expression of SSTR2 was significantly elevated in the SSTR2a-positive group and correlated positively with SUV-max, SUV-median, and SUV-mean. Furthermore, the high RNA scores risk group exhibited significantly higher SUV-max compared to the intermediate- and low-risk groups. Of note, an SUV-max > 15.55 (P = 0.014) was independently associated with high WHO grade. Lastly, a nomogram incorporating SUV-max, gender, and surgical history demonstrated robust performance in preoperatively identifying patients at risk for high WHO grade, offering the potential utility of 68Ga-DOTATATE PET/CT in clinical practice. SUV-max of 68Ga-DOTATATE PET/CT was significantly higher in high grade and high RNA score risk groups of meningioma. Of note, an SUV-max > 15.55 was independently associated with high WHO grade.

BackgroundMalignant brain tumors emphasize the importance of O-(2-18F-fluoroethyl)-L-tyrosine (18F-FET) positron emission tomography (PET) imaging for accurate diagnosis and treatment planning, necessitating standardized quantitative features for reliable assessment. However, the calculation of these features is influenced by acquisition duration, as reducing acquisition time remains a key concern in clinical practice. Furthermore, reconstruction algorithms significantly affect imaging quality. This study aimed to clarify the impact of acquisition duration and reconstruction algorithms on the repeatability of 18F-FET PET quantitative features in brain tumors.MethodsA total of 62 patients performing brain 18F-FET PET/magnetic resonance (MR) examinations were retrospectively enrolled. The PET images were reconstructed using 24 designed schemes, comprising a combination of eight acquisition time windows (3, 5, 7, 10, 13, 15, 17, and 20 min) with three reconstruction algorithms [ordered subset expectation maximization (OSEM), OSEM with time-of-flight (OWT), and HYPER iterative with time-of-flight (HIWT)]. Image quality was evaluated using a 5-point Likert scale. The repeatability of quantitative metabolic and radiomic features between the three algorithms was assessed using intraclass correlation coefficients (ICC), whereas temporal stability between 15, 17, and 20 minutes for each algorithm was validated using the Friedman test.ResultsPET reconstruction images at 15, 17, and 20 minutes were considered to provide diagnostic value. The mean standardized uptake value (SUV) and tumor-to-brain ratio (TBR) showed minimal variation with acquisition duration for all three algorithms, with the relative percentage difference (RPD) <1.2% after 15 minutes. The maximum SUV (SUVmax), maximum TBR (TBRmax), metabolic tumor volume (MTV), and total lesion uptake (TLU) became usable when acquisition time exceeded 15 minutes, with an RPD of around 5% or less. There were 8 common metabolic features and 30 radiomics features which demonstrated excellent repeatability between the three algorithms at 15, 17, and 20 minutes. The HIWT algorithm identified 18 stable radiomics features, whereas the OWT identified 2, and the OSEM identified 3.ConclusionsThis study offers a reference for clinically reducing the acquisition time of 18F-FET PET imaging in brain tumors. It compares the temporal stability of different reconstruction algorithms and identifies metabolic and radiomic features with high repeatability and stability for each. These findings help to optimize imaging protocols and improve the reproducibility of quantitative analysis in 18F-FET PET studies for brain tumors.

Development and validation of a radiomics model based on pretreatment deoxy-2-[fluorine-18]-fluoro-D-glucose positron emission tomography/computed tomography (18F-FDG PET/CT) imaging for predicting lymph node metastasis (LNM) in esophageal squamous cell carcinoma (ESCC). A retrospective analysis was performed on 145 patients with ESCC, using pretreatment 18F-FDG PET/CT imaging data and clinical information. Patients were randomly divided into training and validation cohorts in a 7:3 ratio. In the training cohort, independent risk factors for LNM in ESCC were identified through univariate and multivariate logistic regression analyses. Radiomic features were extracted from the PET images, and the least absolute shrinkage and selection operator (LASSO) regression was used for dimensionality reduction. Features highly correlated with LNM in ESCC were selected. The weighted radiomics score (Radscore) was then calculated based on these selected features. The diagnostic performance of each factor was evaluated using receiver operating characteristic (ROC) curves, and a prediction model nomogram was established. Decision curve analysis (DCA) was conducted to evaluate the clinical utility of the model. Finally, the model was validated using the validation cohort. Maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), total lesion glycolysis (TLG), and gender were significantly associated with LNM in ESCC (all P < 0.05). SUVmax was found to be an independent risk factor for predicting LNM in ESCC. In the training and validation cohorts, the areas under the curve (AUC) for SUVmax combined with Radscore were 0.809 (95% CI: 0.723-0.894) and 0.801 (95% CI: 0.661-0.941), respectively, both of which were higher than those for SUVmax and Radscore alone. A nomogram, a comprehensive predictive model based on SUVmax and Radscore, may improve the net clinical benefit for patients. The nomogram, a predictive model developed using 18F-FDG PET/CT-based radiomics, offers reliable predictive value for LNM in ESCC and is expected to serve as a reference tool for therapeutic decision making in patients with ESCC.