Purpose

To investigate the diagnostic accuracy of CT imaging features for the identification of high-risk GI stromal tumors (GISTs).

Material and methods

This retrospective dual-institution study included patients with pathologically proven GIST meeting the following inclusion criteria: i) preoperative contrast-enhanced CT performed between 2007 and 2019; ii) lack of neoadjuvant treatment; iii) pathological analysis through resection specimens. Tumor risk stratifications were determined according to the National Institutes of Health (NIH) 2008 criteria. Two radiologists evaluated the CT imaging features, including enhancement pattern and tumor characteristics in a blinded fashion. The distribution of CT features between high-risk and low-to-intermediate-risk GISTs was assessed using univariate and multivariate binary logistic regression analyses. Statistical significance was set at p<0.05.

Results

The final population included 86 patients (58 men, 28 women, mean age 60.1±10.9 years) with 37 high-risk and 49 low-to-intermediate risk GISTs (mean diameter 8.6±6.2 cm). High-risk GISTs demonstrated more frequently heterogeneous enhancement (91.9% vs 44.9%, p<0.001), lobulated contours (86.5% vs 32.7%, p<0.001), ill-defined margins (37.8% vs 0%, p<0.001), exophytic growth (73.0% vs 49.0%, p=0.018), intralesional necrosis (67.6% vs 32.7%, p=0.001), cystic degeneration (10.8% vs 0%, p=0.031), intratumoral vessels (37.8% vs 6.1%, p<0.001), and enlarged feeding vessels (67.5% vs 4.1%, p<0.001) compared to low-to-intermediate-risk GISTs. At multivariate analysis, lobulated contours (OR: 0.12, 95% C.I.: 0.03-0.78, p=0.023) and enlarged feeding vessels (OR: 16.07, 95% C.I.: 1.89-136.19, p=0.010) remained independently associated with high-risk GIST.

Conclusion

Morphologic contrast-enhanced CT features are significantly different depending on the risk status and may help to predict patients with high-risk GIST.

Purpose

To evaluate the association between radiomic biomarkers extracted from baseline CT imaging, mitotic count, tumor mutational profile and prognostic Miettinen classification.

Material and methods

This retrospective multicenter observational study includes 63 histologically proven gastrointestinal stromal tumors (GISTs). Each lesion was manually segmented; 37 texture features were extracted either on a single slice and on the entire tumor volume. Reference standards: pathological findings and Miettinen classification. Patients were dichotomized with mitotic count (≤5/50HPF vs >5/50HPF), mutational status (c-KIT mutation vs PDGFRα and wild-type), patients prognosis (good prognosis class: none, very low and low risk vs poor prognosis class: intermediate and high risk). Univariate analysis using the Mann-Whitney test and multivariate analysis were performed; a stepwise logistic regression model was developed to predict patient's prognosis using 70% of patients as the training set and the remaining 30% as the test set.

Results

Eight 3D features discriminated lesions with low or high mitotic count (best AUC 0.81, best sensitivity 86%, best specificity 93%). Six 3D parameters detected GISTs based on the mutational group (best AUC 0.77, best sensitivity 75%, best specificity 79%) and three parameters correlated with risk class (best AUC 0.76, best sensitivity 72%, best specificity 85%). To differentiate between GIST at lower or higher risk of recurrence, the regression model used 6 different features with AUC 0.78, sensitivity 65%, specificity 79%, VPN 71% and VPP 73% on the training set, and AUC 0.83, sensitivity 88% and specificity 75% on the test set.

Conclusion

A good correlation between radiomics features, disease aggressiveness, mutational profile and risk of recurrence was observed. Results are promising; validation on external datasets is necessary to confirm the role as imaging biomarker.

Purpose

To investigate whether there are differential imaging features of intraabdominal desmoid tumor from peritoneal seeding in patients with a history of previous cancer surgery.

Material and methods

From January 2000 to June 2019, 32 patients who had pathologically proven intraperitoneal lesions developed after cancer surgery were enrolled. There were 17 desmoid tumors and 16 peritoneal seedings. Portal-phase CT and/or positron-emission tomography (PET) findings were analyzed by two board-certified radiologists in consensus for the following items: size, shape, margin, degree of enhancement and fluorodeoxyglucose (FDG) uptake, homogeneity, presence of intralesional fat, necrosis, calcification, adjacent organ invasion, peritoneal thickening, and mass effect. For quantitative analysis, the Hounsfield unit (HU) of lesions and psoas muscles as well as maximum standardized uptake value (SUVmax) of the lesions were measured. Imaging findings were compared between desmoid tumor and peritoneal seeding groups using statistical analysis methods.

Results

Desmoid tumors frequently showed iso-attenuation (14/17) while peritoneal seeding depicted high attenuation (12/16) compared to psoas muscle (P=0.001). Intralesional fat was more frequently found in desmoid tumors (8/17) than in peritoneal seeding (1/16) (P=0.017). Desmoid tumors frequently showed well-defined margin (9/17) and smooth contour (13/17) whereas peritoneal seeding had ill-defined margin (13/16) and lobuating contour (11/16) (P=0.041 and 0.009, respectively). HU ratio between the lesion and psoas muscle was not significantly different between desmoid tumor (1.15) and peritoneal seeding (1.23) (P=0.570). SUVmax (4.14) of desmoid tumor did not significantly differ from that (5.19) of peritoneal seeding (P=0.519).

Conclusion

Desmoid tumor can be non-invasively differentiated from peritoneal seeding based on CT findings in patients with a previous history of cancer surgery.

Purpose

To evaluate the role of MRI in local response assessment and detection of early local relapse after chemoradiotherapy (CRT) in patients with anal squamous cell carcinoma (AC).

Material and methods

Twenty-three patients with histologically proven AC who completed pre-(t0), post-CRT 4-week(t1) and 20-week(t2) MRIs and pre-(t0) and post-CRT 20-week(t2) PET scan were included. Tumor size (D, mm), volume (V, cc), apparent diffusion coefficient (ADC, mm²/s), time to peak (TTP, sec) and SUV were recorded. Lesion biopsy performed at 6 months was the gold standard test for the evaluation of response to treatments (responder, r, or non-responders, nr). Data were analyzed using non-parametric test and ROC curve.

Results

Twenty patients were classified as responders. In the responder group, all the parameters significantly changed (decrease of D, V and SUV; increase of ADC and TTP). Non-responders had significantly higher values of D and V than responders at any time: D_t0[nr 80mm vs r 50mm, p=0.03]; D_t1[nr 50mm vs r 26mm, p=0.02]; D_t2[nr 28mm vs r 4mm, p=0.03; Vt₀[180cc vs 35cc, p=0.008]; V_t1[nr 79cc vs r 4cc, p=0.006]; V_t2[nr 13cc vs r 0.13cc, p=0.01]. There were no significant differences of ADC, TTP and SUV in the two groups at any time. ROC curve identified as a threshold for responders: V_t0≤120 cc (AUC=0.99), V_t1≤40 cc (AUC=1.00), ΔV%_t1≥80% (AUC=0.925) and D_t1≤40 mm (AUC=0.92).

Conclusion

V_t0≤120 cc, V_t1≤40 cc, ΔV%_t1≥80% and D_t1≤40 mm seem to be markers of good response to treatment. After therapy, there is a significant increase of ADC and TTP in all patients, without a significant difference between the two groups: in the future, the addition of more non-responder patients could confirm this difference.

Purpose

Early salvage surgery after locoregional failure of chemoradiotherapy (CRT) for anal squamous cell carcinoma (ASCC) is associated with better outcomes. Previous data suggest equivocal findings on 3 months post-treatment MRI in more than half of cases, potentially delaying curative surgery or leading to unnecessary EUA and biopsy. We aimed to evaluate the effectiveness of 3m post-treatment positron emission tomography (PET)/CT in predicting outcomes in ASCC.

Material and methods

257 patients were consecutively managed for ASCC from January 2012 to January 2018. With local ethical approval, reports for 18F-fluorodeoxyglucose (FDG)-PET/CT performed 3 months following CRT were placed into 5 tumour regression categories based on combined qualitative and semiquantitative assessment of anal canal metabolic activity. Outcome measures were overall survival (OS) and locoregional treatment failure/time to progression. Univariable and complete case multivariable landmark analyses, 3-month landmark time, of the covariates of interest were conducted using the Cox proportional hazards model. Presentation of survival curves using the Kaplan-Meier method and p-values from the log-rank test is reported.

Results

243 patients were included. Patients who had residual disease at 3 months as suggested by the PET/CT Tumor regression grading (TRG) score and confirmed on EUA/biopsy had poorer prognosis. Multivariable analysis showed that PET/CT response was a strong predictor of locoregional failure (HR 7.14 (3.89-13.11), p<0.001) and overall survival (HR 7.55 (3.90-14.61), p<0.001).

Conclusion

FDG-PET/CT is highly accurate in assessing early response to treatment in ASCC, an excellent predictor of outcomes and streamline management pathways. This study supports the inclusion of PET/CT in response assessment for anal cancer.

Purpose

To explore the value of CT-based whole-liver radiomics for response evaluation and grade assessment in patients treated with somatostatin analogues (SSAs) for liver-metastasized neuroendocrine tumours (NET).

Material and methods

Thirty-eight consecutive patients with pathologically confirmed small intestine or pancreatic NET (grade 1=24/grade 2=14), treated with SSAs underwent contrast-enhanced CT. The whole liver was semi-automatically delineated on arterial phase with Philips Intellispace Portal, excluding main vessels and bile ducts. Five histogram features were extracted with Pyradiomics, using different Laplacian-of-Gaussian (LoG) filters. Response (progressive disease (PD) vs. stable disease (SD) or partial response (PR)) was assessed with RECIST1.1. Response and grade (1 vs. 2) were compared with Mann–Whitney U or independent t test. Univariate Cox regression was used for overall survival (OS) analyses.

Results

10/38 patients had PD (median follow-up 53 months) and 28 had PR/SD (median follow-up 65 months, p=0.065). Radiomics features did not significantly differ between patients with PD compared to patients with PR/SD. Kurtosis (unfiltered+LoG0.5) was significantly different between patients with grade 1 and 2 NET (p=0.05). Skewness, mean (LoG2.5) and kurtosis (LoG2.5) were significantly associated with OS (HR 0.76 (95%CI: 0.61-0.95); HR 1.96 (95%CI: 1.07-3.60) and HR 1.07 (95%CI: 1.02-1.12), respectively).

Conclusion

These preliminary data show no value for whole-liver CT-based radiomics to identify NET patients treated with SSAs that will show progression. Whole-liver CT-based radiomics might show the potential to identify patients at risk of impaired outcome (i.e. higher grade and OS). These findings need to be validated in a larger cohort and lesion-based analysis will be explored to investigate whether this leads to better results.

Purpose

Y⁹⁰-radioembolization is increasingly used in cancer patients. Even though Y⁹⁰-radioembolization demonstrated to be safe and effective, in some cases it may lead to the so-called radioembolization-induced liver disease (REILD) that is characterized as the occurrence of jaundice and ascites 1-2 months after treatment in patients without tumor progression or bile duct obstruction. The purpose of our study was to identify risk factors of REILD in patients with liver metastases treated with Y⁹⁰-radioembolization.

Material and methods

A retrospective analysis of a prospectively collected database was performed. 116 patients with liver metastases (mean age: 66 years, range 24-91y) were treated with Y⁹⁰-radioembolization (2010-2017). Demographics, radiological, functional and clinical data, number and treatment parameters of SIRT, and administered treatments (chemo-/immunotherapy, liver-directed locoregional therapies) were analyzed. The effect of those parameters was determined by univariate analysis.

Results

78 patients were included in this ongoing analysis. Mean administered dose was 1.9GBq (0.3-5.87). The administered dose correlated with a significant increase in bilirubin and ASAT levels at 1 month post-SIRT (r=0.358, p=0.017 and r=0.39, p=0.006). A trend was observed for bilirubin at 3 months (r=0.288, p=0.068). 9 REILD were observed. Univariate logistic regression demonstrated that baseline y-GT was predictive of REILD (OR=1.302, 95%CI=1.045-1.66; p=0.022). None of the other investigated factors were predictive of the occurrence of REILD.

Conclusion

REILD is an uncommon but potentially severe complication that can happen in patients with liver metastases treated by Y⁹⁰-radioembolization. Patients with elevated baseline y-GT are at increased risk of REILD.

Learning objectives

Learning objectives

Learning objectives

Learning objectives

Learning objectives