Purpose or Learning Objective To assess whether PET-derived semiquantitative parameters (PDSPs) extracted from 68Ga-DOTANOC PET/CT and single radiomic features (RFs) can differentiate low grade pancreatic neuroendocrine tumour (panNET). Methods or Background 13 patients with G1 and 15 with G2 primary panNET demonstrated by pre-surgical 68Ga-DOTANOC PET/CT were included in this study (M:F=13:15; mean age: 56 years old [17-78]). Tumour grading was assessed after surgical excision evaluation. Total lesion receptorial expression (TLRE), Receptorial tumour volume (RTV), andSUVmax were analysed together with 60 first and second-order RFs computed on standardized uptake value (SUV) maps of the primary lesion whole volume. To prevent overfitting, only single RF were considered to generate discriminative radiomic models, where linearly correlated RFs were removed. Discrimination capability was assessed through the two-tail Wilcoxon rank-sum test with Bonferroni correction (pvalue<0.0031). Receiver operating characteristic and area under the curve (AUC) were computed. Features with the lowest p-values and highest AUC were selected. Results or Findings RTV is the only PDSP yielding a significant separation (p-value=0.03) between G2 and G1 panNET patients (Sensitivity=67%, Specificity=92%, Accuracy=79%). Indeed, SUVmax (Sensitivity=93%, Specificity=38%, Accuracy=68%, p-value=0.71) and TLRE (Sensitivity=87%, Specificity=54%, Accuracy=71%, p-value=0.20) were not significant and led to worse performance, accordingly. On the contrary, the first-order kurtosis provided the best performance (p-value=0.0009) at all, with Sensitivity=93%, Specificity=77%, Accuracy=86%. Conclusion Despite PDSPs are more easily accessible by clinicians, RFs proved to be more accurate in discriminating tumour grading in well-differentiated panNETs. The use of this radiomic model, if validated on a larger population, could represent a novel non-invasive approach to avoid biopsy before surgery, especially in selected patients not amenable to biopsy or with poor health conditions. Limitations Reduced patients’ sample size.

Comparing PET-derived semiquantitative parameters and radiomic features in discriminating G1 and G2 primary pancreatic neuroendocrine tumours in 68Ga-DOTANOC PET/CT

Silvia Malavasi;Diletta Calabrò;Alessandro Bevilacqua;Claudio Ricci;Riccardo Casadei;Davide Campana;Stefano Fanti;Valentina Ambrosini
2021

Abstract

Purpose or Learning Objective To assess whether PET-derived semiquantitative parameters (PDSPs) extracted from 68Ga-DOTANOC PET/CT and single radiomic features (RFs) can differentiate low grade pancreatic neuroendocrine tumour (panNET). Methods or Background 13 patients with G1 and 15 with G2 primary panNET demonstrated by pre-surgical 68Ga-DOTANOC PET/CT were included in this study (M:F=13:15; mean age: 56 years old [17-78]). Tumour grading was assessed after surgical excision evaluation. Total lesion receptorial expression (TLRE), Receptorial tumour volume (RTV), andSUVmax were analysed together with 60 first and second-order RFs computed on standardized uptake value (SUV) maps of the primary lesion whole volume. To prevent overfitting, only single RF were considered to generate discriminative radiomic models, where linearly correlated RFs were removed. Discrimination capability was assessed through the two-tail Wilcoxon rank-sum test with Bonferroni correction (pvalue<0.0031). Receiver operating characteristic and area under the curve (AUC) were computed. Features with the lowest p-values and highest AUC were selected. Results or Findings RTV is the only PDSP yielding a significant separation (p-value=0.03) between G2 and G1 panNET patients (Sensitivity=67%, Specificity=92%, Accuracy=79%). Indeed, SUVmax (Sensitivity=93%, Specificity=38%, Accuracy=68%, p-value=0.71) and TLRE (Sensitivity=87%, Specificity=54%, Accuracy=71%, p-value=0.20) were not significant and led to worse performance, accordingly. On the contrary, the first-order kurtosis provided the best performance (p-value=0.0009) at all, with Sensitivity=93%, Specificity=77%, Accuracy=86%. Conclusion Despite PDSPs are more easily accessible by clinicians, RFs proved to be more accurate in discriminating tumour grading in well-differentiated panNETs. The use of this radiomic model, if validated on a larger population, could represent a novel non-invasive approach to avoid biopsy before surgery, especially in selected patients not amenable to biopsy or with poor health conditions. Limitations Reduced patients’ sample size.
ECR 2021 – BOOK OF ABSTRACTS
1
1
Silvia Malavasi, Diletta Calabrò, Alessandro Bevilacqua, Claudio Ricci, Riccardo Casadei, Davide Campana, Stefano Fanti, Valentina Ambrosini
File in questo prodotto:
Eventuali allegati, non sono esposti

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11585/785447
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact