Keywords:
Tissue characterisation, Cancer, Segmentation, Decision analysis, Computer Applications-Detection, diagnosis, Image manipulation / Reconstruction, CT-Quantitative, CT, Thorax, Oncology, Computer applications
Authors:
G. Ficarra1, E. Barabino1, C. Genova1, S. Mennella1, M. Verda2, G. Pittaluga1, F. Grossi1, G. Cittadini1; 1Genova/IT, 2Imperia/IT
Results
Δ-Features were calculated as absolute variation between baseline and follow up examination.
Correlation of Δ-Features and clinical endpoints as Progression Free Survival (PFS) and Overall Survival (OS) were assessed by using the Pearson coefficient; a p-value of 0.05 or lower has been considered statistically significant.
Our analysis showed that variations in CTTA parameters,
specially those features derived from co-occurence matrices,
correlated with PFS and OS in our cohort of patients.
Interestingly,
many features correlated with PFS when the progression was established through the irRECIST criteria (mainly in the feature “Sum of Averages”,
calculated as above; 0.005>p>0.0016,
with an increment in this feature positively related with a longer PFS),
whereas no such correlation was found when the progression was assessed by RECIST 1.1 criteria,
thus demonstrating the strong differences between those tumor response criteria.
The strongest predictors of longer OS was increasing Entropy (calculated as above) at first follow up evaluation,
with respectively p=0.0047 and p=0.0039 for the features “Entropy” and “Sum of Entropies”,
suggesting that an increase in tissue heterogeneity at early CT evaluation possibly indicates and predicts a response to nivolumab in NSCLC.