Red-Green-Blue (RGB) Profiling Of Pancreatic Mass-Elastographies: Validation Of A Predictive Model For Non-Invasive Assessment Of Malignancy

Aims Red-green-blue (RGB) profiling is a well-accepted method in imaging science. We aimed to evaluate the accuracy of quantitative image analysis of pancreatic tumor elastographies obtained by EUS to predict malignancy.

Methods Elastographies of solid pancreatic masses between 04/2017-04/2020 were extracted from our ultrasound device (Arrieta Hitachi-V70). A validation group was defined (since the software update in 08/2019) to validate the predictive model for malignancy. Quantitative RGB analysis was performed using ImageJ software (NIH). The exact amount of blue (hard), green (intermediate), and red (soft tissue-elasticity) was measured and expressed in pixels and percentages. Only the tumor tissues inside well-defined margins set by the operator were analyzed. The color intensity was measured on a scale of 0-255 for an 8-bit image. The intensity ratio for each color was defined as the relation between the absolute value for this color and the intensity of the sum of all three colors (R+G+B). The final diagnosis was made by histopathology, or a combination of radiological findings, tumor markers and clinical follow-up.

Results In main cohort, 59 solid pancreatic lesions evaluated by strain elastography were included: 45(75 %) malignant-60 % adenocarcinomas, 8.3 % metastasis and 6.6 % NETs, and 14(23.3 %) benign masses. In the validation cohort, 20(76.9 %) malignant-73.1 % adenocarcinomas, 3.8 % metastasis, and 6(23.1 %) benign tumors were included. Cut-off values (CO) for 4 variables (criteria) of the main cohort correlating with the presence of malignancy were calculated: blue color (CO >55 %-Se 93.3 %, Spe 35.7 %, AUC 0.62), green color (CO < 42.5 %-Se 97.8 %, Spe 42.5 %, AUC 0.64), green color intensity ratio (CO <56 %-Se 71.1 %, Spe 78.6 %, AUC 0.76), red color intensity ratio (CO <18.5 %-Se 42.2 %, Spe 92.9 %, AUC 0.63). Good concordance between the main and the validation group was seen (Table).

Tab. 1

Criteria	Main cohort (04/2017-07/2019)	Risk of malignancy	Validation cohort (08/2019-04/2020)	Risk of malignancy
0	n = 5	0 %	n = 1	0 %
1 or 2	n = 15	60 %	n = 11	63.6 %
3 or 4	n = 39	92.3 %	n = 14	92.9 %

Conclusions Quantitative image analysis of tumor elastographies obtained by EUS may predict or exclude malignancy with high accuracy.

Citation Razpotnik M, Bota S, Essler G et al. eP452 RED-GREEN-BLUE (RGB) PROFILING OF PANCREATIC MASS-ELASTOGRAPHIES: VALIDATION OF A PREDICTIVE MODEL FOR NON-INVASIVE ASSESSMENT OF MALIGNANCY. Endoscopy 2021; 53: S245.

Publication History

Article published online:
19 March 2021