A Novel Computer Vision Software based on Endoscopic Ultrasound Detective Flow Imaging as Objective Evaluation of Spleen Fine vessels in the Presence of Portal Hypertension: A Prospective Pilot Study

Aims Endoscopic Ultrasound (EUS), particularly when combined with innovative imaging technologies like detective flow imaging (DFI), has emerged as a powerful diagnostic tool for assessing fine vessels (FV) and low-flow velocity without contrast agents. DFI displays clear and accurate information on blood perfusion with greater resolution and sensitivity compared to conventional technologies such as color Doppler and eFlow, and contrast agents. However, up to now these findings were considered in a subjective manner. We developed a EUS-DFI-based computer vision software (CV) that allows objective quantification of FV, providing an average of the number of FV and the maximum number of FV detected during EUS-DFI (Figure 1). In the present study, we aimed to establish the cut-off value of the number of spleen FV detected to differentiate the spleen changes caused by portal hypertension (PTH), and determine the diagnostic accuracy of the cut-off value to differentiate between normal and pathologic spleen.

Methods This is a single-center prospective, observational, pilot study aimed to determine a cut-off value of FV detected and the diagnostic accuracy of the CV to differentiate changes caused by PTH compared with normal subjects. Patients over 18 years-old with chronic liver cirrhosis and indication for EUS were included. Patients were divided into two groups, a control group (patients without PTH) and a cirrhotic group (patients with PTH). All patients underwent an EUS-DFI. A p-value<.05 will be considered statistically significant [1] [2] [3] [4] [5] [6].

Results A total of 27 patients were included in the analysis (12 without PTH and 15 with PTH). Mean age was 61 (SD: 13.9) and 70.4% were female. 38.5% of patients in the control group and 60% of cases had FV observed during EUS-DFI (Table 1). The median number of FV detected by the CV software was 7 (2.8 – 8.0) in the control group and 12 (11-14) for the PTH group (p<.001). The median maximum number of FV detected was 38 (19.5-45.3) in the control group and 32 (29-41.0) in the PTH group (p=.981). The cut-off value for the average FV detected by the software was determined at 10.5, with an area under the receiver operator curve (AUROC) of 0.933. The CV software obtained the following diagnostic accuracy metrics: sensitivity, specificity, positive and negative predictive values, and observed agreement of 93.3%, 91.6%, 93.3%, 91.6%, and 92.5%.

Conclusions In conclusion, the EUS-DFI-based CV software was capable of differentiating between patients without PTH from those with PTH with high diagnostic accuracy using a 10.5 cut-off value. The number of FV detected may potentially be used as an objective measurement to evaluate Spleen vascular changes with EUS-DFI even in the absence of splenomegaly, thus helping in establishing the prognosis of liver cirrhosis.

Publikationsverlauf

Artikel online veröffentlicht:
27. März 2025

• References

• 1 Smith-Laing G, Kingham JGC, Williams R.. Predictive value of endoscopic features of variceal bleeding. Gastrointest Endosc 1982; 28 (1): 1-4
• 2 Tandon P, Garcia-Tsao G.. Portal hypertension and variceal bleeding: An overview. Gastroenterology 2019; 156 (7): 1946-61
• 3 Sharma M, Reddy KR.. Endoscopic ultrasound in the management of esophageal varices: Current perspectives. Clin Liver Dis 2018; 12 (4): 100-5
• 4 Sanyal AJ, Bosch J, Blei A, Arroyo V.. Portal hypertension and its complications. Gastroenterology. 2008; 134 (6): 1715-28
• 5 De Franchis R.. Expanding consensus in portal hypertension: Report of the Baveno VI Consensus Workshop. J Hepatol 2015; 63 (3): 743-52
• 6 Bosch J, García-Pagán JC.. Esophageal varices: Pathophysiology, diagnosis, and treatment. J Hepatol 2020; 71 (5): 1071-83