Utility of dual energy CT angiography in the evaluation of acute non-variceal gastrointestinal hemorrhage: comparison with digital subtraction angiography.

PURPOSE

To evaluate the utility of dual energy CT angiography (DECTA) in acute non-variceal gastrointestinal hemorrhage (ANVGIH) compared to digital subtraction angiography (DSA) as gold standard.

MATERIALS AND METHODS

111 Patients (mean age: 39.2 years; 94 males) of ANVGIH who underwent both DECTA and DSA between January 2016 and September 2021 were included. Virtual monochromatic (VM) images at 10 keV increments from 40 to 70 keV and blended (120kVp equivalent) images of arterial phase of DECTA were evaluated independently by two readers blinded to DSA information. Quantitative analysis included measurement of attenuation in the major arteries (abdominal aorta, celiac artery, superior mesenteric artery), suspected vascular lesion, and lesion feeding artery to calculate contrast-to-noise ratios (CNRs) and signal-to-noise ratios (SNRs). Qualitative analysis assessed the image quality of each data set using a 3-point Likert scale. Findings on DSA were evaluated by a third reader and both DECTA and DSA were compared.

RESULTS

On linear blended images, vascular lesion was identified by reader 1 in 88 (79.3%) and by reader 2 in 87 (78.4%) patients and DSA showed lesion in 92 (82.9%) patients. The sensitivity and specificity of blended images and VM images of DECTA for lesion detection were not significantly different from each other. The CNR and SNR of arteries, vascular lesion and feeding artery were significantly higher at 70 keV (p < 0.005) compared to blended and other VM images. Although subjective scores for image quality were higher for 60 keV images by both readers, the difference was not statistically significant (p = 0.3). The interobserver agreement was mostly good.

CONCLUSION

In the assessment of ANVGIH, the 60 keV and 70 keV VM images improved the image quality and contrast, respectively, but there was no increase in diagnostic accuracy of VM image datasets compared to linearly blended images. Hence, the diagnostic utility of DECTA in ANVGIH is still uncertain.