One of the keys to successful TACE lies in seeing and reaching tumor lesions after accurate and comprehensive assessment of all the feeding arteries. Conventionally, operators rely on fluoroscopy and digital subtraction angiography (DSA). However, the intrinsic two-dimensional nature of DSA and its low contrast resolution limit tumor detection, resulting in a recent adoption of intraprocedural cone beam computed tomography (CBCT) for its axial and three-dimensional (3D) volumetric visualization, as well as superior soft tissue sensitivity [8,9,10,11].
To our knowledge, no study comparing the two modalities has been conducted in such an endemic area as Egypt.
Our included study population were 21 males (84.0%) and 4 females (16.0%) that emphasized the more prevalence of HCC in males going with the results of Liu et al. [12] who showed that HCC was more prevalent in males than females but with no exact clear etiology of that difference except for the hypothesis of the protective effect of female sex hormones against HCC development as most of the detected HCC were in post-menopausal females with higher mean age of females with HCC than that of males. This was also the case in our study as the mean age of the male group was 55.67 ± 6.814 years while that of the female group was 58.25 ± 2.363 years. Holah et al. [13] and Ziada et al. [14] reached the same conclusion that HCC was more prevalent in males than females.
The whole study group mean age was 56.08 ± 6.350 years. Similar results were reported by Ziada et al. [14], Mohamed et al. [15], and Tangkijvanich et al. [16] who reported a mean age of 57.5, 56, and 52.6 years respectively for HCC in high risk areas.
CBCT showed higher tumor detectability than angiography; the number of focal lesions detected by CBCT in our study was 87 while only 51 focal lesions were detected by angiography (Fig. 4). Similar results were reported by Yao et al. [17] who studied 43 HCC patients in which CBCT detected 99 lesions while angiography detected only 82 lesions. That emphasized the higher tumor detectability of dual-phase CBCTs than DSA. The mean number of lesions detected in each patient in our study by CBCT was 3.48 lesions while in the study by Yao et al. [17], it was 2.3 lesions; this could be attributed to the fact that our study was conducted in a high-risk region (Menoufia, Egypt) in which there is a high incidence of HCC.
Similar results about the tumor detectability where reported by Zheng et al. [18], Kakeda et al. [19], and Wang et al. [20] where cone beam CT outperformed angiography (DSA) in all the three studies.
We were able to detect 23 lesions less than 1 cm in diameter by dual-phase cone beam CT while only 1 lesion was detected by angiography. Similar results were reported by Lee et al. [21] in which all tumors less than 1 cm were detected by CBCT (23 tumors). Lee et al. [21] accounted the high detectability of CBCT to the small lesions less than 1 cm to the small focal spot and the high spatial resolution of CBCT more than the multidetector CT and the conventional angiography.
Similar results about the detection of small focal lesions less than 1 cm were reported in the meta-analysis done by Pung et al. [22] who reported that tumors more than 30 mm were easier to be detected by cone beam CT; however, the advantage of the high-contrast, three-dimensional images remained for tumors less than 10 mm and the tumors that could not be detected by angiography (Fig. 5).
In contrast, Kakeda et al. [19] reported that CBCT imaging did not detect tumors but detected the feeding arteries. This is believed to be because their evaluation was based only on MIP and VR images, with small or weakly enhanced nodules not being revealed during three-dimensional reconstruction. In our study, evaluation was made using three-dimensional MPR images as well as MIP and VR images. Evaluation with MPR images is considered necessary for the identification of small nodules and nodules with weak enhancement. It is believed that CBCT imaging could become a useful tool for identification of such small nodules.
As regards the feeding arteries, conventional angiography detected 87 feeding arterial branches to 43 target lesions with a mean of 3.48 ± 1.674 branches for each target lesion. On the other hand, cone beam computed tomography-hepatic artery phase (CBCT-HA) detected 130 branches to the same number of target lesion with a mean of 5.18 ± 2.264 branches for each target lesion. Miyayama et al. [23] reported similar results as among 100 tumor-feeding vessels, 81 were identified with cone beam CT manual feeding branch detection and 38 with non-selective DSA. They compared both methods with multidetector CT revealing statistically significant higher detectability of tumors and tumor-feeding branches with MFD than with non-selective DSA (both P < .001). They were able to calculate the P value as they compared both to multidetector CT results which were not available in our study.
Similar results about the detectability of feeding branches by cone beam CT and angiography were reported by Ushijima et al. [24] who reported that 53 feeding arteries were associated with the studied 39 HCC nodules. Among them, 21 arteries were identified by DSA imaging (angiography from the proximal hepatic artery); however, 26 additional feeding arteries were identified by combined application of DSA and CBCT imaging. Moreover, Minami et al. [25] also reported that the CBCT tracking navigation imaging detected 66 out of 73 feeding branches (90.4%) while celiac trunk non-selective angiography detected only 37 out of 73 feeding branches (50%). They used selective angiography to each lesion as a gold standard, but in our study, we only compared the number of feeders detected by each modality.
The meta-analysis done by Pung et al. [22] concluded that the overall detectability of the tumor feeding branches by CBCT is much better than that of conventional angiography whether selective, non-selective, or celiac trunk angiography. However, there are factors that negatively influenced the accuracy of cone beam CT which includes the small caliber of the feeding artery, poor patients’ breath hold, adjacent arterio-portal shunts, or streak artifact from metal clips.
Our study reported that the confidence of tumor feeding branches was much better using CBCT than angiography; there was good confidence in 80% of the reviewed arteries, fair confidence in 15.3% of the reviewed feeders, and poor confidence in 4.7% of the reviewed feeders. But when observing the feeders of the same lesions detected by angiography, good confidence was in 57.5%, fair in 31%, and poor in 11.5% of the reviewed feeders. This is because of the 3D capability of CBCT over 2D angiography as the liver, specially its right and caudate lobes, show many overlapping anterior and posterior branches in the 2D images of the antero-posterior direction of angiography and DSA. These can be easily separated using the combination of the MPR, MIP, and 3D VR reconstructions of the CBCT images as we did in our study.
Similar results were reported by Ushijima et al. [24] who concluded that CBCT imaging was better in identifying feeding arteries associated with lesions located in the right hepatic lobe, as this lobe is deep antero-posteriorly and its anterior and posterior branches often overlap, leading to poor isolation in 2D images of the antero-posterior direction.
For the similar cause of the overlapping arteries and the better spatial resolution, Lee et al. [26] reported that CBCT was superior to angiography in tumor feeding branches tractability which is an important advantage when performing chemoembolization for HCC with CBCT hepatic arteriography as it enables the demonstration of subtle feeding arteries and provide a three-dimensional roadmap. This also agrees with our results.
In spite of the superiority of CBCT in tumor detection and feeding vessels isolation, we observed that the images’ quality is affected by motion, catheter-induced, and lipiodol-induced artifacts more than does conventional angiography and DSA. Our study reported that out of 145 angiography acquisition, 94 acquisitions had no motion, catheter-induced, or lipiodol-induced artifacts while out of 106 CBCT acquisitions, only 8 acquisitions had no artifacts.
Our study revealed that 6 CBCT acquisitions were severely compromised by all types of artifacts and had bad image quality while only 1 angiography acquisition was severely compromised by all types of artifacts and had a bad image quality. That is attributed to the fact that CBCT acquisition requires more time (about 12 s) of no motion, breath holding, and steady contrast injection rate while angiography takes only about 4 to 6 s and also CBCT has exaggerated beam hardening. Similar results were reported by Lee et al. [26].
The whole liver was covered during image acquisition in 89.6% of angiography acquisitions and 28.3% of CBCT acquisitions which demonstrates the superiority of angiography over CBCT in liver coverage in the same acquisition. Lee et al. [26] reported that in 29 of 100 cases, CBCT covered the whole liver while incomplete coverage was found in the remaining 71 cases. They noticed that the areas most often missed from the field of view of cone beam CT were the left lateral segments and the inferior portion of segment VI, and this was also observed in our study.