In the normal anatomy of the hepatic artery, the CA divides into three branches . The first branch is the LGA, followed by the SA and the CHA, which divides into the GDA and PHA, while the PHA divides into the LHA and RHA. According to autopsy and early DSA report [2, 10–12], some papers even reported the sensitivity of computed tomography (CT) and magnetic resonance (MRI) angiography to describe hepatic artery variation [13–17]. Hepatic artery variation has been found to be diverse and complex. In general, the description of hepatic artery variation has been enriched and the classification has been improved [11–13, 15]. Although DSA is an invasive examination, it is an important method for vascular examination and a means of treatment. Moreover, in this study, we reported some rare variations and deficiency of the complement type. Through retrospective analysis of the origin, shape and branches of the hepatic artery by DSA, we concluded that the incidence of hepatic artery variation rate in 1141 patients with hepatic tumor was 23.7%, which was lower compared with previous reports [1–3]. First of all, our study was performed retrospectively and was based on analysis of DSA images. Moreover, regional differences, ethnic differences, and differences in selected investigation populations who all had liver tumor were considered.
The most common variation is the RHA originating from the SMA, then LHA from LGA, and CHA from SMA. Extensive MEDLINE search showed that the presence or absence of middle hepatic artery which is considered normal is independent from this . It has been reported that both the LHA and RHA can originate from the CA, AA, LGA, GDA, SMA, SA and even right renal artery [18–21]. The CHA may arise from the CA, AA, SMA and LGA . The PHA can arise from the CA, AA and SMA . Compared with the Michels  and Hiatt  classifications, which were confined to origin of the hepatic artery from the LGA and SMA. We found five new variations and proposed a new classification. The first type is normal hepatic artery anatomy.
The second type is variation in origin which is considered to be the primary type of variation, including variation in origin of the LHA, RHA, PHA and CHA. DHA belongs to a special type of variation in origin, which was analyzed separately. Fasel et al  proposed that DHA refers to the situation in which one or two hepatic arteries directly originate from the CA or AA. It is an uncommon but important variant, in which there is no CHA and the GDA may originate from either hepatic artery. As far as we know, many case reports have been made about rare variations in the literature [23–28]. In our study, we discovered uncommon cases arising from the SMA, such as LHA originating from the SMA , PHA from SMA , LHA and RHA both originating from SMA, which has not been reported before. There are rare cases of the hepatic artery originating from the LGA, including CHA originating from the LGA [3, 11]. In addition to replaced or accessory LHA and RHA from vessels other than the SMA or LGA, there are infrequent cases of GDA origin. In this study, we observed that rarely, the LHA arose from the GDA [24, 25], and the accessory RHA arose from the GDA . In rare cases of AA origin, including the CHA originated from the AA alone , and the PHA originated from the AA . Some of the original variants even have double replaced or accessory systems, including: RHA arising from the SMA and accessory RHA arising from the GDA ; accessory LHA derived from the LGA and accessory RHA derived from the SMA [3, 27]; CHA arising from the SMA and accessory LHA arising from the LGA ; and LHA arising from the GDA and RHA from the SMA, which has never been reported before. Thus, variation in origin is variable, while rare variations also play an important role in understanding tumor blood supply.
The third type is variation in bifurcation , including trifurcation, quadrifurcation and pentafurcation of the CHA. Németh et al  referred to one case of pentafurcation that underwent 3D volumetric CT reconstruction. However, it was not observed in this study, and trifurcations accounted for 1.4%, which was less than the 2.0% of patients in the study by Vandamme et al . Quadrifurcation is a peculiar variation, and we only discovered three cases (0.3%). Covey et al10 also described this category.
The fourth type is bifurcation and variation of origin, which has not been reported by DSA examination. We refer to the situation where the CHA bifurcated directly as having any variation of hepatic artery origin. We found 3 new cases that the incidence rate of this specific situation was only 0.3%. Although only a few authors detect and consider bifurcation to be a variant condition [11, 29], it does have critical surgical implications. Especially for transcatheter arterial chemoembolization (TACE), the catheter should be superselected into the liver parenchyma for injection. Obviously, practice has proved that this variation proposed was reasonable, feasible as well as necessary.
Besides, we discovered eight cases of stenosis or occlusion in the CHA, even at the root of the CA [11, 30], with multiple tortuous collateral vessels, forming a continuous anastomotic channel between the CHA and SMA, which only accounted for 0.7%. This variation was not common, and could not be classified as a specific type. Song et al.  reported the ambiguous CA anatomy and the prevalence of this persistent anastomotic channel was 0.2%. It can be seen that in the treatment of such patients with TACE, and it needs to be performed by SMA uplink superselection. The more tortuous the anastomotic channel is, the more difficult the operation.
Moreover, we found one case of liver cancer with SIT. SIT is a rare congenital disease with an incidence rate of 1:500–1:20 000 , in which the abdominal and thoracic viscera are all mirror image shifted. The etiology of visceral transposition is unknown and does not affect normal health or life expectancy, but has important surgical significance. To our knowledge, hepatocellular carcinoma (HCC) with SIT is extremely rare, and only 10 cases have been reported in Pubmed [31–35]. The condition was first described by Kanematsu et al  in 1983. Later Tao Li et al  reported a case of HCC with SIT by writing to the editor in 2007, which was treated by TACE. We also successfully treated such a patient who is still alive, and it was the second report by DSA.
We know that CHA can be divided into four types, arising from the CA, AA, SMA and LGA, and has been analyzed by CT and DSA. All forms were found in our study, including rare CHA separately originating from the AA or LGA. In previous studies the incidence of the latter was 0.4–0.5% [11, 36, 37], with a total of 15 cases. Michels  and Uva  respectively reported a case, which was found at autopsy. Two cases were found by Gruttadauria  during hepatic surgery and Waki  during pancreatic surgery. The remaining 11 cases, which were detected by CT angiography and 3D reconstruction, were reported by Okada  and Song . However, it is the first case of the CHA arising from the LGA with liver tumor treatment reported by DSA image in the world.