Diagnosis of liver hemangioma using magnetic resonance diffusion-derived vessel density (DDVD) pixelwise map: a preliminary descriptive study

Introduction

Liver hemangioma (HG) has an incidence rate ranging between 0.4% and 20.0%, and is commonly discovered during any abdominal imaging work-up. According to the classification system of the International Society for the Study of Vascular Anomalies, liver HG is a vascular tumor (1-3). The pathogenesis of HG is ill-understood. Consistent with a female predilection (ratio of 2–5:1), HG is considered to be a congenital disorder with possible hormonal dependence (4,5). From a histopathological perspective, these neoplasms are characterized by cavernous venous spaces delineated by a lining of vascular endothelial cells and interspersed with connective tissue septa. With the hepatic artery serving as the principal source of vascular supply, the hemodynamics within HGs is notably impaired, exhibiting a markedly reduced flow rate. Morphologically, HGs are well-defined lesion with round or lobulated margins. The primary subtypes include cavernous HG, capillary HG, and sclerosing HG. The principal criterion for this classification is the extent of fibrous tissue present within the body of HG. Cavernous HG represents the most frequent subtype with the presence of larger vascular spaces coupled with a low quantity of connective tissue, and is closely aligned with the typical imaging profile of liver HG. Capillary HGs, also known as flash-filling or rapidly-filling HGs, account for approximately 16% of all liver HGs. This subtype is notably more prevalent in HG measuring less than 1 cm in diameter. Hyalinised or sclerosed HG is unusual and is believed to represent the end stage of a HG. Due to the replacement of the vascular spaces by fibrotic tissue, it is virtually impossible to propose a definitive diagnosis for sclerosed HG based on imaging, thus pathologic proof is necessary.

The vast majority of HG cases do not require treatment or monitoring. Notably, no cases of malignant transformation with liver HG have been documented. Large HG may give rise to complications in 4.5% to 19.7% of cases, consisting of bleeding, compressive effects on adjacent structures such as the bowel and torsion if pedunculated. A giant liver HG may cause serious coagulation disorders, such as Kasabach-Merritt syndrome, which presents as hemolytic anemia, thrombocytopenia, prolonged prothrombin time, and hypofibrinogenemia, as well as Budd–Chiari syndrome (6-10). Surgical resection is rarely indicated, except in the presence of Kasabach-Merritt syndrome (11,12). Transcatheter hepatic embolization can be applied to manage Kasabach-Merritt syndrome, as can a combination of systemic corticosteroids and vincristine (1,13).

Plain CT features of liver HG usually show a hypodense well-defined lesion, with an internal density similar to the vessels. On T1-weighted (T1-w) MR images, liver HGs display low signal intensity, and on T2-weighted (T2-w) images they show very high signal intensity due to the long T2 of its blood-filled vascular channels (14,15). In rare cases, mostly in advanced cirrhosis and hepatic steatosis, HGs can lose their typical imaging features. When MRI is the first line examination for the liver, for the majority of HG cases, the diagnosis is established with the application of a contrast enhanced imaging. Dynamic studies performed after the administration of extracellular compounds exhibit early nodular peripheral enhancement, since its feeding vessels originate from the hepatic artery. Subsequent phases of liver enhancement reveal a progressive slow centripetal fill-in, with hyper-intensity on delayed phase as compared to the normal liver parenchyma. Late intra-tumoral accumulation of contrast can be explained by the slow flowing blood within its vascular channels determining the absence of noticeable washout on the latter phases of the dynamic study (1-3,16,17). Nevertheless, this distinct enhancement pattern may not be discernible in lesions smaller than 5 mm. Less typical imaging findings of HGs can result from three main causes: altered morphology or structure, unusual blood flow patterns or associated liver abnormalities (1,3,18). When the diagnosis cannot be achieved with imaging, percutaneous biopsy may be required. Provided that a cuff of normal hepatic parenchyma is interposed between the capsule and the margin of HG, needle biopsy is not contraindicated. Caldironi et al. (19) reported an overall needle biopsy diagnostic accuracy of 96%.

The diffusion-weighted imaging (DWI)-derived surrogate biomarker diffusion-derived vessel density (DDVD) works on the principle that on spin-echo type echo-planar-imaging DWI, blood vessels (including micro-vessels) show high signal when there is no motion probing gradient [b=0 (s/mm²)], while they show low signal even when very low b-values (such as b=1 or 2) are applied. Thus, the signal difference between images when the motion probing gradient is off and on reflects the extent of tissue vessel density. DDVD is derived from the equation (20):

DDVD(b0b2)=Sb0/ROIarea0−Sb2/ROIarea2[unit: arbitrary unit(au)/pixel][1]

where ROIarea0 and ROIarea2 refer to the number of pixels in the selected region-of-interest (ROI) on b=0 and b=2 DWI, respectively. Sb0 refers to the measured total signal intensity within the ROI when b=0, and Sb2 refers to the measured total signal intensity within the ROI when b=2, thus Sb/ROIarea equates to the mean signal intensity within the ROI. Sb2 and ROIarea2 can also be approximated by other low b-values (such as b=10) DWI. If we consider a pixel is an individual ROI, DDVD pixelwise map (DDVDm) can be constructed pixel-by-pixel with this same principle (21).

In some cases, a HG can be diagnosed based on typical imaging features without the need for contrast enhanced scan. However, when MRI is the first line examination for the liver, contrast enhanced scan is commonly acquired to increase the diagnostic confidence for HG. In this study, we conducted a preliminary evaluation of magnetic resonance DDVDm for diagnosing HG.

Methods

There were three testing datasets of convenient samples initially not collected for the purpose of HG evaluation. All imaging data were acquired at 3.0T. For dataset-1, an intravoxel incoherent motion (IVIM) imaging DWI sequence was initially acquired with TR of 1,600 ms and TE of 59 ms and an acquisition spatial resolution of 3.02×3.11×7 mm³. DDVDm was reconstructed with b=0 and b=2 images. There were initially 47 patients, and three patients [1.0 cm HG; 1.24 cm focal nodular hyperplasia (FNH), 1.38 cm FNH] with a combination of too small lesion and/or position shift between b=0 and b=2 images were excluded. Finally, dataset-1 consisted of 16 HGs (lesion diameter median: 3.83 cm; range, 1.5–8.85 cm), 4 FNHs (lesion diameter median: 2.24 cm; range, 1.72–5.7 cm), and 24 hepatocellular carcinomas (HCCs, lesion diameter median: 5 cm; range, 1.83–12.77 cm). For dataset-2, an IVIM DWI sequence was initially acquired, with TR of 2,500 ms and TE of 84 ms and an acquisition spatial resolution of 2.73×2.73×5 mm³. DDVDm was reconstructed with b=0 and b=10 images. There were initially 7 patients, and one patient (HG lesion diameter: 1.3 cm) with a combination of small lesion and/or position shift between b=0 and b=10 images was excluded. Finally, dataset-2 consisted of 6 HGs (lesion diameter median: 2.76 cm; range, 1.14–6.2 cm). For dataset-3, DWI with two b-values of 0, 2 s/mm² were acquired with TR of 313 ms and TE of 38 ms and an acquisition the spatial resolution of 3.04×3.04×7 mm³. DDVDm was reconstructed with b=0 and b=2 images. Dataset-3 consisted of 28 patients with 28 HCC (lesion diameter median: 5.78 cm; range, 1.91–13.52 cm) with DDVDm quality suitable for visual analysis. Detailed MR acquisition parameters for these three datasets are provided in Appendix 1. FNHs and HCCs all had pathological confirmation. HGs were diagnosed with typical contrast enhanced imaging appearances and/or with pathological confirmation. DDVDm in dataset-1, dataset-2, and dataset-3 were assigned with different pseudo-color scales, but within each dataset the same pseudo-color scale was applied to each patient’s images.

Three readers were involved in the image analysis. Reader-1 (C.Y.L.) was a senior trainee in radiology, and reader-2 (G.W.H.) and reader-3 (Y.X.J.W.) were specialist radiologists. Reader-1 and reader-2 had access to all the imaging data and clinical information of the patients of these three datasets, and some historical liver DDVDm data acquired in our unit. Reader-1 and reader-2 summarized the DDVDm features of HG and mass-forming lesion (MFL, i.e., FNH or HCC) and explained them to reader-3 (Y.X.J.W.) who had prior knowledge of liver DDVDm reading (21), and this was also aided with historical image data of liver DDVDm. Before the testing session, reader-3 had no knowledge of lesion specific information of dataset-1 and dataset-2. During the testing session for dataset-1 and dataset-2, which was completed in about 90 minutes, reader-3 was required to make a diagnosis solely based on DDVDm, and reader-1 was available to explain the location or extent of the lesions to reader-3 when asked. Reader-3 was required to make a decision for a lesion with four choices: (I) HG with confidence; (II) HG without confidence (lesion features suggesting HG but the diagnosis could not be made firmly); (III) solid MFL with confidence; (IV) solid MFL without confidence (lesion features suggesting MFL but the diagnosis could not be made firmly). Therefore, reader-3 was not asked to differentiate FNH from HCC.

After the testing session described above, three readers read dataset-3 together, to confirm whether the DDVDm features summarized from dataset-1 and dataset-2 would be generalizable to dataset-3. Reader-3 was also aware that dataset-3 only contained HCC lesions.

At the end of these readings, with reference to anatomical T1-w/T2-w images and DWI, the causes of misdiagnosis and other potential pitfalls were further analysed.

Results

The diagnostic performance of reader-3 for dataset-1 and dataset-2 during the testing session is shown in Table 1. Correct diagnosis was made in 90.9% of the HGs (72.7% with confidence) and 96.4% of the MFLs (85.7% with confidence). Figure 1 further shows that, for most of the larger lesions, the differentiation between HG and MFL was made with confidence.

Figure 1 A distribution of the diameters for HGs and MFLs according to the diagnostic outcome of reader-3 based on diffusion-derived vessel density map. Diameter is the longest length of the lesion on the slice with the largest area on diffusion-weighted imaging. HG, hemangioma; MFL, mass-forming lesion.

HGs generally showed substantially higher DDVD signal relative to background liver parenchyma (Figures 2-4). Though not necessarily, HG DDVD signals could be similar to those of blood vessels. Kidneys have higher perfusion than those of liver and spleen (22,23). Some HGs showed DDVD signals higher or similar to that of kidneys. Another sign noted in the current study is that a thick very low signal rim around, or partially around, the HG lesion (Figures 5-7, we term this as thick ‘black-out rim’). We did not see such a thick very low signal rim in MFLs in the current study. Two cases of mis-diagnosed HG lesions are shown in Figures 8,9. It was later considered that these two HGs might be associated with slow and stagnant blood flow.

Figure 2 A case of typical HG. (A) The HG lesion (red arrow) shows very high signal on b=0 DWI. (B) The HG lesion signal is attenuated on b=10 DWI. (C) The HG lesion shows high signal on DDVDm. The shape of the stomach liquid changed between b=0 DWI and b=10 DWI, leading to its high signal on DDVDm (green arrow). DWI, diffusion-weighted imaging; CSF, cerebrospinal fluid; HG, hemangioma; DDVDm, diffusion-derived vessel density pixelwise map.

Figure 3 DDVDm of five cases of liver HG. (A,B) The same lesion. An HG with high DDVD signal is indicated by arrows. (C) An HG with high DDVD signal is indicated by arrows. (D) An HG with high DDVD signal is indicated by arrow. (E) A small HG with high DDVD signal is indicated by red arrow. The gallbladder also shows high DDVD signal due to the position shift during the data acquisition of b=0 and b=10 DWI [(E,F) dotted green arrows]. (F) b=0 DWI image of the case in (E). The small HG shows high very signal on b=0 DWI. (G) An HG with high DDVD signal indicated by arrow. CSF can show high DDVD signal. Note that the colors in the image are pseudo-color and can be adjusted manually to best show the contrast between vessels and liver parenchyma. In (A-C), and (G), when a pixel has a DDVD value less than zero, this pixel is labeled as ‘white-out’ without color. DDVDm, diffusion-derived vessel density map; HG, hemangioma; DWI, diffusion-weighted imaging; DDVD, diffusion-derived vessel pixelwise density; CSF, cerebrospinal fluid.

Figure 4 DDVDm of six cases with eight liver HGs. (A) An HG with high DDVD signal is indicated by arrow. (B) An HG with high DDVD signal is indicated by arrows. (C) An HG with high DDVD signal is indicated by arrow. (D) Two HGs indicated by arrows. (E) Two HGs indicated by arrows. (F) An HG with high DDVD signal is indicated by arrow. Note that the colors in the image are pseudo-color and can be adjusted manually to best show the contrast between vessels and liver parenchyma. In the DDVDm in this figure, when a pixel has a DDVD value less than zero, this pixel is labeled as ‘white-out’ without color. DDVDm, diffusion-derived vessel density pixelwise map; HG, hemangioma.

Figure 5 An HG shows a ‘black-out’ rim. (A) b=0 DWI. A high signal HG showing the border with double rims (yellow lines), with a dark rim lying outside the bright rim (blue arrows). (B) b=2 DWI of the lesion in (A), and the same rim patterns are shown. (C) (b=0 DWI) and (D) (b=2 DWI) show the same lesion in (A) and (B) but in another slice section. (E,F) DDVDm of (A) and (B). In (F), when a signal has a DDVD value less than zero, this signal is labeled as ‘white’ without color. (E,F) A corresponding ‘black-out’ rim [denoted with lines in (E,F)]. (G) DDVDm of (C) and (D). A corresponding ‘black-out’ rim is shown (line). With the image data read in this study, such a relatively thick ‘black-out’ rim suggests HG. DWI, diffusion-weighted imaging; HG, hemangioma; DDVDm, diffusion-derived vessel density pixelwise map.

Figure 6 HGs showing a thick ‘black-out’ rim. (A,B) A case of HG (arrow) showing the ‘black out’ rim (red thin line) in (B). In the presentation in (B), this ‘black out’ rim is demonstrated as a ‘white-out’ rim. The internal DDVD signal is also high. (C-E) a case of HG (arrow) showing a thick ‘black out’ rim in (D) (yellow thin red line). (D) and (E) are the same slice section. In the presentation in (E), this ‘black-out’ rim is demonstrated as a ‘white-out’ rim. HG, hemangioma; DDVD, diffusion-derived vessel pixelwise density.

Figure 7 A case with two HGs (orange arrows) showing a relatively thick ‘black out’ rim. (A) and (B) are the same section with different pseudo-color adjustments, and the pseudo-color in (B) was additionally adjusted. A thick ‘black-out’ rim [white arrows in (A) and dotted thin line in (B)] is noted for both HGs. HG, hemangioma.

Figure 8 A case of HG misdiagnosed as MFL by reader-3. (A) The lesion (orange arrow) shows high signal on b=0 DWI, and a vessel shows high signal (green arrow). (B) The lesion (orange arrow) remains high signal on b=10 DWI, while the vessel shows signal void (green arrow). (C) The DDVD signal of the lesion (orange arrow) is lower than what is anticipated for HG. This suggests that the blood flow pattern might be more stagnant in this case. This lesion was mis-diagnosed as MFL (not high confidence). In hindsight, the DDVD signal of this lesion is indeed relatively high compared to that of the liver parenchyma. In fact, the DDVD signal of this lesion is similar to those of adjacent vessels. HG, hemangioma; MFL, mass-forming lesion; DWI, diffusion-weighted imaging; DDVD, diffusion-derived vessel pixelwise density.

Figure 9 A case of HG misdiagnosed as MFL by reader-3. (A) The lesion (orange arrow) shows high signal on b=0 DWI, and the vessel also shows high signal. (B) The lesion (orange arrow) remains high signal on b=2 DWI, while the vessel shows mostly signal void. (C) The DDVD signal of the lesion (orange arrow) is lower than what is anticipated for HG, more consistent with MFL. This suggests that the blood flow pattern might be stagnant in this case. This case had surgical pathology diagnosis of being typical cavernous HG. HG, hemangioma; MFL, mass-forming lesion; DWI, diffusion-weighted imaging; DDVD, diffusion-derived vessel density.

MFL generally showed DDVD signal only slightly higher, similar to, or even slightly lower, than that of background liver parenchyma (Figures 10,11). The case of mis-diagnosed MFL lesion is shown in Figure 12. It was later considered that a combination of peripheral necrotic foci and position shift between b=0 DWI and b=2 DWI caused artificial rim high signal on DDVDm.

Figure 10 DDVDm of six cases with MFL [(A) focal nodular hyperplasia, others: HCC] indicated by red arrows. The interior portion of these MFL demonstrates lower DDVD signal than those of HG, and the signals tend to be inhomogeneous. The overall DDVD signals of MFL tend to be slightly higher than that of liver parenchyma. A rim could be seen in all cases, and these rims are more suggestive of MFL (rather than HG). No thick ‘black-out’ rim is noted for these MFLs. In the DDVDm except that of (D), when a pixel has a DDVD value less than zero, this pixel is labeled as ‘white-out’ without color. All lesions in this figure were diagnosed as MFL with high confidence by reader-3. DDVDm, diffusion-derived vessel density map; MFL, mass-forming lesion; HCC, hepatocellular carcinoma; HG, hemangioma; DDVD, diffusion-derived vessel pixelwise density.

Figure 11 DDVDm of four cases with MFL (all HCC) indicated by red arrows. (A) A typical case of MFL indicated by arrows. This MFL has lower DDVD signal as compared with cases of HG, however the DDVD signal of this MFL is overall higher than that of the liver parenchyma. (B) A typical case of MFL indicated by arrows. The border of this case is not considered to be ‘black-out’ rim related to HG. The lesion border of this case is rather discontinuous and varies in thickness, and the interior signal is also too low for an HG. This MFL has signal lower than the background liver parenchyma. (C) A small MFL, note this lesion has pseudo ‘black-out’ rim. However, this black rim only covers half of the lesion (i.e., a half circle), and opposite this black rim is an artificially high signal half circle (arrows). This symmetrical feature suggests that it was caused by position shift during the data acquisition of b=0 and b=2 DWI. This point was also suggested by the thick dark rims around the kidneys. (D) A small MFL (arrow), the left-upper corner is an insert showing the same lesion at an adjacent slice. This lesion has lower DDVD signal than an HG, and without ‘black-out’ rim. Lesions in (A) and (B) were diagnosed as MFL with confidence by reader-3. Lesions in (C) and (D) were diagnosed as MFL without confidence. DDVDm, diffusion-derived vessel density map; MFL, mass-forming lesion; HCC, hepatocellular carcinoma; HG, hemangioma; DWI, diffusion-weighted imaging; DDVD, diffusion-derived vessel density.

Figure 12 A case of HCC was initially mis-diagnosed as HG. The misdiagnosis was due to the artificially high DDVD signal anterior rim of the lesion [(B,C) arrows], and the artificially high signal shown on (A) (arrow) which is actually a section of the rim of the tumor. T2-w (D) shows a number of focal high signal areas (arrow), which are also shown on DWI [(E,F) arrows]. There is a slight shift of position between b=0 DWI (E) and b=2 DWI (F), as shown by that a few hepatic vein branches are visible on b=2 DWI (dotted arrows) but not on b=0 DWI. Mapping of focal high signal areas on b=0 DWI onto b=2 DWI can lead to unreliable high signal. In hindsight, the internal DDVD signal interior to the rim is too low to be HG (the DDVD signal interior to the rim is close to the liver parenchyma), and this case should be diagnosed as MFL. This case would not be mis-diagnosed if the T2-w images were available for analysis. It also was later considered that the rim of this lesion is suggestive of MFL. T2-w, T2-weighted; DWI, diffusion-weighted imaging; HCC, hepatocellular carcinoma; DDVD, diffusion-derived vessel density; MFL, mass-forming lesion.

Dataset-3 were all HCC cases, and DDVDm of these cases were all consistent with MFL as agreed by the three readers. Figure 13 further shows two typical HCCs from dataset-3.

Figure 13 DDVDm of two HCC cases (from dataset 3) indicated by arrows. A typical tumor rim is noted for both cases. Consistent with the typical appearance of HCC, the interior portion of these HCC only slightly higher signal on DDVDm than that of the liver parenchyma. Images of dataset 3 had overall lower quality. DDVDm, diffusion-derived vessel density pixelwise map; HCC, hepatocellular carcinoma.

Discussion

It has been noted that the blood vessels show high signal when there is no motion probing gradient (b=0 s/mm²) and low signal when even very low b-values are applied (20). This phenomenon has not been fully validated with spin physics theory, but it is likely due to that, for spin-echo type echo-planar imaging (EPI) sequence, the second motion probing gradient after the 180-degree radiofrequency (RF) pulse could not fully re-focus the flowing spins in vessel and micro-vessels after being de-phased by the first motion probing gradient before the 180-degree RF pulse. DDVD measure based on this simple principle appears to be useful as a straightforward imaging biomarker. DDVD is a useful parameter for distinguishing of livers with and without fibrosis, and livers with severer fibrosis tend to have even lower DDVD measurements than those with milder liver fibrosis (20,24,25). Li et al. (26) applied DDVD to assess the perfusion of HCC. DDVD results (ratio of HCC DDVD to background liver DDVD equals around 3.0) approximately agree with other dynamic contrast enhanced CT/MRI literature data. Lu et al. (27) reported earlier clinical grades rectal carcinoma had a higher DDVD ratio (tumor to tumor-free rectal wall) than those of the advanced clinical grades (2.245 for grade 0&I, 1.460 for grade II, 1.430 for grade III, 1.130 for grade IV). These are all consistent with the biological behaviours of HCC and rectal carcinoma. Moreover, He et al. (28) reported that placenta DDVD as a perfusion biomarker of the placenta allows excellent separation of normal and early preeclampsia pregnancies. Lu et al. (29) reported that placenta regional DDVD is significantly higher in pregnant women with placenta accreta spectrum disorders than women with normal placenta, and especially higher in patients with placenta increta and percreta.

Liver HG has higher apparent diffusion coefficient (ADC) and T2 values than those of HCC and liver metastases, and lower ADC and T2 values than those of liver simple cysts. In some cases, a HG can be diagnosed based on typical imaging features without the need for contrast enhanced scan. However, as shown in Figure 14, a substantial portion of HG shows ADC and T2 overlapping with those of HCC and liver metastasis, and cyst (30-37). Therefore, contrast agent enhanced MRI is commonly used to confirm the diagnosis of liver HGs (38).

Figure 14 Liver HG is characteristically associated with high ADC and high T2 values. Liver HG has higher ADC and T2 values than those of HCC and liver Met, and lower ADC and T2 values than those of liver simple cysts. However, as shown in this graph, a substantial portion of HGs showing ADC and T2 values overlapping with those of HCC and liver metastases, and cysts. Therefore, contrast agent enhanced imaging is commonly used to confirm the diagnosis of liver HG. Data replotted from Yamada et al. (30), Parikh et al. (31), Kim et al. (32), Ohtomo et al. (33), Lombardo et al. (34), Cieszanowski et al. (35). (A-C) Data in mean ± standard deviation; (F) data in mean and range. Y-axis: ADC in ×10⁻³ mm²/s or T2 in millisecond (ms). ADC, apparent diffusion coefficient; HCC, hepatocellular carcinoma; Met, metastases; HG, hemangioma; FNH, focal nodular hyperplasia; ML, malignant lesions.

The analysis of DDVD requires only two b-values (with one being b=0 s/mm²), allowing a significantly shorter scanning time than contrast enhanced CT/MRI or IVIM imaging. Compared with contrast enhanced imaging, DDVD protocol does not involve contrast injection, data acquisition is faster. DDVD is conceptually as simple as ADC. Though ADC has been proposed to reflect tissue diffusion, recent analyses suggest that ADC value is heavily affected by T2 value of the tissue (39-41). In other words, T2 contribution to ADC quantification or ‘T2-shining through’ can never be eliminated. It has been recently also noted that non-invasive IVIM imaging cannot characterize HG (42). In fact, it has been shown that IVIM-perfusion fraction (PF) is also heavily affected by tissue’s T2, with longer T2 leading to a ‘depressed’ IVIM-PF measure (26,43,44). It is noted that liver simple cysts demonstrate very low signal on DDVDm (Figure 15). We anticipated that a substantial portion of ‘larger’ HGs could be diagnosed confidently with DDVDm alone, and this was confirmed in the current study. As shown in Table 1, based only on DDVDm, approximately 73% (16/22) of the HGs were diagnosed with confidence; for further 18% (4/22) of the HGs, a correct diagnosis was suggested. For the two cases (9% of the HGs) where an incorrect diagnosis was suggested, both cases might have had atypical blood flow pattern. For the MFLs, based only on DDVDm, approximately 86% (24/28) were diagnosed with confidence; for further 11% (3/28) of the MFLs, a correct diagnosis was suggested. Figure 1 suggests that larger lesions were more likely to be diagnosed with confidence. This is also the subjective experience of all three readers that large lesions could be differentiated with confidence. We noted that some cases of HG were associated with a ‘black-out rim’ (Figures 5-7). The HG’s ‘black-out rim’ differed from the rims of HCC in that HG’s ‘black-out rims’ were thicker, and HCC’s rims were more irregular. HG’s ‘black-out rim’ may be associated with the particular blood flow pattern of these HGs. More studies are needed to clarify this particular sign. It should be emphasized that the experience of reading liver DDVDm remained yet limited for the readers in this study. Diagnostic performance may improve with further accumulation of practical experience.

Figure 15 An example of a liver simple cyst (white arrow). DDVDm [b0b1, (C)] is computed from b=0 DWI (A) and b=1 DWI (B), and DDVDm [b0b2, (D)] is computed from b=0 DWI and b=2 DWI. The cyst typically shows very low signal on DDVDm. Adapted with permission from (21). DDVDm, diffusion-derived vessel density pixelwise map; DWI, diffusion-weighted imaging.

‘Position shift’ between b=0 image and b=2 (or 10) image is a major source of quantification error for DDVD calculation of the liver (21), as the liver is heavily subject to respiratory motion. Figure 16 Illustrates the principle that a ‘position shift’ between b=0 image and b=2 (or 10) image can cause artificially high, or low, or negative DDVD values. Figure 17 and Figure 18 further illustrate two cases with artificially high lesion DDVD signals. This strongly suggests that DDVDm should be viewed together with other structural images and DWI, and the extent of ‘position shift’ between b= 0 image and b=2 (or 10) image should be always noted. In fact, for the case illustrated in Figure 12, if the T2-weighted images were available for analysis, then this case would not be mis-diagnosed as HG by reader-3. For both dataset-1 and dataset-2, the images were initially acquired for IVIM analysis, without efforts being made to minimize the position shift between b=0 image and b=2 (or 10) image. One possible way to overcome this difficulty is to scan the DDVD protocol twice (or three) times, and manually select the pair of images with the most similar positions to reconstruct DDVDm. This approach may be attempted in the future particularly for smaller lesions. Breath-hold method with sufficiently long repetition time (TR), rather than free-breathing in dataset-1 and respiration-gating in dataset-2, is indeed practically feasible (20). In this study, DDVDm of dataset-3 were generally of lower image quality, possibly due to the shorter TR resulting in lower signal-to-noise ratio.

Figure 16 An illustration of potential DDVD signal error due to ‘position shift’ between b=0 and b=2 DWI. In the case of (A) with no ‘position shift’, signals of cells in columns ‘a’, ‘b’, ‘c’ on b=0 DWI are all only slightly higher than signals of corresponding cells on b=2 DWI, and the result is that cells in DDVDm all show low signal (i.e., ‘similar’ minus ‘similar’ = low). In the case of (B) with no ‘position shift’, signals of cells in columns ‘a’, ‘c’ on b=0 DWI are all only slightly higher than signals of corresponding cells on b=2 DWI, and the result is that cells in columns ‘a’, ‘c’ all show low signal on DDVDm (i.e., ‘similar’ minus ‘similar’ = low). For cells in column ‘b’, signal is much higher on b=0 DWI than on b=2 DWI, and the result is that these cells on DDVDm show high signal (i.e., ‘high’ minus ‘low’ = high). In the case of (C), position shift between b=0 and b=2 DWI occurred. Cells in column ‘a’ of b=0 DWI are mapped onto cells in column ‘b’ of b=2 DWI, column ‘a’ on DDVDm is the result of ‘high’ minus ‘low’ =high. Column ‘b’ on DDVDm is the result of ‘low’ minus ‘high’= negative (cells in column ‘b’ on b=0 DWI is mapped onto cells in column ‘c’ of b=2 DWI). Therefore, the DDVDm in case C is erroneous. Note that the illustration presented here is only one possibility for negative DDVD values. Other DDVD signal error possibilities include slice position mismatch in the longitudinal direction and the signal instability of echo-planar sequence imaging. DWI, diffusion-weighted imaging; DDVDm, diffusion-derived vessel density pixelwise map; lo: low; hi: high; (−): negative value.

Figure 17 A case of HCC with erroneous high signals on DDVDm. (A) An HCC nodule (arrow) is moderately high signal on b=0 DWI. (B) For the corresponding b=2 DWI of slice in A, position shift occurred and the HCC is not shown. (C) On DDVDm, the HCC (arrow) is shown with erroneous high signal. This is due to that the high signals of the HCC on b=0 DWI is mapped onto lower signal liver parenchyma on b=2 DWI. DWI, diffusion-weighted imaging; HCC, hepatocellular carcinoma; DDVDm, diffusion-derived vessel density pixelwise map.

Figure 18 One case of HCC with unreliable high signal on DDVDm. (A) A large HCC with heterogenous high signals on b=0 DWI. (B) Position shift occurred for b=2 DWI relative to b=0 DWI, as shown with liver fissure shape change (arrows) and change in the shown upper pole of right kidney (dotted arrows). The heterogenous high signals are not shown on b=2 DWI. (C) On DDVDm, subtraction of (A) and (B) leads to unreliable high signals (arrow) in the region of HCC and in the upper pole of the right kidney (dotted arrow). However, the heterogenous and more localized high signals in the region of the HCC on DDVDm will not lead to mistaking this tumor as HG. These erroneous heterogenous high signals should not be considered as intra-tumoral vessels. DWI, diffusion-weighted imaging; DDVDm, diffusion-derived vessel density pixelwise map; HCC, hepatocellular carcinoma.

There are many limitations to this study. This is a preliminary descriptive study with small sample size, the full diversity of HG and MFL might not have been well presented in our samples. Notably, the lesions included in the analysis of the current study were mostly relatively large (Figure 1). Smaller lesions more likely cause diagnostic challenges. Future studies should include smaller lesions with well position-matched b=0 and b=2 images. Another limitation is that we did not include liver metastases. In this study, we did not conduct quantitative analysis, as in clinical practice the radiologists are more likely to make a diagnosis based on subjective assessment of the lesion’s signal intensity (as compared with the adjacent liver parenchyma and vessels such as the aorta and the portal/hepatic veins), the signal homogeneity of the lesion, and the morphology of the lesion, rather than conducting quantitative measurement. This study grouped together FNH and HCC as MFL, the potential separation of FNH and HCC based on DDVDm will be another topic of research for the future. For the HG lesions in this study, a thick ‘black-out rims’ was noted in some cases. The reliability of this particular sign for suggesting HG diagnosis shall be assessed with more studies. We did not compare the relative performance of DDVDm constructed from b=0 and b=2 images (as in dataset-1) and from b=0 and b=10 images (as in dataset-2), but our subjective impression was that these two types of DDVDm were comparable in HG evaluation. Note that, some of the clinical MRI scanners do not allow a non-zero b-value less than 10. In this study, we did not explore the optimal pseudo-color scale to demonstrate the lesions. Our future study will allow the pseudo-color scale to be adjusted individually during the image-reading, so that liver parenchyma show (dark) blue color, main vessels show red color, and then HGs show yellow, orange, or red color (Figures 19,20). Finally, this study only assessed the diagnostic performance of DDVDm, without considering the integration of other T2, ADC, and morphological features. We can anticipate that the integration of T2, ADC, and morphological features of the lesions can further improve the diagnostic performance of DWI. Note that, in this study, we did not compare relative diagnostic performance DDVDm with ADC map or T2 map. Our goal was to demonstrate that DDVDm can be another additional parameter based on a different contrast mechanism.

Figure 19 A case of focal nodular hyperplasia. (A-C) are of one section, and (D-F) are of one section. In (C) and (F), when a pixel has a DDVD value less than zero, this pixel is labeled as ‘white-out’ without color. The lesion on [(A)/(B)/(C)] shows higher signal than the liver parenchyma. However, it is noted that the high signal is also associated with the liver capsule (arrows), and (C) shows there are large areas of mismatch during the mapping (‘white-out’ pixels). Thus, these high signals may be less reliable. Note that the liver parenchyma signal is better presented in [(D)/(E)/(F)] with dark blue color. The lesion (arrows) on section [(E)/(F)] shows almost DDVD iso-signal or only slightly higher signal relative to the liver parenchyma. The insert in (D) is b=0 DWI showing the lesion. This case was diagnosed by reader-3 as mass-forming lesion without confidence. DDVD, diffusion-derived vessel density; DWI, diffusion-weighted imaging.

Figure 20 A case of focal nodular hyperplasia (arrows). (A) b=0 DWI, (B,C) are of one section, and (D) b=0 DWI, (E,F) are of another section adjacent to [(A)/(B)/(C)]. We may prefer the pseudo-color scale shown with (B), with the liver parenchyma showing dark blue. On (C), the lesion has some yellowish colors, however it can be noted that the lesion has only slightly higher DDVD signal than the liver parenchyma. On the other hand, the lesion on section [(E)/(F)] shows almost DDVD iso-signal to the liver parenchyma. This case was diagnosed by reader-3 as MFL with confidence. DWI, diffusion-weighted imaging; DDVD, diffusion-derived vessel density; MFL, mass-forming lesion.

In conclusion, when DDVDm is used to evaluate the liver, HG can be diagnosed with confidence in a substantial portion of patients. A tentative pattern is noted that a small portion of HGs might be misdiagnosed as MFL, but a MFL is quite less likely to be misdiagnosed as HG. In practice, DDVDm should be considered together with lesion morphology and T2 signal and/or ADC features so to improve the diagnostic confidence. We anticipate that, with the integration of DDVDm to liver MRI, the number of gadolinium-contrast enhanced scans can be saved in a high proportion of patients, particularly for patients with large lesions.

Acknowledgments

We thank Mr. Dian-Qi Yao, a research student at the Chinese University of Hong Kong, for the help in image data post-processing.

Funding: This work was supported by National Natural Science Foundation of China (No. 82171893); Hong Kong GRF Project No. 14112521; Shenzhen People’s Hospital Physician Scientist Training “Five Three” Program (No. SYWGSLCYJ202404); the Hong Kong Research Grants Council (RGC) Research Impact Fund (RIF) with project No. R4015-21.

Footnote

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-24-1837/coif). Y.X.J.W. serves as the Editor-in-Chief of Quantitative Imaging in Medicine and Surgery. Y.X.J.W. is the founder of Yingran Medicals Ltd., which develops medical image-based diagnostics software. W.C. is an employee of Philips Healthcare. M.S.Y.Z. contributed to the development of Yingran Medicals Ltd. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study re-used historical image data for new analysis. All the diffusion weighed image data were acquired with the institutional ethical approval and with informed consent obtained from individual patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Kacała A, Dorochowicz M, Matus I, Puła M, Korbecki A, Sobański M, Jacków-Nowicka J, Patrzałek D, Janczak D, Guziński M. Hepatic Hemangioma: Review of Imaging and Therapeutic Strategies. Medicina (Kaunas) 2024;60:449. [Crossref] [PubMed]
2. Sandulescu LD, Urhut CM, Sandulescu SM, Ciurea AM, Cazacu SM, Iordache S. One stop shop approach for the diagnosis of liver hemangioma. World J Hepatol 2021;13:1892-908. [Crossref] [PubMed]
3. Caseiro-Alves F, Brito J, Araujo AE, Belo-Soares P, Rodrigues H, Cipriano A, Sousa D, Mathieu D. Liver haemangioma: common and uncommon findings and how to improve the differential diagnosis. Eur Radiol 2007;17:1544-54. [Crossref] [PubMed]
4. Glinkova V, Shevah O, Boaz M, Levine A, Shirin H. Hepatic haemangiomas: possible association with female sex hormones. Gut 2004;53:1352-5. [Crossref] [PubMed]
5. Giannitrapani L, Soresi M, La Spada E, Cervello M, D'Alessandro N, Montalto G. Sex hormones and risk of liver tumor. Ann N Y Acad Sci 2006;1089:228-36. [Crossref] [PubMed]
6. Liu X, Yang Z, Tan H, Xu L, Sun Y, Si S, Liu L, Zhou W, Huang J. Giant liver hemangioma with adult Kasabach-Merritt syndrome: Case report and literature review. Medicine (Baltimore) 2017;96:e7688. [Crossref] [PubMed]
7. Erdogan D, Busch OR, van Delden OM, Bennink RJ, ten Kate FJ, Gouma DJ, van Gulik TM. Management of liver hemangiomas according to size and symptoms. J Gastroenterol Hepatol 2007;22:1953-8.
8. Gandolfi L, Leo P, Solmi L, Vitelli E, Verros G, Colecchia A. Natural history of hepatic haemangiomas: clinical and ultrasound study. Gut 1991;32:677-80. [Crossref] [PubMed]
9. Sun JH, Nie CH, Zhang YL, Zhou GH, Ai J, Zhou TY, Zhu TY, Zhang AB, Wang WL, Zheng SS. Transcatheter Arterial Embolization Alone for Giant Hepatic Hemangioma. PLoS One 2015;10:e0135158. [Crossref] [PubMed]
10. Sharma V, Aggarwal A, Singla R, Kalra N, Chawla YK. Giant hemangioma causing budd-Chiari syndrome. J Clin Exp Hepatol 2014;4:380-1. [Crossref] [PubMed]
11. Hall GW. Kasabach-Merritt syndrome: pathogenesis and management. Br J Haematol 2001;112:851-62. [Crossref] [PubMed]
12. O'Rafferty C, O'Regan GM, Irvine AD, Smith OP. Recent advances in the pathobiology and management of Kasabach-Merritt phenomenon. Br J Haematol 2015;171:38-51. [Crossref] [PubMed]
13. Tlougan BE, Lee MT, Drolet BA, Frieden IJ, Adams DM, Garzon MC. Medical management of tumors associated with Kasabach-Merritt phenomenon: an expert survey. J Pediatr Hematol Oncol 2013;35:618-22. [Crossref] [PubMed]
14. Mamone G, Miraglia R. The "light bulb sign" in liver hemangioma. Abdom Radiol (NY) 2019;44:2327-8. [Crossref] [PubMed]
15. McFarland EG, Mayo-Smith WW, Saini S, Hahn PF, Goldberg MA, Lee MJ. Hepatic hemangiomas and malignant tumors: improved differentiation with heavily T2-weighted conventional spin-echo MR imaging. Radiology 1994;193:43-7. [Crossref] [PubMed]
16. Brannigan M, Burns PN, Wilson SR. Blood flow patterns in focal liver lesions at microbubble-enhanced US. Radiographics 2004;24:921-35. [Crossref] [PubMed]
17. Wilson SR, Burns PN. An algorithm for the diagnosis of focal liver masses using microbubble contrast-enhanced pulse-inversion sonography. AJR Am J Roentgenol 2006;186:1401-12. [Crossref] [PubMed]
18. Ghai S, Dill-Macky M, Wilson S, Haider M. Fluid-fluid levels in cavernous hemangiomas of the liver: baffled? AJR Am J Roentgenol 2005;184:S82-5. [Crossref] [PubMed]
19. Caldironi MW, Mazzucco M, Aldinio MT, Paccagnella D, Zani S, Pontini F, De Bellis ME, Rebuffi AG, Costantin G. Echo-guided fine-needle biopsy for the diagnosis of hepatic angioma. A report on 114 cases. Minerva Chir 1998;53:505-9.
20. Wáng YXJ. Living tissue intravoxel incoherent motion (IVIM) diffusion MR analysis without b=0 image: an example for liver fibrosis evaluation. Quant Imaging Med Surg 2019;9:127-33. [Crossref] [PubMed]
21. Yao DQ, Zheng CJ, Deng YY, Lu BL, Lu T, Hu GW, Li XM, Xiao BH, Ma FZ, Sabarudin A, King AD, Wáng YXJ. Potential diverse applications of diffusion-derived vessel density (DDVD) pixel-by-pixel mapping. Quant Imaging Med Surg 2024;14:2136-45. [Crossref] [PubMed]
22. Vallée JP, Lazeyras F, Khan HG, Terrier F. Absolute renal blood flow quantification by dynamic MRI and Gd-DTPA. Eur Radiol 2000;10:1245-52. [Crossref] [PubMed]
23. Winter JD, St Lawrence KS, Cheng HL. Quantification of renal perfusion: comparison of arterial spin labeling and dynamic contrast-enhanced MRI. J Magn Reson Imaging 2011;34:608-15. [Crossref] [PubMed]
24. Xiao BH, Huang H, Wang LF, Qiu SW, Guo SW, Wáng YXJ. Diffusion MRI Derived per Area Vessel Density as a Surrogate Biomarker for Detecting Viral Hepatitis B-Induced Liver Fibrosis: A Proof-of-Concept Study. SLAS Technol 2020;25:474-83. [Crossref] [PubMed]
25. Hu GW, Zheng CJ, Zhong WX, Zhuang DP, Xiao BH, Wáng YXJ. Usefulness of diffusion derived vessel density computed from a simplified IVIM imaging protocol: An experimental study with rat biliary duct blockage induced liver fibrosis. Magn Reson Imaging 2021;84:115-23. [Crossref] [PubMed]
26. Li XM, Yao DQ, Quan XY, Li M, Chen W, Wáng YXJ. Perfusion of hepatocellular carcinomas measured by diffusion-derived vessel density biomarker: Higher hepatocellular carcinoma perfusion than earlier intravoxel incoherent motion reports. NMR Biomed 2024;37:e5125. [Crossref] [PubMed]
27. Lu BL, Yao DQ, Wáng YXJ, Zhang ZW, Wen ZQ, Xiao BH, Yu SP. Higher perfusion of rectum carcinoma relative to tumor-free rectal wall: quantification by a new imaging biomarker diffusion-derived vessel density (DDVD). Quant Imaging Med Surg 2024;14:3264-74.
28. He J, Chen C, Xu L, Xiao B, Chen Z, Wen T, Wáng YXJ, Liu P. Diffusion-Derived Vessel Density Computed From a Simplified Intravoxel Incoherent Motion Imaging Protocol in Pregnancies Complicated by Early Preeclampsia: A Novel Biomarker of Placental Dysfunction. Hypertension 2023;80:1658-67. [Crossref] [PubMed]
29. Lu T, Wang L, Li M, Wang Y, Chen M, Xiao BH, Wáng YXJ. Diffusion-derived vessel density (DDVD) computed from a simple diffusion MRI protocol as a biomarker of placental blood circulation in patients with placenta accreta spectrum disorders: A proof-of-concept study. Magn Reson Imaging 2024;109:180-6. [Crossref] [PubMed]
30. Yamada I, Aung W, Himeno Y, Nakagawa T, Shibuya H. Diffusion coefficients in abdominal organs and hepatic lesions: evaluation with intravoxel incoherent motion echo-planar MR imaging. Radiology 1999;210:617-23. [Crossref] [PubMed]
31. Parikh T, Drew SJ, Lee VS, Wong S, Hecht EM, Babb JS, Taouli B. Focal liver lesion detection and characterization with diffusion-weighted MR imaging: comparison with standard breath-hold T2-weighted imaging. Radiology 2008;246:812-22. [Crossref] [PubMed]
32. Kim HC, Seo N, Chung YE, Park MS, Choi JY, Kim MJ. Characterization of focal liver lesions using the stretched exponential model: comparison with monoexponential and biexponential diffusion-weighted magnetic resonance imaging. Eur Radiol 2019;29:5111-20. [Crossref] [PubMed]
33. Ohtomo K, Itai Y, Furui S, Yashiro N, Yoshikawa K, Iio M. Hepatic tumors: differentiation by transverse relaxation time (T2) of magnetic resonance imaging. Radiology 1985;155:421-3. [Crossref] [PubMed]
34. Lombardo DM, Baker ME, Spritzer CE, Blinder R, Meyers W, Herfkens RJ. Hepatic hemangiomas vs metastases: MR differentiation at 1.5 T. AJR Am J Roentgenol 1990;155:55-9. [Crossref] [PubMed]
35. Cieszanowski A, Szeszkowski W, Golebiowski M, Bielecki DK, Grodzicki M, Pruszynski B. Discrimination of benign from malignant hepatic lesions based on their T2-relaxation times calculated from moderately T2-weighted turbo SE sequence. Eur Radiol 2002;12:2273-9. [Crossref] [PubMed]
36. Duran R, Ronot M, Kerbaol A, Van Beers B, Vilgrain V. Hepatic hemangiomas: factors associated with T2 shine-through effect on diffusion-weighted MR sequences. Eur J Radiol 2014;83:468-78. [Crossref] [PubMed]
37. Holzapfel K, Bruegel M, Eiber M, Ganter C, Schuster T, Heinrich P, Rummeny EJ, Gaa J. Characterization of small (≤10 mm) focal liver lesions: value of respiratory-triggered echo-planar diffusion-weighted MR imaging. Eur J Radiol 2010;76:89-95. [Crossref] [PubMed]
38. EASL Clinical Practice Guidelines on the management of benign liver tumours. J Hepatol 2016;65:386-98. [Crossref] [PubMed]
39. Wáng YXJ, Zhao KX, Ma FZ, Xiao BH. The contribution of T2 relaxation time to MRI-derived apparent diffusion coefficient (ADC) quantification and its potential clinical implications. Quant Imaging Med Surg 2023;13:7410-6. [Crossref] [PubMed]
40. Wáng YXJ, Ma FZ. A tri-phasic relationship between T2 relaxation time and magnetic resonance imaging (MRI)-derived apparent diffusion coefficient (ADC). Quant Imaging Med Surg 2023;13:8873-80. [Crossref] [PubMed]
41. Wáng YXJ. The very low magnetic resonance imaging apparent diffusion coefficient (ADC) measure of abscess is likely due to pus's specific T2 relaxation time. Quant Imaging Med Surg 2023;13:8881-5. [Crossref] [PubMed]
42. Wáng YXJ, Sabarudin A. Underestimation of liver hemangioma perfusion fraction by standard intravoxel incoherent motion diffusion magnetic resonance imaging. Quant Imaging Med Surg 2024;14:2128-35. [Crossref] [PubMed]
43. Ma FZ, Wáng YXJ. T2 relaxation time elongation of hepatocellular carcinoma relative to native liver tissue leads to an underestimation of perfusion fraction measured by standard intravoxel incoherent motion magnetic resonance imaging. Quant Imaging Med Surg 2024;14:1316-22. [Crossref] [PubMed]
44. Yu WL, Ma FZ, Huang H, Xiao BH, Li XM, Wáng YXJ. Age and gender differences of normative values of spleen diffusion MRI parameters. Rofo 2024; Epub ahead of print. [Crossref]