Liver: Detection and characterization of surgically staged focal liver lesions: Accuracy, limitations and pitfalls of dual phase spiral CT

Ihab R. Kamel, MD, PhD 1; H. J. V. Braga, MD 1; Bernard A. Birnbaum, MD 2;
Elliot K. Fishman, MD 1; Karen M. Horton, MD 1; David A. Bluemke, MD, PhD 1

Introduction

Dual phase CT is commonly utilized in the initial evaluation of patients with suspected focal liver lesions. The purpose of this exhibit is to present the use of dual phase spiral CT in the pre-surgical patient population, based on a blinded reviewer analysis of 237 surgically and histopathologically lesions, present in 77 patients. The high reliability of CT, as well as common pitfalls and limitations of dual phase helical CT are presented. Specific learning objectives include: 1) recognize the usefulness of dual phase helical CT in the detection and characterization of focal liver lesions; 2) discuss the accuracy in classifying lesions as benign or malignant; and 3) identify common errors and limitations in the classification of lesions, and potential pitfalls in lesion detection and characterization.

Results

A total of 237 lesions were confirmed at surgery and pathology, with average size of 2.3 cm (range 0.2 - 20 cm). There were 73 benign lesions (31%) including cyst (n=43), hemangioma (n=14), adenoma (n=8), focal nodular hyperplasia (n=5), focal fat (n=1), post-surgical scar (n=1), and chronic inflammation (n=1). The remaining 164 lesions (69%) were malignant. These included metastasis (n=154), hepatocellular carcinoma (n=7), cholangiocarcinoma (n=2), and adult hepatoblastoma (n=1).

Sensitivity, specificity and AUC values with the corresponding 95% confidence intervals for each of the 3 reviewers are shown in Table 1. Overall performance of all 3 reviewers was very similar, as denoted by the AFROC curve (Figure 1). Table 2 demonstrates the distribution of lesions as reported by each reviewer. The difference in the distribution of lesions between the 3 reviewers was not statistically significant (p= 0.673) as determined by Chi-squared analysis.

Errors and pitfalls in dual phase CT

A) Lesion size
Small lesions may be difficult to detect because of volume averaging with surrounding liver parenchyma, resulting in reduced lesion conspicuity. In the current study 19 of 237 lesions (8%) in 9 patients were not seen by any of the reviewers, nor by retrospective evaluation. All the lesions were small, measuring up to 1 cm (mean = 0.6 cm, range = 0.2 - 1 cm), and were all malignant. These lesions were only identified by intraoperative ultrasound and manual palpation.

B) Lesion location
Lesion detection rate varied according to the lobar and segmental distribution of the lesion. Fifteen out of 160 lesions in the right lobe were not reported, resulting in a false negative rate of 9.4%. Dome lesions (segment 8) were more likely to be missed than lesions in the remaining right lobe (Figure 2).

Left lobe lesions, particularly in the caudate lobe (segment 1) (Figures 3, 4) and left lateral segment (segments 2 and 3) (Figure 5) are more likely to be missed than right lobe lesions. Thirteen out of 68 lesions (19.1%) in the left lobe were not reported, resulting in a false negative rate that is approximately twice the rate in the right lobe.

Large predominantly exophytic lesions may be misclassified as arising from surrounding organs. In our study a 10 cm exophytic hepatocellular carcinoma was misclassified as gastrointestinal stromal tumor arising from the stomach (Figure 6).

C) Parenchymal changes
- Fatty infiltration
Lesion detection in a fatty liver may be a challenging task because the contrast between the low-attenuation lesions and the liver is decreased. In addition, fatty infiltration of the liver may occasionally be regional or focal in distribution. This could be mistaken for a liver mass or infarction. Normal vessels can be seen coursing through the areas of fatty infiltration, and there should be no mass effect. Focal fatty changes commonly appear adjacent to the falciform ligament and near the porta hepatis (Figure 7). However, when masses occur at these locations they can be mistaken for fatty infiltration (Figure 8). Nonunifrom fatty infiltration of the liver may result in focal areas of fatty sparing, which are especially common in the periportal region. These could be difficult to distinguish from mass lesions (Figure 9).

- Perfusion abnormality
Arterial phase scanning of the liver may demonstrate foci of hyperenhancement known as transient hepatic attenuation differences (THAD). These have been described in conditions that alter the balance between the hepatic arterial and portal venous blood flow. When focal, these can mimic a hypervascular tumor (Figure 10). Areas of abnormal perfusion may also be mistaken for masses (Figure 11).

- Volume averaging
Volume averaging artifacts can occur when 2 adjacent tissues differ significantly in attenuation, resulting in blurring of the borders between the different tissues. Volume averaging near the liver surface or abutting major fissures may result in false positive lesion (Figure 12).

- Pseudolesions
A pseudolesion may result from surrounding vascular structures within the liver parenchyma. These are more common near the hilum, were major vessels may overlap. Accessory or aberrant veins may be a contributing factor (Figure 13).

D) Atypical features
- Lesions mistaken for hemangioma
Hemagiomas are low-attenuating lesions with globular peripheral contrast enhancement on the portal venous phase images. They are usually subcapsular in location. Calcification (Figure 14) or vessels (Figure 15) in the wall of a lesion could result in misclassification of metastasis or other malignant lesions as hemangioma.

- Atypical hemangioma
If hemangiomas lack the classical features, they may be misclassified as malignant. Small lesions may fill completely in the arterial phase, and may be difficult to distinguish form hypervascular metastasis. When partially exophytic, hemangioma may lack the peripheral nodular enhancement, resulting in misdiagnosis (Figure 16).

- Focal nodular hyperplasia
Focal nodular hyperplasia are well defined lesions with intense homogeneous enhancement on the arterial phase images, that disappears on the portal venous phase. If a central scar is present it demonstrates delayed enhancement. The intense enhancement on the arterial phase may be difficult to distinguish from hypervascular metastasis, especially if enhancement is heterogeneous (Figure 17).

- Hepatic adenoma
Hepatocellular adenoma is a rare benign lesion, most often seen in young women, with a history of oral contraceptive use. Uncomplicated adenomas are typically solitary, sharply marginated and may contain fat. Small lesions enhance rapidly and are hyperattenuating relative to the liver. Adenomas may be mistaken for malignant lesions (Figure 18) especially if they are large, multiple or hemorrhagic.

E) Indeterminate lesions
When benign lesions have atypical features they may be considered indeterminate (Figures 19, 20). In our study 26 out of 237 lesions (11%) were considered indeterminate by at least 1 reviewer. The mean size of these lesions was 1.5 cm (range 0.2 - 4 cm). At pathology 9 lesions (35%) were benign and 17 (65%) were malignant.

A limitation of our study is that multidetector CT was not utilized. The thin collimation results in high image resolution and is expected to improve lesion detection and characterization. Multidetector dual phase CT is, therefore, expected to reduce the incidence false positive and indeterminate lesions. However, the rapid advance in image technology hinders performing the study on a large number of patients, while obtaining strong histo-pathologic correlation.

Conclusions

In our series we have found that:
1. Dual phase CT has good sensitivity and high specificity in the detection and characterization of focal liver lesions.
2. Interpretation is highly reproducible among experienced reviewers.
3. Approximately 8% of all lesions are not detected by dual phase technique.
4. Left lobe lesions are twice as likely to be missed compared to right lobe lesions. Lesions in the dome of the liver and the caudate lobe are more likely to be missed than lesions in the remaining liver segments.
5. Parenchymal liver disease, particularly fatty infiltration and focal sparing may cause difficulty in diagnosing or excluding liver masses.
6. Hemangioma is frequently the cause of false positive as well as false negative diagnosis.
7. Indeterminate lesions account for approximately 11% of lesions, two-thirds of which are malignant.

References

1. Kuszyk BS, Bluemke DA, Urban BA, et al. Portal-phase contrast-enhanced helical CT for the detection of malignant hepatic tumors: sensitivity based on comparison with intraoperative and pathologic findings. AJR 1996;166:91-95.
2. Semelka RC, Cance WG, Marcos HB, Mauro MA. Liver metastases: comparison of current MR techniques and spiral CT during arterial portography for detection in 20 surgically staged cases. Radiology 1999; 213:86-91.
3. Urban BA, McGhie PA, Fishman EK. Helical CT: diagnostic pitfalls of arterial phase imaging of the upper abdomen. AJR 2000; 174:455-461
4. Horton KM, Fishman EK. The liver: helical (spiral) and conventional CT. In: Shirkhoda A, ed. Variants and pitfalls in body imaging. Philadelphia, Pa: Lippincott Williams and Wilkins, 2000; 193-218.

Table 1: Sensitivity, specificity, and AUC values with corresponding 95% confidence interval for each of the 3 reviewers

Table 2. Distribution of lesions by reviewer. True positive: reviewer correctly identified a true lesion and correctly classified it as being benign or malignant; type I false positive: reviewer identified a false lesion, not seen by pathology; type II false positive: reviewer incorrectly classified a benign lesion as being malignant; type I false negative: reviewer missed a true lesion; and type II false negative: reviewer incorrectly scored a malignant lesion as being benign.