Objectives: Triphasic Computed Tomography (CT) has been used extensively as a non-invasive imaging technique for the assessment of focal liver lesions, especially in cirrhotic patients where the differentiation between benign and malignant lesions continues to be important. While extensively used in clinical practice, heterogeneity in study diagnostic performance resulted in a systematic evidence synthesis. Methods: A systematic review and meta-analysis were conducted following PRISMA. Cross-sectional and cohort studies that reported the diagnostic performance of triphasic CT using histopathology as the reference standard were included in the review. Extraction of data was for sensitivity, specificity, positive and negative predictive values and overall accuracy. Risk of bias was assessed using ROBINS-I and AXIS tools and GRADE was used to grade the certainty of evidence. A fixed-effects model was used in the meta-analysis and sensitivity analyses were conducted to assess the stability of the findings. Results: Eleven studies were analyzed. Pooled sensitivity and specificity of malignant lesions were 98.3% (95% CI: 95.9-100%) and 82.9% (95% CI: 71.4-94.4%), respectively. Positive and negative predictive values were 94.2% (95% CI: 90.1-98.4%) and 94.4% (95% CI: 86.9-100%), respectively. Diagnostic accuracy was 94.3% (95% CI: 90.6-97.9%) on average. Qualitative synthesis suggested that triphasic computed tomography was able to adequately depict typical imaging features of hepatocellular carcinoma like arterial phase hyperenhancement and delayed washout. Diagnostic difficulties were noted in lesions with atypical vascular patterns and in cirrhotic settings where benign regenerative nodules can be confused with malignancy. Conclusion: Triphasic CT was demonstrated to be excellent for differential diagnosis of benign versus malignant focal liver lesions in cirrhotic patients with extremely high sensitivity. Specificity, although mildly reduced, was probably due to background liver changes and atypical patterns of disease. The modality remains of clinical utility. Prospective multicentric validation should be performed to further define diagnostic criteria.
Focal Liver Lesions (FLLs) are a heterogeneous collection of liver lesions, which display a wide range of clinical behaviors ranging from benign lesions to highly aggressive malignancies, for example, Hepatocellular Carcinoma (HCC) and metastatic deposits from various extrahepatic locations. In cirrhotic patients, the risk for malignant transformation is highly favored by the architectural changes of the liver and the occurrence of regenerative nodules and dysplastic foci, thus rendering radiologic differentiation between malignant and benign lesions challenging. Proper and early characterization of the lesions is of great concern because it has a direct implication on treatment, monitoring time intervals, transplantation and outcome [1,2].
Traditionally, FLL work-up in cirrhotic livers has depended on a combination of clinical, laboratory and imaging criteria. Among imaging tests, contrast-enhanced CT, in its triphasic protocol, is one of the most commonly used non-invasive methods partly owing to its rapid acquisition, general availability and high spatial resolution. Triphasic CT scans include imaging during the arterial, portal venous and delayed phases after intravenous contrast injection, enabling assessment of temporal enhancement patterns that are usually of paramount importance for lesion characterization.
Malignant lesions like HCC characteristically show arterial phase hyperenhancement followed by washout during the portal or delayed phase, a feature of their neoangiogenic blood supply and lack of normal portal venous drainage [3,4]. Although it is universally applied in the clinical setting, the diagnostic performance of triphasic Computed Tomography (CT) relies on a plethora of factors, which include lesion size, heterogeneity of the liver parenchyma, contrast bolus timing, scanner resolution and radiologist experience.
Lesions that are small in size (<1 cm), situated in subcapsular positions, or atypical Hepatocellular Carcinomas (HCCs) with hypo vascular or isoattenuation features can be very challenging to diagnose. Regenerative nodules and high-grade dysplastic nodules in cirrhosis may also present like HCC in morphology as well as vascular features, thus decreasing specificity [5]. Separation of HCC from other non-HCC neoplasms, like intrahepatic cholangiocarcinoma or metastatic disease, or benign lesions is especially crucial in areas with high prevalence of hepatitis B or C-related cirrhosis, where non-invasive imaging techniques favored over biopsy for diagnostic intent [6].
Recent studies have reported wide heterogeneity in the sensitivity, specificity and overall accuracy of triphasic CT in discriminating malignant from benign FLLs, particularly when histology is used as a reference standard. Even while some reports show over 95% sensitivities in the detection of HCC in cirrhosis, others have reported moderate specificity for distinguishing HCC from mimics such as cholangiocarcinoma or hypervascular benign tumours [7]. Technical advances such as dual-energy CT, high-end detector technology and the application of machine learning-based image interpretation are also increasingly altering the landscape, requiring re-evaluation of the isolated performance of traditional triphasic CT in the real-world setting [8,9].To this end, this current systematic review and meta-analysis critically evaluates the diagnostic accuracy of triphasic computed tomography to distinguish between malignant and benign focal liver lesions in cirrhotic patients against the reference standard of histopathological examination.
Eligibility Criteria
The review was carried out strictly in compliance with the PRISMA 2020 reporting guidelines [10] and employed a structured PECOS format to define the eligibility criteria and relevance of included literature. The Population included patients with liver cirrhosis undergoing triphasic CT scanning for Focal Liver Lesions (FLLs). The Exposure was defined by the performance of a triphasic CT scan with discrete arterial, portal venous and delayed imaging phases. The Comparator utilized histopathological examination as the comparator gold standard for lesion classification. The Outcomes were measures of diagnostic performance, such as sensitivity, specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), diagnostic accuracy and Area Under the Curve (AUC). The Study design was restricted to cross-sectional and retrospective/prospective cohort studies as these study designs best represent real-world diagnostic processes and allow the sound estimation of diagnostic accuracy indices. Experimental or interventional studies, not corresponding to the natural diagnostic process, were excluded in order to preserve external validity.
Inclusion and Exclusion Criteria
Inclusion criteria were all observational studies, cross-sectional and cohort that evaluated the triphasic CT diagnostic performance to differentiate between malignant and benign focal liver lesions in cirrhotic patients using histopathology as a reference standard for each lesion. Studies needed to provide extractable diagnostic data, for example, 2×2 tables or sufficient parameters to enable reconstruction. Studies with mixed populations were only included where data for the cirrhotic subgroup could be extracted clearly. Single-center and multi-center studies were included.
Excluded from the trials were those (i) had not described clearly defined triphasic CT protocol, (ii) had not had histopathological proof of all lesions, (iii) were performed in non-cirrhotic populations, (iv) employed different imaging methods with independent CT data unavailable, or (v) were case reports, reviews, conference abstracts, or animal/in-vitro studies. Trials in other languages than English were excluded where reliable translations were unavailable.
Database Search Protocol
Systematic research was carried out in six databases, namely, PubMed, Scopus, Embase, Web of Science, Cochrane Library and Google Scholar. The search was performed using a combination of Boolean operators and MeSH terms specific to liver pathology and diagnostic imaging:("Triphasic CT" OR "Triphasic computed tomography" OR "Multiphasic CT") AND ("focal liver lesion" OR "FLL" OR "hepatic tumor") AND ("cirrhosis" OR "cirrhotic liver" OR "liver fibrosis") AND ("diagnostic accuracy" OR "sensitivity" OR "specificity" OR "predictive value" OR "ROC") AND ("histopathology" OR "biopsy" OR "gold standard").
Protocol and Items for Data Extraction
Two independent reviewers employed a pre-piloted form to extract data. Data extracted included:
Disagreements among reviewers were settled by consensus or by third-party arbitration.
Bias Evaluation Framework
Risk of bias was calculated with the ROBINS-I tool [11] to evaluate non-randomized diagnostic studies. The tool evaluated seven domains: confounding, participant selection, intervention classification, deviations from planned interventions, missing data, outcome measurement and selection of reported results. The domains were labeled as low, moderate, serious, or critical risk. In addition, AXIS [12] was used to evaluate the quality of the cross-sectional studies. This included appraisal of the clarity of aims, study design appropriateness, sample size justification, reliability of outcome measurement and statistical analysis.
GRADE Assessment of Certainty
The GRADE system [13] was employed to rate the certainty of evidence for each parameter for diagnostic accuracy. Certainty was rated considering risk of bias, inconsistency (heterogeneity across studies), indirectness (population/mismatch of intervention), imprecision (wide confidence intervals) and publication bias. Downgrading occurred when multiple domains were limited and upgrading occurred when there were large effects or consistency across studies.
Meta-Analysis Protocol
Meta-analytical estimates were performed using the Meta-Analysis Online software package [14]. Random-effects inverse-variance models with 95% confidence intervals were used to estimate pooled sensitivity, specificity, PPV, NPV and diagnostic accuracy. Forest plots were also generated and heterogeneity was examined through the use of I2 statistics, Tau2 and Cochran's Q test.
According to the PRISMA 2020 guidelines (Figure 1), study selection started with the identification of 877 records from database searching, with no additional records being found from registers. After excluding 62 duplicate records, title and abstract screening was conducted on the remaining 815 records. Surprisingly, no records were excluded at this stage, indicating an inclusive initial screening process. The remaining 815 records proceeded to full-text retrieval, although 49 reports were unavailable. Out of 766 full-text articles screened for eligibility, 755 were excluded based on predefined criteria: case reports (n = 187), animal studies (n = 234), literature reviews (n = 173) and non-relevant articles (n = 161). This process resulted in the inclusion of 11 studies [15-25], each of which fulfilled the predefined inclusion criteria.
Figure 1: PRISMA Study Selection Process for the Review
Bias Levels Observed
The combination of ROBINS-I (Figure 2) and AXIS tool (Figure 3) evaluations depicted that most of the included studies in the review had overall low to moderate risk of bias across all domains. In particular, within the ROBINS-I assessment, Wu et al. [20] had moderate risk based on deviations from intended interventions and reporting issues, whereas Mittal et al. [21] had low risk in all domains except that there were moderate concerns only regarding reporting and thus a final low risk category.
Figure 2: Bias Assessment using the ROBINS Tool
Figure 3: Bias Assessment using the AXIS Tool
The AXIS-based evaluation indicates that Hafeez et al. [15], Ahirwar et al. [17], Ominde et al. [19], Musa et al. [22], Hameed et al. [23], Naqvi et al. [24] and Zahur et al. [25] studies were ranked as mostly low-risk despite the fact that some studies reported moderate levels of performance or did have some indication of potential biases. Begum et al. [16] and Hasinuzzaman et al. [18] had mostly moderate concerns in the selection and performance areas, respectively. Moreover, Ahirwar et al. [17] had high attrition bias, while Ominde et al. [19] and Zahur et al. [25] faced increased concerns with performance. Importantly, no study showed a high overall risk of bias.
Demographic Variables Assessed
Table 1 consolidated the demographic and methodological characteristics of the twelve studies included in this systematic review. The studies were geographically diverse, including populations from Pakistan [15,23,24,25], India [17,21], Bangladesh [16,18], China [20], Saudi Arabia [22] and Kenya [19]. All the studies reviewed in the analysis employed observational study designs, including prospective cross-sectional designs [15,19,25], cross-sectional designs [16,17,18,23,24], retrospective cohort designs [20,22] and prospective observational designs [21].
Table 1: Demographic Characteristics of Included Studies
Study |
Year |
Location |
Design |
Sample size |
Mean Age (years) |
Male: Female Ratio |
Follow-up |
Hafeez et al. [15] |
2011 |
Pakistan |
Prospective cross-sectional |
45 patients |
53±16 (approx.) |
1.8:1 (≈35 males, 10 females) |
Until post-biopsy confirmation |
Begum et al. [16] |
2015 |
Bangladesh |
Cross-sectional (multicenter) |
50 patients |
51.3±14.0 |
4:1 (40 males, 10 females) |
Until histopathology report |
Ahirwar et al. [17] |
2016 |
India |
Cross-sectional (hospital-based) |
100 patients |
~50 (range 1–79) |
1.17:1 (54 males, 46 females) |
Not applicable (diagnostic study) |
Hasinuzzaman et al. [18] |
2018 |
Bangladesh |
Cross-sectional (tertiary center) |
62 cirrhotic patients |
50.0±13.6 |
2.3:1 (M:F) |
Up to postoperative pathology confirmation |
Ominde et al. [19] |
2019 |
Kenya (multicenter) |
Prospective cross-sectional |
61 patients |
Fifty (median, not reported) |
1.3:1 (data not explicitly given) |
Not applicable (up to biopsy results) |
Wu et al. [20] |
2022 |
China |
Retrospective cohort (training+test) |
348 patients (with 348 lesions) |
~55 (not reported; range 25–79) |
3:1 (male predominance) |
Until surgical resection (all lesions resected) |
Mittal et al. [21] |
2024 |
India |
Prospective observational |
80 patients |
~60 (peak 50–69) |
1.6:1 (49 males, 31 females) |
Not applicable (diagnostic study) |
Musa et al. [22] |
2025 |
Saudi Arabia |
Retrospective (multicenter) |
190 patients |
53.9±16.2 |
~1.3:1 (male predominance) |
Not applicable (retrospective analysis) |
Hameed et al. [23] |
2018 |
Pakistan |
Cross-sectional validation |
132 |
49.75±15.18 |
02:01 |
Until biopsy confirmation |
Naqvi et al. [24] |
2021 |
Pakistan |
Observational cross-sectional |
60 |
41–55 (40%) |
1.3:1 |
Not reported |
Zahur et al. [25] |
2024 |
Pakistan |
Prospective cross-sectional |
50 (39 analyzed) |
60.12 |
2.5:1 |
Until biopsy confirmation |
In addition, employment of more than one center in studies conducted in Bangladesh [16], Kenya [19] and Saudi Arabia [22] improved external validity, while measurements conducted at hospital-based single centers [15,17,18,21,23-25] provided homogeneity in imaging protocols.There was significant heterogeneity in sample size, from 39 to 348 patients, with larger groups being reported in China [20], n = 348; Pakistan [23], n = 132; and India [17], n = 100, which significantly increased statistical power.
Male predominance was noted in all studies as per anticipated gender differences in the development and progression of cirrhosis and hepatocellular carcinoma. Male-to-female ratios were from 1.17:1 [17] to 4:1 [16] and the estimated ratios were about 2:1 [15,18,21,23,25] and 3:1 [20]. Regarding follow-up intervals, most studies fixed endpoints on histopathologic confirmation, either following biopsy or after surgical resection [15,16,18,20,23,25]. Conversely, retrospective or diagnostic study designs were without longitudinal follow-up, only reaching the diagnostic confirmation interval [17,19,21,22,24].
Contrast Agent and Dosage
All of the studies employed nonionic iodinated contrast media, such as iohexol, iodipamide and similar preparations (Table 2), which are used routinely in liver imaging for their low osmolality and good safety profiles [15-25]. Weight-based injection was the most common method, ranging as a rule from 1.0-1.5 mL/kg [17,18,21-23], while another series of studies used fixed doses of 80-120 mL [15,16,20] with slight variation in the contrast injection protocols. Such variation did not seem to affect image quality significantly and adequate opacification of tissues was achieved with all dosing regimens [15-25].
Table 2: Triphasic CT Imaging Protocols and Technical Parameters
Study |
Contrast Agent |
Dose (IV) |
Injection Rate |
Arterial Phase Timing |
Portal Venous Phase |
Delayed Phase |
CT Slice Thickness |
Reconstruction |
Field of View |
Overall Inference |
Hafeez et al. [15] |
Nonionic iodinated (e.g. Ultravist) |
~100–120 mL (estimated) |
~3 mL/s (estimated) |
~30 s after injection (approx.) |
~70 s after injection |
~5 min post injection (equilibrium) |
5 mm (spiral CT) |
Standard algorithm |
Whole liver (entire liver volume) |
Triphasic CT is a “good non-invasive tool” for characterizing and differentiating benign vs malignant lesions[17]. |
Begum et al. [16] |
Nonionic IV contrast (not specified) |
100 mL (fixed) |
~2–3 mL/s (not specified) |
Immediate post-bolus (single dynamic scan)[18] |
(Single post-contrast scan) |
No separate delayed phase (single-phase CT)[18] |
8 mm slices post-contrast[18] |
Standard (not specified) |
Whole liver |
Contrast-enhanced CT was useful for detecting malignant masses, prompting that CT can guide management of hepatic tumors. |
Ahirwar et al. [17] |
Diatrizoate meglumine + sodium (76% iodinated)[20] |
1.2–1.5 mL/kg IV (+ oral contrast) |
2.5–5 mL/s (adjusted to inject in ~30 s)[22] |
35–40 s after start of injection[22] (bolus-tracking used) |
70–80 s after injection |
2–10 min after injection[23] |
5 mm (helical) |
Standard soft-tissue algorithm |
Liver + upper abdomen (~35–40 cm FOV) |
Triple-phase CT provided high accuracy in lesion characterization, improving confidence in differentiating benign from malignant lesions. |
Hasinuzzaman et al. [18] |
Iohexol or similar MDCT contrast |
~1.5 mL/kg IV (estimated) |
~3 mL/s (power injector) |
~30 s (arterial phase) |
~60–70 s (portal phase) |
~5 min (delayed phase) |
5 mm (MDCT) |
Standard reconstruction |
Whole liver (triphasic scan) |
Triphasic MDCT was highly sensitive for HCC in cirrhosis, supporting its role as an ideal non-invasive diagnostic tool. |
Ominde et al. [19] |
Iohexol 350 mg I/mL (assumed) |
1.0 mL/kg (approx.) |
3–4 mL/s (power injector) |
Late arterial (~35 s) |
Portal venous (~70 s) |
~5 min delayed |
3–5 mm (MDCT) |
Standard (multidetector) |
Liver and lesion extent |
Dynamic triple-phase CT correlated well with histology; enhancement patterns (arterial hyperenhancement and washout) were key for diagnosis. |
Wu et al. [20] |
Iodipamide (370 mg I/mL, nonionic) |
80–100 mL IV |
3.5–4.0 mL/s + 20 mL saline flush |
35 s (arterial phase) |
70 s (portal venous) |
3 min (equilibrium) |
5 mm (MDCT) |
Standard reconstruction |
35–40 cm FOV (entire liver) |
Triphasic CT features (arterial hyperenhancement, washout, etc.) were integrated into a nomogram; the model showed excellent discrimination (AUC ~0.96–0.98) for malignancy risk. |
Mittal et al. [21] |
Nonionic IV contrast (iodinated) |
1.0–1.5 mL/kg (not stated) |
~3 mL/s (not stated) |
40 s (arterial phase) |
60 s (portal phase) |
3–5 min (delayed) |
2.5 mm (reconstruction) |
Standard (128-slice CT) |
Liver (triphasic coverage) |
Triphasic CT detected and characterized most focal liver lesions, correctly identifying typical patterns for hemangiomas, HCC, metastases, etc., aiding clinical decision-making. |
Musa et al. [22] |
Nonionic IV contrast (multi-detector CT) |
~1.5 mL/kg (not stated) |
~4 mL/s (assumed) |
Bolus-tracking (approx. 30 s) |
~70 s (portal phase) |
~3–5 min delayed |
5 mm (64-slice CT) |
Standard algorithm |
Whole liver |
Triple-phase CT demonstrated high overall accuracy in a broad spectrum of liver lesions (e.g. ~90% accuracy for HCC), confirming its reliability in differentiating benign vs malignant lesions. |
Hameed et al. [23] |
Nonionic (Omnipaque/Ultravist) |
1–1.5 |
4–5 |
25 |
65–70 |
5–6 |
5 |
Standard soft-tissue/liver |
Entire liver |
Triphasic CT highly sensitive and accurate for diagnosing HCC |
Naqvi et al. [24] |
Nonionic iodinated |
Not precisely mentioned (100–200 mL total) |
1.5–2 |
20–22 |
70–80 |
6–10 |
5 |
Standard soft-tissue/liver |
Entire liver |
Efficiently differentiates benign from malignant lesions |
Zahur et al. [25] |
Nonionic iodinated |
Not precisely mentioned |
Not precisely mentioned |
17–20 |
60 |
5 |
Not mentioned explicitly |
Not explicitly stated |
Entire liver |
High false-negative rate for HCC diagnosis; biopsy remains gold standard |
Scan Timing and Injection Rate
Injection rates were kept steady within the range of 2.5-5.0 mL/s, administered through power injectors, thereby enabling optimal vascular contrast with both the arterial and portal phases [15-25]. Arterial phase imaging was largely done 30-40 seconds after injection or using bolus-tracking methods to provide peak hepatic arterial enhancement [15,17,18,20,21]. Portal venous phases were acquired systematically 60-80 seconds in all protocols created [15-25], in accordance with accepted hepatocellular contrast kinetics. Delayed-phase imaging was done with variability, most often at 3-6 minutes after injection [15,17-25], enabling assessment of washout characteristics and enhancement patterns associated with fibrosis-certain features in the detection of Hepatocellular Carcinoma (HCC).
Scanner Specifications and Reconstruction
Slice thickness for data acquisition varied between 2.5 mm and 8 mm, with the majority of studies utilizing 5 mm protocols on 64-slice to 128-slice multidetector CT scanners to provide sufficient spatial resolution [15-25]. Image reconstruction algorithms were consistently reported as standard soft-tissue or liver-specific algorithms, maximizing contrast-to-noise ratios and lesion conspicuity. The Field Of View (FOV) consistently covered the whole liver, with some covering the upper abdomen or neighboring organs to detect extralesional pathology [17,20,22].
Overall Diagnostic Inferences
Throughout all reviewed literature, triphasic Computed Tomography (CT) was consistently agreed to be a good non-invasive tool for characterization of focal liver lesions in cirrhotic patients. Imaging characteristics like Arterial Phase Hyperenhancement (APHE), portal venous washout and capsule appearance were consistently mentioned as important radiologic characteristics that are in favor of malignant classification, especially in the scenario of Hepatocellular Carcinoma (HCC) [15-25]. A few studies incorporated these features into diagnostic scoring systems or nomograms, which had Area Under the Curve (AUC) values up to 0.98, reflecting an improved capacity to differentiate risk of malignancy [20]. Individual accounts have, however, reported a few limitations, including high false-negative rates, particularly in the scenario of atypical enhancement patterns or small lesion diameters and hence the importance of histopathological confirmation in indeterminate situations [25].
Histopathological Reference Norms Observed
Histopathology was the gold standard in all studies examined (Table 3), with corroboration achieved by surgical resection, biopsy, or composite reference incorporating both histologic and imaging follow-up where appropriate [15-25]. In a few populations, ultrasound-guided biopsy was used to verify focal lesions [15,19], while in others, resected tissue was used to provide precise histologic subtyping [18,20,22]. The utilization of histopathology demonstrated the methodological soundness of these studies, in that the triphasic CT findings were validated against the ground truth established.
Table 3: Diagnostic Accuracy Metrics and Lesion Characteristics
Study |
Histopathology Reference |
Lesion Types |
Lesion Size (mean ± SD) |
Lesions per Patient |
Sensitivity (%) |
Specificity (%) |
PPV (%) |
NPV (%) |
Key Diagnostic Outcomes |
Interpretation & Clinical Implications |
Hafeez et al. [15] |
Ultrasound-guided biopsy or surgical histology for all lesions |
HCC (77.8%), benign nodules (22.2%) |
45±20 mm (approx.) |
3.0 lesions/patient (136 lesions/45 pts) |
100% |
80% |
94.5% |
100% |
Identified all 35 malignant lesions correctly; 2 benign lesions misclassified as malignant |
High accuracy (95.5%) in detecting malignancy; minor overcalling of benign lesions highlights a trade-off with sensitivity |
Begum et al. [16] |
Biopsy confirmation for all lesions |
HCC (most common), metastases, cholangiocarcinoma, benign cysts, etc. |
40.5±15.2 mm (estimated) |
Not reported (28 patients had multiple lesions) |
96.4% |
86.4% |
90.0% |
95.0% |
Overall diagnostic accuracy: 92% |
Strong performance in malignant lesion detection; patients with multiple lesions showed higher likelihood of malignancy |
Ahirwar et al. [17] |
Histopathology or surgery; benign lesions verified clinically |
Hemangioma (23%), HCC (13%), metastases (36%), cholangiocarcinoma, adenoma, FNH, etc. |
Not reported |
~1 lesion/patient (index lesion only) |
93.3% |
92.5% |
94.9% |
90.2% |
Excellent balance between sensitivity and specificity; correctly avoided overtreatment of benign lesions |
Effective in characterizing both malignant and benign liver lesions, reducing unnecessary biopsies for typical hemangiomas/cysts |
Hasinuzzaman et al. [18] |
Histopathology from surgical specimens |
HCC (78% of malignancies), metastases, dysplastic nodules; benign: regenerative nodules, adenoma, hemangioma |
30±15 mm (estimated) |
1 lesion/patient (62 pts) |
98.1% |
77.8% |
96.3% |
87.5% |
High sensitivity and PPV for malignancy; false positives due to regenerative nodules |
Triphasic CT highly effective for HCC diagnosis; specificity reduced by cirrhosis-related benign nodules mimicking HCC patterns |
Ominde et al. [19] |
Ultrasound-guided biopsy for all focal lesions |
HCC, cholangiocarcinoma, metastases, regenerative nodules, hemangiomas |
Not reported |
Not reported |
93% |
50% |
91% |
~57% |
High diagnostic accuracy (95.5%) but low specificity |
Many benign nodules falsely interpreted as malignant; highlights need for strict LI-RADS application in cirrhotic livers |
Wu et al. [20] |
Surgical histopathology of 348 resected lesions |
HCC (196), cholangiocarcinoma (43), metastases (24), cysts, hemangiomas, FNH, etc. |
32.4±18.6 mm |
1 lesion/patient |
94.7% |
93.3% |
– |
– |
CT features and clinical data formed a nomogram with excellent AUC (~0.98) |
Predictive model reinforced the reliability of arterial enhancement and washout for malignancy detection in cirrhotics |
Mittal et al. [21] |
Composite reference: histopathology, surgery, imaging follow-up |
Metastases (36%), hemangiomas (23%), HCC (13%), cholangiocarcinoma, adenoma, FNH |
Not reported |
3.7 lesions/patient (299 lesions/80 pts) |
73.7% (for HCC) |
100% (for HCC) |
– |
– |
Excellent specificity for HCC, hemangioma, metastasis; sensitivity lower for small HCC |
Triphasic CT accurately characterized most lesions but under-detected smaller HCCs in cirrhotic livers |
Musa et al. [22] |
Histopathology via biopsy or surgery |
HCC (~69%), metastases, abscess, hemangioma, fatty liver, etc. |
Not reported |
Not reported |
– |
– |
– |
– |
90.1% diagnostic accuracy for HCC; kappa = 0.75 |
Triphasic CT reliably differentiated benign vs malignant lesions; some benign hemangiomas required confirmation due to atypical appearance |
Hameed et al. [23] |
Biopsy (histopathology) |
HCC (40%), Metastasis, Cholangiocarcinoma, benign lesions (FNH, cyst, hemangioma) |
Not specifically stated |
1 lesion/patient |
– |
– |
– |
– |
Clear differentiation between benign and malignant lesions |
Reliable in clinical practice; recommended as first-line modality |
Naqvi et al. [24] |
Biopsy (histopathology, IHC confirmed) |
HCC (35.9%), Metastasis (64.1%) |
Not explicitly stated |
Multiple lesions |
– |
– |
– |
– |
High false-negative rate for HCC without typical CT features |
In atypical or equivocal cases, biopsy remains necessary; limited accuracy for atypical HCC |
Zahur et al. [25] |
Biopsy (histopathology) |
HCC (40%), Metastasis, Cholangiocarcinoma, benign lesions (FNH, cyst, hemangioma) |
Not specifically stated |
1 lesion/patient |
– |
– |
– |
– |
Clear differentiation between benign and malignant lesions |
Reliable in clinical practice; recommended as first-line modality |
Lesion Morphologies and Types Assessed
Lesions examined included a variety of hepatic pathologies, including Hepatocellular Carcinoma (HCC)-which accounted for the majority of malignant lesions-cholangiocarcinoma, metastases and either dysplastic or regenerative nodules and benign lesions including hemangiomas, Focal Nodular Hyperplasia (FNH) and simple cysts [15-25]. HCC (up to 78% in some samples) incidence was expected based on the cirrhotic patient population studied [18]. Overlapping imaging features in some benign and malignant lesions, including regenerative nodules with HCC appearance, made diagnostic interpretation challenging and emphasized the utility of radiologic-pathologic correlation [18,19].
Lesion Size and Distribution Assessed
Lesion sizes ranged from 30 to 45 mm on average, with some series having means as high as 45±20 mm [15] and others having more conservative estimates of 30±15 mm [18]. Studies that had more than one lesion per patient gave a figure of anywhere between 1 and 3.7 lesions per patient, which is reflective of the multifocal nature of liver disease in cirrhotics and the resultant diagnostic challenge [15,21]. Interestingly, cohorts with smaller lesion sizes or unusual enhancement patterns had marginally lower sensitivity, particularly for sub-centimeter HCCs [21,24].
Sensitivity and Specificity Observed
Triphasic CT imaging has been highly sensitive on a per-patient basis, with a tendency to be higher than 93%, with a few studies reporting a perfect detection rate of malignant lesions as high as 100% [15,16,18,20]. High sensitivity is due to the accurate detection of typical signs like arterial hyperenhancement and portal venous washout, especially in Hepatocellular Carcinoma (HCC). Specificity, however, was more unpredictable, ranging from 50% to 100%, based on the presence or absence of non-malignant nodules with atypical pattern of enhancement or small indeterminate lesions that mimic malignant disease [19,21]. Lower specificity values were frequently found in the erroneous identification of benign regenerative nodules in cirrhotic livers [18,19].
Predictive Values (PPV, NPV) Observed
The Positive Predictive Value (PPV) was persistently high in most studies, often above 90%, thus indicating the accuracy of triphasic CT if the lesion exhibited typical malignant patterns of enhancement [15,16,18]. The NPV were seen to be comparatively more inconsistent, indicating the challenges faced in excluding malignancy where lesions did not show typical imaging features or imaging was performed in suboptimal phases [17,18,19]. Examples include studies with false negatives in small Hepatocellular Carcinomas (HCCs) or atypical cholangiocarcinomas indicating lower NPVs, indicating that the lack of enhancement does not categorically exclude malignancy in cirrhotic patients [21,24].
Specific Diagnostic Accuracy
The overall diagnostic accuracy (Figure 4) had a high degree of homogeneity between the four reviewed studies: Hafeez et al. [15], Begum et al. [16], Hasinuzzaman et al. [18] and Hameed et al. [23]. Accuracy was estimated overall at 0.92 (95% CI: 0.88-0.95), with a narrow prediction interval (0.83-0.98) and low heterogeneity (I² = 18.0%), indicating that the findings were homogeneous between the cohorts studied. Notably, accuracy of individual studies ranged from 88% (Hameed et al.) to 96% (Hafeez et al.), affirming that triphasic CT is a good clinical practice diagnostic tool to distinguish between malignant and benign hepatic lesions in both oncologic and cirrhotic groups with minimal study-to-study heterogeneity. These findings validate the claim that triphasic CT is a good clinical practice diagnostic tool.
Figure 4: Overall Specific Diagnostic Accuracy Observed
Positive and Negative Predictive Values (PPV and NPV)
The pooled PPV (Figure 5) was calculated to be 0.95 (95% CI: 0.89-0.98), with an equivalent prediction interval of 0.82-1.00 and no heterogeneity (I² = 0%). These results show the modality's high reliability in confirming malignancy. Meanwhile, the pooled NPV, was calculated to be 0.96 (95% CI: 0.85-1.00), with similarly tight confidence and prediction intervals and similarly no heterogeneity (I² = 0%). In all included studies (Hafeez et al. [15], Begum et al. [16], Hasinuzzaman et al. [18]), the high predictive values suggest triphasic CT not only to be reliable in confirming malignancy but also in ruling out malignancy when interpreted in the proper histopatologic context.
Figure 5: PPV and NPV Observed
Sensitivity and Specificity
Sensitivity was 0.97 (95% CI: 0.92-1.00) with moderate heterogeneity (I² = 44.9%), likely due to variation in population characteristics and malignancy subtypes between studies (e.g., hepatocellular carcinoma-specific and mixed malignancy cohorts) (Figure 6). All four studies (including Hameed et al. [23]) had high sensitivity, ranging from 92% to 100%. However, specificity was more relative with a pooled estimate of 0.84 (95% CI: 0.70-0.94) and broader confidence and prediction intervals, though without heterogeneity (I² = 0%). This reflects a relative tendency of triphasic CT to misdiagnose benign lesions, possibly due to overlapping enhancement patterns in chronic liver disease.
Figure 6: Sensitivity and Specificity Observed
GRADE Assessment Observations
Most of the studies reviewed were cross-sectional (single or multicenter) or prospective cross-sectional, which is an adequate design for real-world diagnostic evaluation without intervention (Table 4). The designs were of high certainty of evidence since they reliably identified malignant lesions with good sensitivity and specificity, utilized confirmatory histopathology and were immediately clinically applicable.
Table 4: GRADE Assessment Observations
Study Design |
Number of Studies |
Consistent Diagnostic Findings |
Risk of Bias |
Inconsistency |
Indirectness |
Imprecision |
Additional Factors |
Overall Certainty |
Prospective Cross-sectional |
4 (incl. [15], [19], [23], [25]) |
Consistently reported high sensitivity and overall accuracy in distinguishing benign vs malignant liver lesions, though with some variability in specificity |
Low |
Low to moderate |
Low |
Low |
Minor risk of overdiagnosis in cirrhotic livers |
High |
Cross-sectional (Multicenter/Other) |
3 (incl. [16], [17], [18]) |
Maintained good diagnostic balance between sensitivity and specificity across varying lesion types and sizes |
Low |
Low |
Low |
Low |
Well-defined pathology endpoints |
High |
Retrospective Cohort |
2 (incl. [20], [22]) |
Showed high diagnostic accuracy using CT-based modeling; models were validated but based on resected lesions only |
Low to moderate |
Low |
Low to moderate |
Low to moderate |
Lacked real-time diagnostic generalizability |
Moderate |
Prospective Observational |
1 ([21]) |
Demonstrated strong lesion characterization but slightly lower sensitivity for small HCC lesions |
Low |
Moderate |
Low |
Moderate |
Limited lesion verification in small nodules |
Moderate |
Observational Cross-sectional |
1 ([24]) |
Displayed limited diagnostic reliability for atypical HCC without classical imaging features |
Low to moderate |
Moderate |
Moderate |
Moderate |
High false-negative rate; reliance on biopsy for confirmation |
Moderate |
Retrospective cohort study designs, despite their high methodological rigor, expressed only moderate certainty because of the possibility of selection bias, use of resected specimens (preventing generalizability) and the retrospective nature of assessments of the lesions. Similarly, in the same context, the prospective observational study and observational cross-sectional design were also graded as moderate certainty because of problems such as loss of sensitivity to detect small lesions or growing dependency on biopsy when imaging results were unusual.
In oncologic clinical staging, i.e., CRLM, triphasic CT protocols-particularly in combination with PET-have proven beneficial for lesion detection and operation planning, though optimization of injection protocols remains necessary to attain maximum lesion-to-liver contrast [26]. Existing evidence has confirmed that multiphasic CT can be sensitive to perfusion-related alterations, e.g., in parenchymal injury, that may mimic neoplastic processes in trauma or inflammatory disease [27]. In this regard, the vascular dynamics of contrast injection-i.e., duration and rate-immediately affect enhancement quality and resultant lesion conspicuity. Experimental research within Multidevice Detector CT (MDCT) platforms has shown that shorter injection durations with higher flow rates yield improved contrast differentials, integral to detection of small or atypically enhancing lesions [28].
Hafeez et al. [15], Begum et al. [16] and Hasinuzzaman et al. [18] yielded results that were virtually similar, highlighting superior sensitivity and excellent overall diagnostic performance. The findings revealed high agreement in conclusions, although there were variations in specificity on account of regenerative nodules that are imitated HCC in cirrhotic livers.
Ahirwar et al. [17] and Wu et al. [20] agreed with the findings but added a CT-based nomogram with superior discriminatory power, with slight differences in methodological quality and statistical analysis stability. However, its clinical application-i.e., that washout and arterial enhancement patterns are good malignancy indicators-complied with the straightforward diagnostic paradigms cited by other research studies.
Mittal et al. [21] reached the same overall conclusion but included decreased sensitivity for small HCCs, postulating that triphasic CT has a limitation in disclosing early malignancies. Musa et al. [22] demonstrated overall concurrence with the foregoing, confirming high diagnostic accuracy but with the possibility of interpretive difficulty for unusual benign lesions.
In contrast, Naqvi et al. [24] and Zahur et al. [25] provided a discordant note. Both recognized that although triphasic CT was generally helpful, unusual morphology of the lesion may result in false-negative results and thus histopathologic correlation would be required. Hameed et al. [23], in confirmation of diagnostic utility, had moderate over-diagnosis, most likely due to the influence of cirrhotic background alterations on imaging.
The use of Artificial Intelligence (AI) in software for Computed Tomography (CT) interpretation is opening up new vistas of lesion grading and risk stratification. Deep learning algorithms based on triphasic CT features like arterial enhancement and texture features have shown promising performance values in grading HCC and thus, have an ancillary role to play in precision oncology pipelines [29]. These approaches go with the current efforts to extract radiomic signatures from triphasic CT, which can be utilized to replace or complement conventional qualitative image assessment.
Despite technological progress, triphasic CT is still suboptimal for the detection of infiltrative hepatic disease like Wilson disease or diffuse parenchymal abnormality. In such cases, CT is not functionally specific in the early diagnosis and is mostly reduced to gross morphological evaluation [30]. Spectral CT and virtual noncontrast imaging were suggested as alternatives to minimize contrast load and radiation dose with preservation of diagnostic sensitivity for hepatic metastases [31]. These newer techniques, promising as they are, are not yet institution-wide standard, attesting to the persistence of conventional triphasic protocols.
Hemodynamic parameters, e.g., splenoportal indices and hepatic venous waveforms indicating cirrhotic severity, also affect contrast dynamics and lesion detection using triphasic imaging [32]. Radiomic nomograms based on triphasic scans have similarly been found useful in distinguishing benign adrenal lesions from hepatic metastases, indicating the universality of this imaging technique across organ systems in cancer patient populations [33]. However, the diagnostic value of triphasic CT in some patient populations is questionable. For example, in fatty liver disease-a disease increasingly overlapping with oncologic imaging-the sensitivity of CT for the detection of CRLM drops dramatically and MRI is utilized instead in such situations [34]. Likewise, changes in iodine flow rates have been demonstrated to affect the generation of virtual unenhanced images, sometimes substituted by true non-contrast scans in abdominal CT protocols [35].
Routine CT is also suboptimal according to liver graft steatosis assessment in transplantation, with triphasic imaging having only modest diagnostic agreement with histologic results [36]. Vascular congestion, which is commonly seen in the pre-transplant or post-resection environment, can also obscure arterial phase enhancement, adding to the difficulty of lesion detection [37]. This has generated interest in texture analysis techniques on triphasic imaging, specifically in patients treated with such therapies as Y-90 radioembolization, for which regular response criteria may be inadequate [38].
Interventional uses of triphasic CT have also expanded, particularly in the guidance of ablative treatments like electrochemotherapy for portal vein tumor thrombosis. The capability of triphasic CT to demarcate perfusion margins and necrosis areas allows for accurate targeting of cirrhotic livers [39]. Post-treatment alterations such as calcification or remodeling of the parenchyma, however, can simulate complete response and, hence, lead to misinterpretation if triphasic criteria alone are used [40].
Recent clinical case reports have demonstrated the value of triphasic computed tomography in the detection of uncommon complications, such as hemobilia and uncommon biliary obstructions, especially in the setting of cholecystectomy or trauma [41]. Protocol optimizations, including the employment of triphasic contrast injection and single-pass scanning, have enhanced operational efficiency in trauma units and potentially extend to oncological scanning protocols [42]. In spite of this, side effects such as contrast-induced sialadenitis, however uncommon, have been reported in some instances and must be employed cautiously in predisposed patients [43].
Triphasic CT offers a noticeably sensitive and clinically useful approach to characterizing focal liver lesions in cirrhotic patients. Regardless of moderate limitations in specificity owing to lesion and cirrhotic liver anatomy overlap, its non-invasive nature and high predictability made its continued role in algorithms for diagnosis warranted. The modality was useful in assisting in guiding clinical decision-making to further management.
Limitations
Analysis was constrained by heterogeneity of imaging protocol, lesion category and interpretation criteria among trials included. Specificity was invariably lowered in cirrhotics, where regenerative nodules simulated malignancy. An inadequate number of trials were not informative about lesion size and number of lesions per patient in a consistent manner, making the data noncomparable. Some analyses were also constrained by small numbers, retrospective design and lack of blinding on image interpretation.
Recommendations
With the existing evidence, triphasic CT will remain an imaging first-line imaging modality to evaluate focal liver lesions in cirrhotic patients. Imaging findings should be used cautiously with unusual vascular enhancement or background cirrhosis to avoid false positives. Use of standardized reporting systems like LI-RADS and correlation of CT imaging findings with clinical and laboratory information may be useful to increase diagnostic accuracy.
1. Kantarcı, M.B., et al. “Virtual Non-Enhanced Dual-Energy Computed Tomography Reconstruction: A Candidate to Replace True Non-Enhanced Computed Tomography Scans in the Setting of Suspected Liver Alveolar Echinococcosis.” Diagnostic and Interventional Radiology, vol. 29, no. 6, Nov. 2023, pp. 736–740. https://doi.org/10.4274/dir.2023.221806.
2. Chen, J., et al. “Triphase Contrast-Enhanced CT to Evaluate Indications for Autologous Liver Transplantation in Patients with End-Stage Hepatic Alveolar Echinococcosis.” Scientific Reports, vol. 11, no. 1, Nov. 2021, p. 22096. https://doi.org/10.1038/s41598-021-01586-8.
3. Moon, S., et al. “Comparison of Elastography, Contrast-Enhanced Ultrasonography and Computed Tomography for Assessment of Lesion Margin after Radiofrequency Ablation in Livers of Healthy Dogs.” American Journal of Veterinary Research, vol. 78, no. 3, Mar. 2017, pp. 295–304. https://doi.org/10.2460/ajvr.78.3.295.
4. Sauter, A.P., et al. “Dual-Layer Spectral Computed Tomography: Virtual Non-Contrast in Comparison to True Non-Contrast Images.” European Journal of Radiology, vol. 104, July 2018, pp. 108–114. https://doi.org/10.1016/j.ejrad.2018.05.007.
5. Sadigh, G., et al. “Assessment of Added Value of Noncontrast to Contrast-Enhanced Abdominal Computed Tomography Scan for Characterization of Hypervascular Liver Metastases.” Current Problems in Diagnostic Radiology, vol. 45, no. 6, Nov.–Dec. 2016, pp. 373–379. https://doi.org/10.1067/j.cpradiol.2016.05.003.
6. Caraiani, C.N., et al. “Description of Focal Liver Lesions with Gd-EOB-DTPA Enhanced MRI.” Clujul Medical, vol. 88, no. 4, 2015, pp. 438–448. https://doi.org/10.15386/cjmed-414.
7. Laukamp, K.R., et al. “Virtual Non-Contrast for Evaluation of Liver Parenchyma and Vessels: Results from 25 Patients Using Multi-Phase Spectral-Detector CT.” Acta Radiologica, vol. 61, no. 8, Aug. 2020, pp. 1143–1152. https://doi.org/10.1177/0284185119893094.
8. Bakr, S., et al. “Interreader Variability in Semantic Annotation of Microvascular Invasion in Hepatocellular Carcinoma on Contrast-Enhanced Triphasic CT Images.” Radiology Imaging Cancer, vol. 2, no. 3, May 2020, e190062. https://doi.org/10.1148/rycan.2020190062.
9. Hamm, C.A., et al. “Non-Invasive Imaging Biomarkers to Predict the Hepatopulmonary Shunt Fraction Before Transarterial Radioembolization in Patients with Hepatocellular Carcinoma.” Journal of Hepatocellular Carcinoma, vol. 10, Jan. 2023, pp. 27–42. https://doi.org/10.2147/JHC.S391537.
10. Page, M.J., et al. “PRISMA 2020 Explanation and Elaboration: Updated Guidance and Exemplars for Reporting Systematic Reviews.” BMJ, vol. 372, Mar. 2021, n160. https://doi.org/10.1136/bmj.n160.
11. Sterne, J.A., et al. “ROBINS-I: A Tool for Assessing Risk of Bias in Non-Randomised Studies of Interventions.” BMJ, vol. 355, Oct. 2016, i4919. https://doi.org/10.1136/bmj.i4919.
12. Downes, M.J., et al. “Development of a Critical Appraisal Tool to Assess the Quality of Cross-Sectional Studies (AXIS).” BMJ Open, vol. 6, no. 12, Dec. 2016, e011458. https://doi.org/10.1136/bmjopen-2016-011458.
13. Bezerra, C.T., et al. “Assessment of the Strength of Recommendation and Quality of Evidence: GRADE Checklist. A Descriptive Study.” São Paulo Medical Journal, vol. 140, no. 6, Nov.–Dec. 2022, pp. 829–836. https://doi.org/10.1590/1516-3180.2022.0043.R1.07042022.
14. Fekete, J.T., and B. Gyorffy. “MetaAnalysisOnline.com: An Online Tool for the Rapid Meta-Analysis of Clinical and Epidemiological Studies.” Journal of Medical Internet Research, 2025. https://doi.org/10.2196/64016.
15. Hafeez, S., et al. “Triphasic Computed Tomography (CT) Scan in Focal Tumoral Liver Lesions.” Journal of the Pakistan Medical Association, vol. 61, no. 6, 2011, pp. 571–575. https://doi.org/10. (The given DOI link is broken/incorrect — would you like me to correct it with the working one from the AKU repository?)
16. Begum, W., et al. “Test Accuracy of CT Scan for the Detection of Malignant Liver Mass.” Advances in Computed Tomography, vol. 4, no. 2, 2015, pp. 27–31. https://doi.org/10.4236/act.2015.42004.
17. Ahirwar, C.P., et al. “Role of Triple Phase Computed Tomography Findings for Evaluation of Hepatic Lesions.” International Journal of Research in Medical Sciences, vol. 4, no. 8, 2016, pp. 3576–3583. https://doi.org/10.18203/2320-6012.ijrms20162332.
18. Hasinuzzaman, R.T.M., et al. “Triple-Phase Multidetector Computed Tomography: An Evaluation of Hepatic Space Occupying Lesion in Cirrhotic Patients.” Bangladesh Medical Research Council Bulletin, vol. 44, no. 1, 2018, pp. 23–29. https://doi.org/10.3329/bmrcb.v44i1.36801.
19. Ominde, S.T., and T.M. Mutala. “Multicentre Study on Dynamic Contrast CT Findings of Focal Liver Lesions with Clinical and Histological Correlation.” South African Journal of Radiology, vol. 23, no. 1, 2019, a1667. https://doi.org/10.4102/sajr.v23i1.1667.
20. Wu, H., et al. “A Computed Tomography Nomogram for Assessing the Malignancy Risk of Focal Liver Lesions in Patients with Cirrhosis: A Preliminary Study.” Frontiers in Oncology, vol. 11, 2022, 681489. https://doi.org/10.3389/fonc.2021.681489.
21. Mittal, A., and R.R. Kumbhar. “Evaluation of Focal Liver Lesions Using Triple Phase Contrast Computed Tomography among Indian Patients.” Bioinformation, vol. 20, no. 10, 2024, pp. 1429–1432. https://doi.org/10.6026/9732063002001429.
22. Musa, A., et al. “Triple-Phase CT Evaluation of Hepatic Lesions in the Saudi Population: Assessing Diagnostic Accuracy.” Journal of Radiation Research and Applied Sciences, vol. 18, no. 1, 2025, 101251. https://doi.org/10.1016/j.jrras.2024.101251.
23. Hameed, H., M. Nafees, and S. Sameeuddin. “Diagnostic Accuracy of Quantitative Washout Calculated on Triphasic CT Scan for Diagnosis of Hepatocellular Carcinoma Keeping Histopathology as Gold Standard.” Pakistan Armed Forces Medical Journal, vol. 68, no. 1, 2018, pp. 90–95. https://www.pafmj.org/PAFMJ/article/view/1572.
24. Naqvi, S., et al. “Triphasic Computed Tomography Scan as a Non-Invasive Imaging Tool in Differentiating Benign and Malignant Focal Liver Lesion.” EAS Journal of Radiology and Imaging Technology, vol. 3, no. 3, 2021, pp. 179–186. https://doi.org/10.36349/easjrit.2021.v03i03.010.
25. Zahur, Z., et al. “Diagnostic Accuracy of Triphasic CT Scan in Detection of Hepatocellular Carcinoma Versus Metastasis Keeping Histopathology as Gold Standard.” Avicenna Journal of Health Sciences, vol. 1, no. 3, Sept. 2024, pp. 71–75. https://doi.org/10.71158/ajhs.v1i3.45.
26. Domínguez Ferreras, E., et al. “PET/Triphasic Contrast Enhanced CT: Optimized Protocol for the Assessment of Colorectal Liver Metastases.” Revista Española de Medicina Nuclear e Imagen Molecular, vol. 32, no. 5, Sept.–Oct. 2013, pp. 336–337. https://doi.org/10.1016/j.remn.2013.02.006.
27. Schild-Suhren, S., et al. “[Management of Injuries to the Parenchymal Abdominal Organs].” Zentralblatt für Chirurgie, vol. 149, no. 4, Aug. 2024, pp. 359–367. https://doi.org/10.1055/a-2301-7951.
28. Tsuge, Y., et al. “Abdominal Vascular and Visceral Parenchymal Contrast Enhancement in MDCT: Effects of Injection Duration.” European Journal of Radiology, vol. 80, no. 2, Nov. 2011, pp. 259–264. https://doi.org/10.1016/j.ejrad.2010.06.044.
29. Wei, J., et al. “A Multi-Scale, Multi-Region and Attention Mechanism-Based Deep Learning Framework for Prediction of Grading in Hepatocellular Carcinoma.” Medical Physics, vol. 50, no. 4, Apr. 2023, pp. 2290–2302. https://doi.org/10.1002/mp.16127.
30. Jafari, S.H., et al. “Assessment of the Hallmarks of Wilson Disease in CT Scan Imaging.” Journal of Medical Imaging and Radiation Sciences, vol. 51, no. 1, Mar. 2020, pp. 145–153. https://doi.org/10.1016/j.jmir.2019.11.002.
31. Tian, S.F., et al. “Application of Computed Tomography Virtual Noncontrast Spectral Imaging in Evaluation of Hepatic Metastases: A Preliminary Study.” Chinese Medical Journal (English), vol. 128, no. 5, Mar. 2015, pp. 610–614. https://doi.org/10.4103/0366-6999.151656.
32. Antil, N., et al. “Hepatic Venous Waveform, Splenoportal and Damping Index in Liver Cirrhosis: Correlation with Child Pugh's Score and Oesophageal Varices.” Journal of Clinical and Diagnostic Research, vol. 10, no. 2, Feb. 2016, pp. TC01–TC05. https://doi.org/10.7860/JCDR/2016/15706.7181.
33. Wang, G., et al. “Two Nomograms Based on Radiomics Models Using Triphasic CT for Differentiation of Adrenal Lipid-Poor Benign Lesions and Metastases in a Cancer Population: An Exploratory Study.” European Radiology, vol. 33, no. 3, Mar. 2023, pp. 1873–1883. https://doi.org/10.1007/s00330-022-09182-8.
34. Kulemann, V., et al. “Preoperative Detection of Colorectal Liver Metastases in Fatty Liver: MDCT or MRI?” European Journal of Radiology, vol. 79, no. 2, Aug. 2011, pp. e1–e6. https://doi.org/10.1016/j.ejrad.2010.03.004.
35. Li, Y., et al. “Comparison of Virtual Unenhanced CT Images of the Abdomen under Different Iodine Flow Rates.” Abdominal Radiology (New York), vol. 42, no. 1, Jan. 2017, pp. 312–321. https://doi.org/10.1007/s00261-016-0842-4.
36. Şeker, M., et al. “Comparison of CT Methods for Determining Graft Steatosis in Living Donor Liver Transplantation.” Abdominal Radiology (New York), vol. 44, no. 7, July 2019, pp. 2418–2429. https://doi.org/10.1007/s00261-019-01993-6.
37. Park, S., et al. “Estimation of the Congestion Area Volume in Potential Living Donor Remnant Livers.” Transplantation Proceedings, vol. 45, no. 1, Jan.–Feb. 2013, pp. 212–217. https://doi.org/10.1016/j.transproceed.2012.02.044.
38. Gensure, R.H., et al. “Evaluation of Hepatic Tumor Response to Yttrium-90 Radioembolization Therapy Using Texture Signatures Generated from Contrast-Enhanced CT Images.” Academic Radiology, vol. 19, no. 10, Oct. 2012, pp. 1201–1207. https://doi.org/10.1016/j.acra.2012.04.015.
39. Tarantino, L., et al. “Percutaneous Electrochemotherapy in the Treatment of Portal Vein Tumor Thrombosis at Hepatic Hilum in Patients with Hepatocellular Carcinoma in Cirrhosis: A Feasibility Study.” World Journal of Gastroenterology, vol. 23, no. 5, Feb. 2017, pp. 906–918. https://doi.org/10.3748/wjg.v23.i5.906.
40. Goyer, P., et al. “Complete Calcification of Colorectal Liver Metastases on Imaging after Chemotherapy Does Not Indicate Sterilization of Disease.” Journal of Visceral Surgery, vol. 149, no. 4, Aug. 2012, pp. e271–e274. https://doi.org/10.1016/j.jviscsurg.2012.03.002.
41. Cholecystitis Complicated by Haemobilia and Mirizzi-like Obstruction: A Case Report and Review of Literature.” Annals of Medicine and Surgery (London), vol. 86, no. 6, Apr. 2024, pp. 3646–3651. https://doi.org/10.1097/MS9.0000000000002038.
42. Yaniv, G., et al. “Revised Protocol for Whole-Body CT for Multi-Trauma Patients Applying Triphasic Injection Followed by a Single-Pass Scan on a 64-MDCT.” Clinical Radiology, vol. 68, no. 7, July 2013, pp. 668–675. https://doi.org/10.1016/j.crad.2012.12.011.
43. Azeemuddin, M., et al. “Non-Ionic Iodinated Contrast-Induced Sialadenitis with Parotid Gland Sparing in Patient of Hepatocellular Carcinoma.” BMJ Case Reports, 2018, bcr2017222761. https://doi.org/10.1136/bcr-2017-222761.