Accuracy of software-assisted detection of tumour feeders in transcatheter hepatic chemoembolization using three target definition protocols

doi:10.1016/j.crad.2013.09.005

Clinical Radiology

Volume 69, Issue 2, February 2014, Pages 145-150

https://doi.org/10.1016/j.crad.2013.09.005 Get rights and content

Aim

To compare the accuracy of computer software analysis using three different target-definition protocols to detect tumour feeder vessels for transarterial chemoembolization of hepatocellular carcinoma.

Materials and methods

C-arm computed tomography (CT) data were analysed for 81 tumours from 57 patients who had undergone chemoembolization using software-assisted detection of tumour feeders. Small, medium, and large-sized targets were manually defined for each tumour. The tumour feeder was verified when the target tumour was enhanced on selective C-arm CT of the investigated vessel during chemoembolization. The sensitivity, specificity, and accuracy of the three protocols were evaluated and compared.

Results

One hundred and eight feeder vessels supplying 81 lesions were detected. The sensitivity of the small, medium, and large target protocols was 79.8%, 91.7%, and 96.3%, respectively; specificity was 95%, 88%, and 50%, respectively; and accuracy was 87.5%, 89.9%, and 74%, respectively. The sensitivity was significantly higher for the medium (p = 0.003) and large (p < 0.001) target protocols than for the small target protocol. The specificity and accuracy were higher for the small (p < 0.001 and p < 0.001, respectively) and medium (p < 0.001 and p < 0.001, respectively) target protocols than for the large target protocol.

Conclusion

The overall accuracy of software-assisted automated feeder analysis in transarterial chemoembolization for hepatocellular carcinoma is affected by the target definition size. A large target definition increases sensitivity and decreases specificity in detecting tumour feeders. A target size equivalent to the tumour size most accurately predicts tumour feeders.

Introduction

Transarterial chemoembolization (TACE) is an accepted locoregional therapy for managing unresectable hepatocellular carcinoma (HCC).1, 2 Accurate detection of tumour feeder vessels by intraprocedural imaging is indispensable for the technical success of this procedure. However, in manual assessments using two-dimensional (2D) angiography, sequential angiographic acquisitions are usually necessary to identify feeder vessels accurately. Further, additional angiographic acquisitions at different angles are often required in patients with complex hepatic arterial vasculature.

Recent advances in C-arm cone-beam technologies have enabled the visualization of three-dimensional (3D) vascular anatomy with a single acquisition using non-selective C-arm computed tomography (CT).3, 4 Further, C-arm CT has comparable ability to multidetector CT in detecting HCC lesions.5, 6 Such capabilities of C-arm CT allow the angiographic operator to visualize the vascular tree associated with the target tumour from multiple projections during the interventional procedure.7, 8, 9, 10, 11

A computer software program specifically designed to assist in planning selective liver tumour embolization (FlightPlan for Liver; GE Healthcare, Waukesha, WI, USA) was recently developed to detect tumour feeders using 3D C-arm CT data.¹² When the catheter entry site and the target tumour are chosen on multiplanar reformatted (MPR) C-arm CT images, the software automatically predicts feeder vessels by displaying a colour-coded image on the workstation screen. Previous pilot studies12, 13, 14, 15 reported that in comparison with manual angiographic assessments, the software showed better sensitivity in detecting tumour feeders and a shorter processing time. However, none of these studies investigated the specificity and overall accuracy of the software in detecting tumour feeder vessels.

In the present study, software-assisted detection of tumour feeder vessels was attempted using three different target definition sizes, and the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy obtained was compared for each target protocol.

Section snippets

Study design

The records of 57 patients with 81 HCC lesions (mean size 18.1 mm; size range 5–47 mm) who underwent selective TACE with computer software assistance using a C-arm cone-beam angiographic system for detection of tumour feeder vessels between September 2011 and March 2013 were reviewed. Patients with diffuse and infiltrative HCC or an extrahepatic supply to the tumour were excluded from this study, because TACE was performed in these patients without software assistance. Non-selective C-arm CT

Results

The TACE results showed 108 tumour feeders supplying 81 HCC lesions. By using the small, medium, and large target sizes, the software analysis identified 92, 111, and 154 possible tumour feeders, respectively. By using the extra-large target definition, 208 possible tumour feeders were identified. The numbers of false-positive vessels identified with the small, medium, and large target sizes were five, 12, and 50, respectively. The numbers of false-negative vessels identified with the small,

Discussion

Vessel-detection software programs that can facilitate selective transcatheter hepatic embolization have been recently developed. Pichon et al.¹² first described the clinical feasibility of such software programs. They analysed 15 liver tumours and found that the sensitivity and PPV of software-assisted detection of tumour feeder vessels to be 89% and 94%, respectively. Solomon et al.¹³ also reported that the software-assisted approach had a sensitivity of 80% for tumour feeder detection in six

References (20)

C.M. Lo et al.
Randomized controlled trial of transarterial lipiodol chemoembolization for unresectable hepatocellular carcinoma
Hepatology
(2002)
J.M. Llovet et al.
Arterial embolisation or chemoembolisation versus symptomatic treatment in patients with unresectable hepatocellular carcinoma: a randomised controlled trial
Lancet
(2002)
M.J. Wallace et al.
Three-dimensional C-arm cone-beam CT: applications in the interventional suite
J Vasc Interv Radiol
(2008)
R.C. Orth et al.
C-arm cone-beam CT: general principles and technical considerations for use in interventional radiology
J Vasc Interv Radiol
(2008)
J. Iwazawa et al.
Clinical utility and limitations of tumor-feeder detection software for liver cancer embolization
Eur J Radiol
(2013)
B. Schulz et al.
Radiation exposure to operating staff during rotational flat-panel angiography and C-arm cone beam computed tomography (CT) applications
Eur J Radiol
(2012)
J. Iwazawa et al.
Detection of hepatocellular carcinoma: comparison of angiographic C-arm CT and MDCT
AJR Am J Roentgenol
(2010)
H. Higashihara et al.
Diagnostic accuracy of C-arm CT during selective transcatheter angiography for hepatocellular carcinoma: comparison with intravenous contrast-enhanced, biphasic, dynamic MDCT
Eur Radiol
(2012)
S. Hirota et al.
Cone-beam CT with flat-panel-detector digital angiography system: early experience in abdominal interventional procedures
Cardiovasc Intervent Radiol
(2006)
B.C. Meyer et al.
Contrast-enhanced abdominal angiographic CT for intra-abdominal tumor embolization: a new tool for vessel and soft tissue visualization
Cardiovasc Intervent Radiol
(2007)

There are more references available in the full text version of this article.

Cited by (12)

Accuracy of a CBCT-based virtual injection software for vessel detection during hepatic arterial embolization
2022, European Journal of Radiology
To assess the accuracy, sensitivity, positive predictive value (PPV) and interobserver agreement of a virtual injection (VI) software that simulates selective arterial injection from nonselective cone-beam CT (CBCT) arteriography.
From March 2019 to May 2020, 20 consecutive patients in whom a nonselective injected CBCT and a selective CT angiography (CTA) were completed in the same procedure, were retrospectively included. The position of the microcatheter tip used for selective CTA injection was identified. The VI was simulated from the exact same point on the nonselective CBCT and the two volumes were merged. VI was compared to the real injection on the selective CTA. Three interventional radiologists evaluated the accuracy using a 6-point scale (Perfect; Good; Fair; Incorrect Origin; False Negative; Non existing). Sensitivity, PPV, and Fleiss’ kappa were calculated. Numerical variables were presented as means ± standard deviations.
Twenty procedures and 195 vessel segments were analyzed. Most vessels were 4th order (57/195; 29%) and 5th order (96/195; 49%). VI was classified as perfect to good in 96.8% ± 1.4 of 1st–3rd order arteries and in 83.4% ± 0.4 of 4th–5th order arteries. Interobserver agreement was substantial (Fleiss’ kappa = 0.79; 95% confidence interval = 0.73–0.84, P < 0.01). False negatives were reported with a mean of 9.4% ± 0.3. Average sensitivity was 90.6% ± 0.3 and average PPV was 92.7% ± 0.02. Fourteen false positives were noted.
CBCT-based VI software accurately simulated distal injections in the liver with high sensitivity and a substantial interobserver agreement.
Assessment of automated cone-beam CT vessel identification software during transarterial hepatic embolisation: radiation dose, contrast medium volume, processing time, and operator perspectives compared to digital subtraction angiography
2018, Clinical Radiology
Citation Excerpt :
Further studies could consider associations with tumour size, shape, location, and type as predictors of an AMS clinical value. Technical limitations of the AMS, related to ROI size selection and vessel segmentation parameters have been reported to impact relevant vessel detection,10,12,15,21 and may have been amplified by real-time use of the AMS during TAE relative to previous retrospective studies. In conclusion, A-CBCT is known to aid in the identification and catheterisation of arterial vascular supply to hypervascular liver tumours.
To evaluate arterial cone-beam computed tomography (A-CBCT) automated analysis software for identification of vessels supplying tumours during transarterial hepatic embolisation (TAE).
This study was approved by the institutional review board, with waiver of consent. Consecutive TAE procedures using arterial mapping software (AMS), and performed between February 2014 and August 2014, were reviewed. Hepatic arteries were imaged using digital subtraction angiography (DSA) as well as A-CBCT processed with AMS. Interventional radiologists reported¹ potential embolisation target vessels computed using AMS versus DSA alone,² modification of the embolisation plan based on AMS, and³ operator confidence related to technical success. Imaging set-up, processing time, radiation dose, and contrast media volume were recorded.
Thirty of 34 consecutive procedures were evaluated retrospectively. At least one additional embolisation target vessel was identified using AMS in 13 procedures (43%, 95% confidence interval [CI]: 26–61%) and embolisation plan modified in 11 (37%, 95% CI: 19–54%). Radiologists reported AMS increased operator confidence and reduced the number of DSA acquisitions in 25 (83%, 95% CI: 70–97%) and 15 cases (50%, 95% CI: 32–68%), respectively. The average A-CBCT acquisition and processing time was 4 minutes 53 seconds and 3 minutes 45 seconds, respectively. A-CBCT contributed to 11% of the radiation dose and 18% of the contrast media volume.
Physicians report increased tumour supplying vessel detection and intraprocedural confidence using AMS during TAE without substantial impact on radiation dose, contrast media volume, and procedure time.
Sensitivity and Reproducibility of Automated Feeding Artery Detection Software during Transarterial Chemoembolization of Hepatocellular Carcinoma
2018, Journal of Vascular and Interventional Radiology
To evaluate the performance of automated feeder detection (AFD) software (EmboGuide; Philips Healthcare, Best, The Netherlands) on hepatocellular carcinoma (HCC) tumors during transarterial chemoembolization.
Forty-four first-time transarterial chemoembolization patients (37 men; mean age, 62 ± 11 years) were enrolled between May 2012 and July 2013. A total of 86 HCC lesions were treated (2.0 ± 1.4 lesions per patient; 27.6 ± 15.9 mm maximum diameter). One hundred forty-seven feeding arteries were found with digital subtraction angiography (DSA), cone-beam computed tomography (CT), and AFD software with the option of manual adjustment (MA). Three independent interventional radiologists analyzed the cone-beam CT images retrospectively with and without AFD and MA. Compared with the number of treated vessels, the number of true positives, false positives, false negatives, sensitivity, and interreader agreement were determined using clustered binary data analysis.
Cone-beam CT enabled detection of 100 ± 3.5 feeding arteries (70% sensitivity) with 68.6% agreement among readers. AFD software significantly improved detection to 127±0.6 feeding arteries (86% sensitivity, P = .008) with 99.7% reader agreement and reduced the number of false negatives from an average of 47 ± 3.5 to 20 ± 0.6 (P = .008). MA of the AFD results produced similar feeding artery detection rates (127 ± 5.1, 86% sensitivity, P = .8), with lower interreader agreement (91.6%) and slightly fewer false positives (16 ± 0.0 to 14 ± 2.5, P = .4).
AFD software significantly improved feeding artery detection rates during transarterial chemoembolization of HCC lesions with better user reproducibility compared with cone-beam CT alone. In conjunction with DSA, AFD enables maximum feeding artery detection in this setting.
The Role of Cone-Beam CT in Transcatheter Arterial Chemoembolization for Hepatocellular Carcinoma: A Systematic Review and Meta-analysis
2017, Journal of Vascular and Interventional Radiology
Citation Excerpt :
Finally, the study by Miyayama et al (11) compared detection rates between 2 phases of a dual-phase cone-beam CT scan but without a gold standard. Thus, there were 15 study series included (10,12–25), with 2 studies reporting both tumor detection and detection of tumor supplying arteries (12,20). The baseline characteristics of the 15 included studies are presented in Tables 1 and 2.
To review available evidence for use of cone-beam CT during transcatheter arterial chemoembolization in hepatocellular carcinoma (HCC) for detection of tumor and feeding arteries.
Literature searches were conducted from inception to May 15, 2016, in PubMed (MEDLINE), Scopus, and Cochrane Central Register of Controlled Trials. Searches included “cone beam,” “CBCT,” “C-arm,” “CACT,” “cone-beam CT,” “volumetric CT,” “volume computed tomography,” “volume CT,” AND “liver,” “hepatic*,” “hepatoc*.” Studies that involved adults with HCC specifically and treated with transcatheter arterial chemoembolization that used cone-beam CT were included.
Inclusion criteria were met by 18 studies. Pooled sensitivity of cone-beam CT for detecting tumor was 90% (95% confidence interval [CI], 82%–95%), whereas pooled sensitivity of digital subtraction angiography (DSA) for tumor detection was 67% (95% CI, 51%–80%). Pooled sensitivity of cone-beam CT for detecting tumor feeding arteries was 93% (95% CI, 91%–95%), whereas pooled sensitivity of DSA was 55% (95% CI, 36%–74%).
Cone-beam CT can significantly increase detection of tumors and tumor feeding arteries during transcatheter arterial chemoembolization. Cone-beam CT should be considered as an adjunct tool to DSA during transcatheter arterial chemoembolization treatments of HCC.
Cone-beam CT angiography for determination of tumor-feeding vessels during chemoembolization of liver tumors: Comparison of conventional and dedicated-software analysis
2016, Journal of Vascular and Interventional Radiology
To compare the ability of dedicated software and conventional cone-beam computed tomography (CT) analysis to identify tumor-feeding vessels in hypervascular liver tumors treated with chemoembolization.
Between January 2012 and January 2013, 45 patients (32 men, mean age of 61 y; range, 27–85 y) were enrolled, and 66 tumors were treated (mean, 32 mm ± 18; range, 10–81 mm) with conventional chemoembolization with arterial cone-beam CT. Data were independently analyzed by six interventional radiologists with standard postprocessing software, a computer-aided analysis with FlightPlan for liver (FPFL; ie, “raw FPFL”), and a review of this computer-aided FPFL analysis (“reviewed FPFL”). Analyses were compared with a reference reading established by two study supervisors in consensus who had access to all imaging data. Sensitivities, positive predictive values (PPVs), and false-positive (FP) ratios were compared by McNemar, χ², and Fisher exact tests. Analysis durations were compared by Mann–Whitney test, and interreader agreement was assessed.
Reference reading identified 179 feeder vessels. The sensitivity of raw FPFL was significantly higher than those of reviewed FPFL and conventional analyses (90.9% vs 83.2% and 82.1%; P < .0001), with lower PPV (82.9% vs 91.2% and 90.6%, respectively; P < .0001), higher FP ratio (17.1% vs 9.4% and 8.8%, respectively; P < .0001), and greater interreader agreement (92% vs 80% and 79%, respectively; P < .0001). Reviewed FPFL analysis took significantly longer than both other analyses (P < .0001).
The FPFL analysis software enabled a fast, accurate, and sensitive detection of tumor feeder vessels.
Automatic three-dimensional detection of prostatic arteries using cone-beam CT during prostatic arterial embolization
2015, Journal of Vascular and Interventional Radiology
Citation Excerpt :
The present study evaluates software that automatically detects PAs by using nonselective dual-phase cone-beam CT. Automatic vessel-tracking software has shown promising results in the detection of tumor-feeding branches to hepatocellular carcinoma during transarterial chemoembolization (14). The present study suggests that vessel-tracking software can also be used to detect PAs (sensitivity, 92%) and to avoid multiple DSA runs before embolization.
The purpose of this study was to evaluate the automatic three-dimensional detection of prostatic arteries (PAs) with the use of dual-phase cone-beam computed tomography (CT) imaging and vessel-tracking software during prostatic artery (PA) embolization (PAE). In six patients, six right PAs and five left PAs were detected by using the software (sensitivity, 92%). The false-positive arteries (right side, 14%; left side, 25%) were deleted after cone-beam CT review. Automatic software detection of PAs from cone-beam CT may permit identification of the PA during PAE.

View all citing articles on Scopus

View full text

Accuracy of software-assisted detection of tumour feeders in transcatheter hepatic chemoembolization using three target definition protocols

Aim

Materials and methods

Results

Conclusion

Introduction

Section snippets

Study design

Results

Discussion

Hepatology

Lancet

J Vasc Interv Radiol

J Vasc Interv Radiol

Eur J Radiol

Eur J Radiol

Detection of hepatocellular carcinoma: comparison of angiographic C-arm CT and MDCT

AJR Am J Roentgenol

Diagnostic accuracy of C-arm CT during selective transcatheter angiography for hepatocellular carcinoma: comparison with intravenous contrast-enhanced, biphasic, dynamic MDCT

Eur Radiol

Cone-beam CT with flat-panel-detector digital angiography system: early experience in abdominal interventional procedures

Cardiovasc Intervent Radiol

Contrast-enhanced abdominal angiographic CT for intra-abdominal tumor embolization: a new tool for vessel and soft tissue visualization

Cardiovasc Intervent Radiol