Application of Classification Trees for the Qualitative Differentiation of Focal Liver Lesions Suspicious for Metastasis in Gadolinium-EOB-DTPA-Enhanced Liver MR Imaging

Abstract

Purpose: To evaluate the diagnostic accuracy of qualitative descriptors alone and in combination for the classification of focal liver lesions (FLLs) suspicious for metastasis in gadolinium-EOB-DTPA-enhanced liver MR imaging.

Materials and Methods: Consecutive patients with clinically suspected liver metastases were eligible for this retrospective investigation. 50 patients met the inclusion criteria. All underwent Gd-EOB-DTPA-enhanced liver MRI (T2w, chemical shift T1w, dynamic T1w). Primary liver malignancies or treated lesions were excluded. All investigations were read by two blinded observers (O1, O2). Both independently identified the presence of lesions and evaluated predefined qualitative lesion descriptors (signal intensities, enhancement pattern and morphology). A reference standard was determined under consideration of all clinical and follow-up information. Statistical analysis besides contingency tables (chi square, kappa statistics) included descriptor combinations using classification trees (CHAID methodology) as well as ROC analysis.

Results: In 38 patients, 120 FLLs (52 benign, 68 malignant) were present. 115 (48 benign, 67 malignant) were identified by the observers. The enhancement pattern, relative SI upon T2w and late enhanced T1w images contributed significantly to the differentiation of FLLs. The overall classification accuracy was 91.3 % (O1) and 88.7 % (O2), kappa = 0.902.

Conclusion: The combination of qualitative lesion descriptors proposed in this work revealed high diagnostic accuracy and interobserver agreement in the differentiation of focal liver lesions suspicious for metastases using Gd-EOB-DTPA-enhanced liver MRI.

Zusammenfassung

Ziel: Evaluation der diagnostischen Genauigkeit qualitativer Diagnosekriterien einzeln und in Kombination für die Klassifikation metastasenverdächtiger fokaler Leberläsionen (FLL) in der Gadolinium-EOB-DTPA-angehobenen Leber MRT.

Material und Methoden: Konsekutive Patienten mit klinischem Verdacht auf Lebermetastasen kamen für diese retrospektive Untersuchung infrage. 50 Patienten erfüllten die Einschlusskriterien. Alle unterzogen sich einer Gd-EOB-DTPA angehobenen Leber-MRT (T2w, chemical shift T1w, dynamische T1w). Primäre maligne Neoplasien der Leber sowie anbehandelte Läsionen wurden ausgeschlossen. Alle Untersuchungen wurden durch 2 geblindete Untersucher (U1, U2) evaluiert. Diese bewerteten unabhängig voneinander das Vorhandensein von Läsionen sowie definierte qualitative Deskriptoren (Signalintensitäten, Anreicherungsmuster und Morphologie). Der Referenzstandard wurde unter Berücksichtigung aller klinischen Informationen gebildet. Die statistische Analyse beinhaltete neben Kontingenztafeln (Chi-Quadrat, kappa) die Kombination von Einzelkriterien mittels Klassifizierungsbäumen (CHAID-Methode) sowie eine ROC-Analyse.

Ergebnisse: In 38 Patienten fanden sich 120 FLL (52 benigne, 68 maligne), 115 (48 benigne, 67 maligne) wurden durch die Untersucher identifiziert. Anreicherungsmuster, relative SI in der T2w und Spätphasen-T1w-Bilder trugen signifikant zur Differenzierung von FLL bei. Die Gesamtgenauigkeit betrug 91,3 % (U1) und 88,7 % (U2), kappa = 0,902.

Schlussfolgerung: Die in dieser Arbeit dargelegte Kombination von qualitativen Läsionsdeskriptoren demonstrierte hohe Treffsicherheit und Interobserver-Übereinstimmung in der Differenzierung von metastasensuspekten FLL in der Gd-EOB-DTPA-angehobenen Leber-MRT.

Key words

abdomen - MR diffusion/perfusion - MR imaging - decision analysis - metastases - statistics

• References

• 1 Mulier S, Ni Y, Jamart J et al. Radiofrequency ablation versus resection for resectable colorectal liver metastases: time for a randomized trial?. Ann Surg Oncol 2008; 15: 144-157
• 2 Tsoulfas G, Pramateftakis MG, Kanellos I. Surgical treatment of hepatic metastases from colorectal cancer. World J Gastrointest Oncol 2011; 3: 1-9
• 3 Jeffery M, Hickey BE, Hider PN. Follow-up strategies for patients treated for non-metastatic colorectal cancer. Cochrane Database Syst Rev 2007; CD002200
• 4 Hammerstingl R, Huppertz A, Breuer J et al. Diagnostic efficacy of gadoxetic acid (Primovist)-enhanced MRI and spiral CT for a therapeutic strategy: comparison with intraoperative and histopathologic findings in focal liver lesions. Eur Radiol 2008; 18: 457-467
• 5 Huppertz A, Haraida S, Kraus A et al. Enhancement of focal liver lesions at gadoxetic acid-enhanced MR imaging: correlation with histopathologic findings and spiral CT--initial observations. Radiology 2005; 234: 468-478
• 6 Zech CJ, Grazioli L, Breuer J et al. Diagnostic performance and description of morphological features of focal nodular hyperplasia in Gd-EOB-DTPA-enhanced liver magnetic resonance imaging: results of a multicenter trial. Invest Radiol 2008; 43: 504-511
• 7 Danet IM, Semelka RC, Leonardou P et al. Spectrum of MRI appearances of untreated metastases of the liver. Am J Roentgenol 2003; 181: 809-817
• 8 Lowenthal D, Zeile M, Lim WY et al. Detection and characterisation of focal liver lesions in colorectal carcinoma patients: comparison of diffusion-weighted and Gd-EOB-DTPA enhanced MR imaging. Eur Radiol 2011; 21: 832-840
• 9 Seale MK, Catalano OA, Saini S et al. Hepatobiliary-specific MR contrast agents: role in imaging the liver and biliary tree. Radiographics 2009; 29: 1725-1748
• 10 Frericks BB, Loddenkemper C, Huppertz A et al. Qualitative and quantitative evaluation of hepatocellular carcinoma and cirrhotic liver enhancement using Gd-EOB-DTPA. Am J Roentgenol Am J Roentgenol 2009; 193: 1053-1060
• 11 Guiu B, Cercueil JP. Liver diffusion-weighted MR imaging: the tower of Babel?. Eur Radiol 2010; 21: 463-467
• 12 Santoro L, Grazioli L, Filippone A et al. Resovist enhanced MR imaging of the liver: does quantitative assessment help in focal lesion classification and characterization?. J Magn Reson Imaging 2009; 30: 1012-1020
• 13 Bossuyt PM, Reitsma JB, Bruns DE et al. Towards complete and accurate reporting of studies of diagnostic accuracy: The STARD Initiative. Radiology 2003; 226: 24-28
• 14 Landis JR, Koch GG. An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers. Biometrics 1977; 33: 363-374
• 15 Deeks JJ, Altman DG. Diagnostic tests 4: likelihood ratios. Bmj 2004; 329: 168-169
• 16 Kass GV. An Exploratory Technique for Investigating Large Quantities of Categorical Data. Applied Statistics 1980; 29: 119-127
• 17 Liao SC, Lee IN. Appropriate medical data categorization for data mining classification techniques. Medical informatics and the Internet in medicine 2002; 27: 59-67
• 18 Bland M. An introduction to medical statistics. Oxford, UK: Oxford University Press; 2000
• 19 Grossi E, Buscema M. Introduction to artificial neural networks. European journal of gastroenterology & hepatology 2007; 19: 1046-1054
• 20 Muller R, Mockel M. Logistic regression and CART in the analysis of multimarker studies. Clinica chimica acta; international journal of clinical chemistry 2008; 394: 1-6
• 21 Bleck JS, Ranft U, Gebel M et al. Random field models in the textural analysis of ultrasonic images of the liver. IEEE Trans Med Imaging 1996; 15: 796-801
• 22 Sugimoto K, Shiraishi J, Moriyasu F et al. Computer-aided diagnosis of focal liver lesions by use of physicians' subjective classification of echogenic patterns in baseline and contrast-enhanced ultrasonography. Acad Radiol 2009; 16: 401-411
• 23 Obuchowski NA, Mazzone PJ, Dachman AH. Bias, underestimation of risk, and loss of statistical power in patient-level analyses of lesion detection. Eur Radiol 2010; 20: 584-594
• 24 Hardie AD, Naik M, Hecht EM et al. Diagnosis of liver metastases: value of diffusion-weighted MRI compared with gadolinium-enhanced MRI. Eur Radiol 2010; 20: 1431-1441