Automatic scan range for dose-reduced multiphase CT imaging of the liver utilizing CNNs and Gaussian models

Multiphase CT scanning of the liver is performed for several clinical applications; however, radiation exposure from CT scanning poses a nontrivial cancer risk to the patients. The radiation dose may be reduced by determining the scan range of the subsequent scans by the location of the target of interest in the first scan phase. The purpose of this study is to present and assess an automatic method for determining the scan range for multiphase CT scans. Our strategy is to first apply a CNN-based method for detecting the liver in 2D slices, and to use a liver range search algorithm for detecting the liver range in the scout volume. The target liver scan range for subsequent scans can be obtained by adding safety margins achieved from Gaussian liver motion models to the scan range determined from the scout. Experiments were performed on 657 multiphase CT volumes obtained from multiple hospitals. The experiment shows that the proposed liver detection method can detect the liver in 223 out of a total of 224 3D volumes on average within one second, with mean intersection of union, wall distance and centroid distance of 85.5%, 5.7 mm and 9.7 mm, respectively. In addition, the performance of the proposed liver detection method is comparable to the best of the state-of-the-art 3D liver detectors in the liver detection accuracy while it requires less processing time. Furthermore, we apply the liver scan range generation method on the liver CT images acquired from radiofrequency ablation and Y-90 transarterial radioembolization (selective internal radiation therapy) interventions of 46 patients from two hospitals. The result shows that the automatic scan range generation can significantly reduce the effective radiation dose by an average of 14.5% (2.56 mSv) compared to manual performance by the radiographer from Y-90 transarterial radioembolization, while no statistically significant difference in performance was found with the CT images from intra RFA intervention (p = 0.81). Finally, three radiologists assess both the original and the range-reduced images for evaluating the effect of the range reduction method on their clinical decisions. We conclude that the automatic liver scan range generation method is able to reduce excess radiation compared to the manual performance with a high accuracy and without penalizing the clinical decision.

[Comparison of radiation exposure in common hepatic interventions : A retrospective analysis of DeGIR registry data]

Hintergrund

Die transarterielle Chemoembolisation (TACE) oder auch Gallenganginterventionen stellen häufige Leberinterventionen dar.

Ziel der Arbeit

In dieser retrospektiven Studie soll die Strahlenexposition der Patienten mit einem hepatischen Eingriff in Abhängigkeit von Art und Feinziel der Intervention analysiert und verglichen werden.

Material und Methoden

Dies ist eine Analyse von 7003 DeGIR-Registerdatensätzen aus den Jahren 2016 bis 2018 für TACE und Gallenganginterventionen. Das Dosisflächenprodukt (DFP) und die Durchleuchtungszeit (DL) sowie die Interventionsart und das anatomisch definierte Feinziel der Interventionen wurden erfasst.

Ergebnisse

Insgesamt lagen Dosiswerte für 4985 durchgeführte TACE und 2018 Gallenganginterventionen vor. Bei Gallenganginterventionen lag der Median des DFP bei 2594 (Interquartilbereich [IQR] = 1174–5858) cGy*cm². Bei der TACE betrug der Median des DFP 11.632 [IQR = 5530–22.800] cGy*cm² und lag damit signifikant höher als bei Gallenganginterventionen (p < 0,0001). Gallengangeingriffe mit dem höchsten DFP sind Interventionen am Ductus hepaticus, während Eingriffe mit der längsten DL an der Hepatikusgabel stattfinden.

Diskussion

Die individuelle Strahlendosis für einen Patienten bei einer Leberintervention hängt weniger von der Komplexität des Eingriffs bzw. Durchleuchtungszeit ab, sondern von der Art des Eingriffs und vom Feinziel der Intervention. Die vorliegenden Dosisdaten können eine Hilfe sein, die Strahlenexposition bei einer Leberintervention bereits vor dem Eingriff grob abzuschätzen.

Phantom Study Investigating the Accuracy of Manual and Automatic Image Fusion with the GE Logiq E9: Implications for use in Percutaneous Liver Interventions

Purpose

To determine the accuracy of automatic and manual co-registration methods for image fusion of three-dimensional computed tomography (CT) with real-time ultrasonography (US) for image-guided liver interventions.

Materials and Methods

CT images of a skills phantom with liver lesions were acquired and co-registered to US using GE Logiq E9 navigation software. Manual co-registration was compared to automatic and semiautomatic co-registration using an active tracker. Also, manual point registration was compared to plane registration with and without an additional translation point. Finally, comparison was made between manual and automatic selection of reference points. In each experiment, accuracy of the co-registration method was determined by measurement of the residual displacement in phantom lesions by two independent observers.

Results

Mean displacements for a superficial and deep liver lesion were comparable after manual and semiautomatic co-registration: 2.4 and 2.0 mm versus 2.0 and 2.5 mm, respectively. Both methods were significantly better than automatic co-registration: 5.9 and 5.2 mm residual displacement (p < 0.001; p < 0.01). The accuracy of manual point registration was higher than that of plane registration, the latter being heavily dependent on accurate matching of axial CT and US images by the operator. Automatic reference point selection resulted in significantly lower registration accuracy compared to manual point selection despite lower root-mean-square deviation (RMSD) values.

Conclusion

The accuracy of manual and semiautomatic co-registration is better than that of automatic co-registration. For manual co-registration using a plane, choosing the correct plane orientation is an essential first step in the registration process. Automatic reference point selection based on RMSD values is error-prone.