Optimization of a chest computed tomography protocol for detecting pure ground glass opacity nodules: A feasibility study with a computer-assisted detection system and a lung cancer screening phantom.

Seongmin Kang,Tae Hoon Kim,Kyunghwa Han,Chul Hwan Park,Ji Young Kim,Baeggi Min,Jae Min Shin,Vadim E Degtyar

doi:10.1371/journal.pone.0232688

Seongmin Kang, Tae Hoon Kim + Show 6 more

Open Access

https://doi.org/10.1371/journal.pone.0232688

Copy DOI

Abstract

ObjectiveThis study aimed to optimize computed tomography (CT) parameters for detecting ground glass opacity nodules (GGNs) using a computer-assisted detection (CAD) system and a lung cancer screening phantom.MethodsA lung cancer screening phantom containing 15 artificial GGNs (−630 Hounsfield unit [HU], 2–10 mm) in the left lung was examined with a CT scanner. Three tube voltages of 80, 100, and 120 kVp were used in combination with five tube currents of 25, 50, 100, 200, and 400 mA; additionally, three slice thicknesses of 0.625, 1.25, and 2.5 mm and four reconstruction algorithms of adaptive statistical iterative reconstruction (ASIR-V) of 30, 60, and 90% were used. For each protocol, accuracy of the CAD system was evaluated for nine target GGNs of 6, 8, or 10 mm in size. The cut-off size was set to 5 mm to minimize false positives.ResultsAmong the 180 combinations of tube voltage, tube current, slice thickness, and reconstruction algorithms, combination of 80 kVp, 200 mA, and 1.25-mm slice thickness with an ASIR-V of 90% had the best performance in the detection of GGNs with six true positives and no false positives. Other combinations had fewer than five true positives. In particular, any combinations with a 0.625-mm slice thickness had 0 true positive and at least one false positive result.ConclusionLow-voltage chest CT with a thin slice thickness and a high iterative reconstruction algorithm improve the detection rate of GGNs with a CAD system in a phantom model, and may have potential in lung cancer screening.

Highlights

The term ‘subsolid nodules (SSNs)’ include both pure ground-glass nodules (GGNs) and partsolid nodules [1]
Among the 180 combinations of tube voltage, tube current, slice thickness, and reconstruction algorithms, combination of 80 kVp, 200 mA, and 1.25-mm slice thickness with an ASIRV of 90% had the best performance in the detection of GGNs with six true positives and no false positives
Baeggi Min is an employee of GE healthcare, Korea, and received support in the form of salary from GE healthcare

Summary

Introduction

The term ‘subsolid nodules (SSNs)’ include both pure ground-glass nodules (GGNs) and partsolid nodules [1]. Up to 19% of the pulmonary nodules detected on baseline studies are SSNs [3,4]. 37–70% of the SSNs are transient according to previous studies [3,5,6], persistent SSNs identified on computed tomography (CT) have a higher possibility of malignancy than solid nodules. Persistent SSNs detected on screening have a malignancy rate of up to 34%; the previously reported malignancy rate of GGNs and part-solid nodules was 18 and 63%, respectively [4,7]. Visualization of SSNs can be missed on CT screens, and according to Li et al, 91% of missed lung cancer lesions on CT were SSNs [8,9]

Methods

Results

Conclusion