Detection of potentially clinically relevant lung nodules with breath-hold CT compared to free-breathing CT in PET-CT in oncology patients and the value of MIP reconstructions

Background: The value of an additional thoracic breath-hold (BH) CT in PET-CT for detecting small nodules is controversial. Purpose: To determine the value of a BH-CT compared to the routinely used free-breathing (FB) CT, and the value of maximum intensity projection (MIP) technique in these two CT series for detection and its potential clinical relevance in staging of small lung nodules in oncology patients. Material and methods: This retrospective study included 200 patients from February to September 2016 who were referred for oncological FDG PET-CT. Thin slice (1.25 mm) and MIP (10 mm) images were analyzed for detection of small lung nodules (1-10 mm). Binominal test and descriptive statistics were analyzed. Results: It was possible to evaluate 186 patients, with a mean age of 70.0 yrs ± 11.1 yrs, range 27-96 yrs, consisting of 84 females (mean age 68.5 yrs ± 12.2, 27-90 yrs) and 102 men (71.3 yrs ± 10.1, 27-96 yrs). In FB-CT, 393 nodules were detected in thin slice CT images, with MIP in 578. In BH-CT, 534 nodules were detected and 905 with MIP (p< 0,001). The extra detected nodules were considered as having a potential clinical relevance in 4.8% (9 of 186) of the patients. The total radiation dose from the routine PET-CT, including FB-CT, was 26 mSv and an additional 4 mSv from the BH-CT. Conclusion: An additional BH-CT in deep inspiration compared with FB-CT detects more small lung nodules, which potentially could alter the TNM stage.


Introduction
With the introduction of a national Swedish healthcare program in 2015 called "Standardized Care Pathway" (SCP), the clinical work up for cancer changed [1]. Previously, a full radiation dose contrast enhanced Computed Tomography (CECT) often existed for staging and treatment planning. The value of Positron Emission Tomography (PET) CT was primarily to detect metabolic active lesions, increasing the likelihood of malignant tumor or spread. In practice, SCP means that the patient should be examined within one week after referral.
It is quite common that a PET-CT is performed directly after detection of a suspected lung lesion found on a plain chest X-ray. PET-CT then becomes the baseline examination if the suspected lesion proves to be a cancer.
Due to the lack of a previous diagnostic CT in these patients, the protocol for the CT performed in conjunction with PET has Citation: Montserrat  been changed in most PET facilities in Sweden. A full radiation dose CECT is performed albeit in Free-Breathing (FB) due to the need for morphological correlation to the PET data. Images of the lungs in FB often contain moving artifacts, which could hide small nodules crucial for cancer staging and management.
The follow-up of cancer patients is most frequently done with CT. Later, when therapy response is assessed with the CT from the previous PET-CT, comparison could be problematic especially for the basal parts of the lungs. Changes in size and number of small nodules are diffi cult to assess. Therefore, some institutions add an additional thoracic Breath-Hold (BH) CT in deep inspiration to the routine PET-CT examination. This practice has created controversy based on concerns regarding the increased radiation dose. The clinical value of adding an additional thoracic BH-CT has not been proven and is also controversial in literature [2,3].
Another strategy for detection of nodules could be the addition of 10 mm Maximum Intensity Projection (MIP) reconstructions to the thin slice CT images of the lungs, based on the knowledge that MIP increases the detection of small nodules [4].
The aim of this study was to determine the value of an additional BH-CT in deep inspiration compared to FB-CT in PET-CT and the value of MIP reconstructions in these two CT series for the detection of potentially clinically relevant small lung nodules in oncology patients.

Patient population
This retrospective study included 200 oncology patients from February to September 2016 who were referred for fl uorodeoxyglucose (FDG) PET-CT. The most common indication was characterization of a newly detected lung lesion (Table 1).

Image acquisition
The examinations were performed in a Discovery VCT PET-CT scanner (GE Healthcare, Waukesha, WI, U.S.A.). The system was equipped with a 64 slice CT detector. The patients rested while 4 MBq/kg dose of FDG was injected intravenously. After 60 minutes, the patients were placed on the tabletop with arms over the head. After an initial CT scout image, a low radiation dose CT (100 kVp, 40-110 mAs, noise index 20, 3.75 mm slice collimation) images in free-breathing from the base of the skull to the mid-thigh were acquired and later used for the attenuation correction (AC) of PET data. These images were not transferred to the Picture Archive and Communication System (PACS) due to their low image quality. Thereafter, the PET emission scans were collected for three minutes for each bed position; thus, 6 to 8 bed positions were acquired depending on the patient's height. Iodine contrast agent was injected intravenously, and a breath-hold thoracic CT during deep inspiration was performed after 35 sec (120 kV, 70-780 mAs, noise index 48, 0.625 mm slice collimation). After 70 seconds from the start of the injection, the routine shallow freebreathing CT in venous phase was scanned (120 kV, 70-780 mAs, noise index 48, 0.625 mm slice collimation). The FB-CT images were used for the PET-CT fusion images. The wholebody CT scan with free-breathing was positioned as the range for PET acquisition from the base of the skull to the mid-thigh and included the neck, thorax, and abdomen. The CT scan with deep inspiration only covered the thorax.
CT images of the lung parenchyma were reconstructed with a sharp lung algorithm, with 1.25 mm slice thickness and no overlap. The CT images were analyzed in a PACS workstation (Carestream Health, Rochester, NY, U.S.A.). MIPs were reconstructed with 10 mm slice thickness.
A "small lung nodule" was defi ned as a non-calcifi ed rounded opacity, smaller than 10 mm surrounded by lung parenchyma. A "potentially clinically relevant nodule" was defi ned as a nodule that could potentially infl uence the patient's management by altering the stage based on the TNM classifi cation system.

Image analysis:
Two radiologists with 15 years (MA), respectively, 12 years (TH) of experience in oncology staging analyzed the 4 sets of CT images for 100 patients each. The series were reviewed in the following order to minimize any bias from pre-existing knowledge of the presence of a nodule: First, the FB scans, then the BH series. Afterwards, the MIP reconstructions of the FB series and fi nally, the MIP of the BH series were reviewed. The original report for the PET-CT examination was blinded for the radiologists. The number of small lung nodules and the location of the nodules considered as potentially clinically relevant was registered. Consensus readings were performed only for the potentially clinically relevant nodules.
Data were analyzed with institutional radiation review board approval and a waiver of the requirement for informed consent.

Statistical analysis
Binominal test for proportions were used for test of signifi cance of the number of nodules detected on BH MIP versus the other series. Signifi cance level were defi ned as p< 0.05. Descriptive statistics were calculated for the sample. Mean, ± standard deviation, and range are given. Graphpad (San Diego, CA, USA) were used for statistical analysis.
Small lung nodules were detected in 64.5% (120 of 186) of the patients that were possible to evaluate. With the additional BH-CT, it was possible to detect more small nodules than FB-CT in 42 patients. This fi gure increased to 51 patients when the MIP reconstructions of the two series of images were analyzed.
The number of the detected nodules for the thin slice FB-CT and the BH-CT series with 1.25 mm slice thickness and MIP reconstructions are given in Table 2     This fi nding supports a previous report which stated that BH-CT detected more nodules than FB-CT [2]. In contrast, in a study reporting on the use of thin 2 mm slices instead of thicker 3-5 mm slices for the FB-CT, the 2 mm slices were found suffi cient to detect nodules, and an additional BH-CT was considered to be unnecessary [3]. In the present study, thin 1.25 mm slices were used on both FB-CT and BH-CT, and our results support the fi nding that BH-CT detects more nodules than FB-CT. A previous report also stated that more nodules are detected with the use of MIP reconstructions as compared with thin slice images, and the result from the present study supports that statement [5].
This study reports on the ability of the radiologist to detect nodules. Perception is multi-factorial, including components such as image quality, experience, and reader fatigue [6].
Even though a nodule could be clearly depicted on an image, it could still be overlooked by the radiologist and be undetected.
There are high expectations regarding the concept of artifi cial intelligence to support the radiologist in this detection process [7]. The fi ndings in this study support the theory that better image quality, i.e., images of the lungs in deep inspiration and breath-hold for reducing motion artifacts and post processing such as MIP reconstructions, helps the radiologist to detect more nodules [5].
The concept of "potentially clinically relevant nodule" was used in the present study because merely detecting a nodule  8 in table  3) referred for PET-CT for characterization of a 11 mm nodule situated in the right upper lobe, which was proven to be an adenocarcinoma on histological specimen from a trans thoracic biopsy afterwards. An additional 11 mm nodule representing a metastasis in the same lobe and a 13 mm metastatic lymph node in the right hilum were reported in the PET-CT report (these fi ndings are not shown in the images). All these three fi ndings were FDG avid. An additional 3 mm nodule indicated by a grey circle in the right lower lobe was not detected on the axial free-breathing The potential benefi t of detecting more nodules is associated with a risk for the patient, in terms of increased radiation dose. If the patient would be under-staged due to nodules not perceived, the likelihood of a more aggressive treatment plan with potential hazards for the patient increases. The use of health care resources would also increase, possibly without benefi t for the patient.
The BH-CT added 4 mSv to the patient's total radiation dose of 26 mSv. In literature, PET dose ranges from 3.5-10.5 mSv, WB CT 11-20 mSv, and CT for AC 3-6 mSv [9]. As technical development occurs, the CT incorporated in the PET-CT could be able to achieve similar image quality for both the FB-CT and the BH-CT series with a reduced radiation dose [10]. In this context and in oncology patients, it seems to possible to justify the additional 4 mSv in most cases.

Conclusion
The addition of BH-CT in deep inspiration and in conjunction with MIP reconstructions of FB-CT and BH-CT can detect more potentially clinically relevant nodules in oncology patients referred for PET-CT.