Using 3D polylines to improve cone-beam CT-guided percutaneous transthoracic needle biopsy

Abstract

Objectives:

To investigate the diagnostic accuracy of using 3D polylines (3DPs) to improve cone-beam CT (CBCT) virtual navigation (VN)-guided percutaneous transthoracic needle biopsies (PTNB) of pulmonary lesions.

Methods:

From May 2021 to November 2021, patients (81 males and 41 females; age, 65 ± 12 years) who underwent CBCT VN with 3DPs for PTNB of pulmonary lesions were retrospectively reviewed. Fluoroscopic visibility of target lesions was evaluated using captured images from a Bull’s eye view. Diagnostic accuracy was calculated, and complications were assessed.

Results:

The mean size of biopsied lesions was 23 ± 13 mm (range: 6–75 mm). Overall, 13.9% (17/122) were small pulmonary nodules (diameter ≤1 cm), and 68.0% (83/122) of biopsied lesions were fluoroscopic visible. The overall diagnostic accuracy was 94.3%. The diagnostic accuracy for visible and invisible lesions was 94.0 and 94.9%, respectively (p = 0.843), and 100% for small pulmonary nodules. Major complications occurred in 8.2% (10/122; eight pneumothorax with chest tube insertion, one hemoptysis with transfusion, and one air embolism) of patients.

Conclusion:

CBCT VN with 3DP guidance provide a real-time outline of pulmonary lesions, thus enabling a reliable and accurate PTNB.

Advances in knowledge:

3DP guidance could be useful technique for CBCT-guided PTNB, especially in small pulmonary nodules.

Introduction

Cone-beam CT (CBCT) percutaneous transthoracic needle biopsy (PTNB) has been reported as an accurate method for diagnosing pulmonary lesions. ^1–3 CBCT offers flexibility in orienting the C-arm gantry and flat-panel detector around the patient compared to conventional CT guided PTNB. It provides multiplanar reconstructed CT images with real-time fluoroscopy. Another advantage of CBCT PTNB is that it allows for virtual navigation (VN), which offers automatic angulation of the X-ray tube. ^4–7 A puncture needle can easily be inserted through an overlaid virtual pathway on the screen; and as a consequence, high accuracy has been reported in diagnosing pulmonary lesions, even for small pulmonary nodules. ⁸ However, VN guidance is reconstructed based on static CBCT images. The location of a lesion may shift during respiration; thus, registration between VN guidance and the real target may be hampered. In these cases, the target nodule can be tracked with real-time fluoroscopy through respiratory cycles instead of with VN guidance.

However, for small pulmonary nodules that are subcentimeter-sized, visual tracking of nodules with real-time fluoroscopy can be difficult, as they are often invisible due to lung parenchymal abnormalities or their low density. A previous study reported a 76.7% diagnostic accuracy for CBCT VN-guided PTNB of small invisible nodules, which was significantly lower than for visible nodules (89.1%). ⁸ Even with visible nodules, taking advantage of real-time tracking during a biopsy is challenging because of radiation exposure to the operator’s hand. To avoid radiation exposure, vendors created a “progression view” projection of the C-arm that is perpendicular to the initial C-arm position. However, while minimizing radiation exposure, visibility is often hampered in the progression view due to overlapping structures, such as the heart, diaphragm, or vertebrae.

With the recent development of the CBCT system and 3D reconstruction software, polygonal diagrams can be drawn in 3D reconstructed space. Polylines can be drawn along the outline of a pulmonary lesion in multiplanar planes, thus depicting the location of a pulmonary nodule in any projection, regardless of its visibility. With the addition of 3D polylines (3DPs), the limitation of decreased visibility in the progression view may be overcome. In this study, we overlaid 3DPs of pulmonary lesions to aid in targeting and performed PTNB with real-time tracking of the biopsy needle under lateral or oblique projection. The purpose of our study was to investigate the diagnostic accuracy of CBCT VN with 3DPs for PTNB of pulmonary lesions.

Methods and materials

Patients

This retrospective study was approved by the Institutional Review Board of Jeonbuk National University Hospital, and written informed consent was waived. From May 2021 to November 2021, 131 consecutive patients were initially included according to the following inclusion criteria: (i) patients who underwent CBCT VN with 3DPs-guided PTNB and (ii) patients with complete electronic medical records of procedure and pathology. Among initially included patients, nine were excluded according to the following exclusion criteria: (i) cases of repeated PTNB (n = 7) and (ii) mediastinal PTNB (n = 2). Finally, 122 patients (81 male, 41 female) aged 69 ± 12 years (range: 23–87 years) were included in this study.

Registration of pulmonary lesions

All biopsies were performed using a CBCT system (Artis Q, Siemens Healthcare, Erlangen, Germany) with the aid of a dedicated virtual guidance software program (Syngo i-Guide, Best, Netherlands). Pre-procedural planning CBCT was performed at full expiration. If an appropriate needle pathway could not be obtained during expiration, planning CBCT was conducted again at the full inspiration state.

After reviewing pre-procedural CBCT, the operator determined the virtual needle pathway to maximize the diagnostic yield and to minimize the risk of complications. After establishing the virtual needle pathway, the multiplanar orientation automatically adjusts to the axis of the virtual needle pathway. Then, the operator determined representative slices by choosing the slice with the largest area from the stack of multiplanar planes of the pulmonary lesion. 3D polylines were manually drawn by creating multiple straight lines along the marginal area of the pulmonary lesion on the representative slices, to be overlaid on fluoroscopy. A “Bull’s-eye” projection, which is an automatic vertical alignment from the skin entry site to the target lesion, was used (Figure 1). Briefly, the C-arm automatically rotates in the direction of the virtual needle pathway. Then, a laser navigation system on the flat panel detector allows the center of the detector to align to the virtual needle pathway. A virtual color spot on the fluoroscopic image and crisscross laser navigation presents a skin entry site. The puncture was performed using a 19-gauge coaxial needle (TSK Starcut Biopsy Needle, TSK Laboratory, Japan). The needle was inserted under VN with an overlaid 3DP.

Figure 1.

Examples of CBCT PTNB with 3D Polylines. The nodule was invisible on fluoroscopy due to an obscuring pulmonary artery.

(a) Based on the preprocedural CBCT, 3D Polylines were drawn along the margin of the nodule. Then, a virtual needle pathway was determined by the operator. (b) Overlaid 3D Polylines provided the additional guidance of 3D Polylines with an aid of a Bull’s eye view.

Outlining with 3dp and PTNB

After inserting the core needle, a control CBCT was performed to verify the location of the needle. The sagittal and axial planes of the control CBCT were adjusted according to the needle pathway. After determining the representative slice on the sagittal and axial plane, 3DPs were drawn along the sagittal with or without an axial margin of the lesion. Also, a straight line was drawn along the needle pathway. These 3DPs were overlaid on the fluoroscopy. After a review of the volume rendering reconstruction model, the operator determined an optimal angle of the C-arm for biopsy. If the needle tip was located at the desired point, a lateral projection was generally used, so that the whole length of the needle and the sagittal plane of the lesion were demonstrated (Figure 2). If the needle tip was mislocated, an oblique view using an LAO or RAO projection with cranial or caudal angulation was used. In this view, the axial and the sagittal margin are simultaneously demonstrated obliquely, enabling manual tilting of the needle toward the lesion. The biopsy was performed with a 20G needle. Between 1 and 3 tissue samples were obtained. Collimation was usually used to minimize radiation exposure to the operator’s hand. Postprocedural CBCT was acquired to identify procedure-related complications.

Figure 2.

Examples of 3D Polyline-guided biopsy.

(a) A control CBCT shows a core needle and the nodule. 3D polylines were drawn along the needle pathway as well as the sagittal margin of the nodule. (b) Fluoroscopy shows the needle overlapping the needle pathway drawn by 3D polylines, which demonstrates the relative location of the needle to be the same as in the control CBCT. (c) Biopsy needle is slightly tilted toward the caudal direction to obtain proper direction to the nodule.

Data collection

Data were obtained from the electronic medical records. Patient data (age and sex), biopsy-related data (size, location, pleura-to-target distance, patient’s position, total fluoroscopy time, and visibility of nodules on the bull’s eye view), and data on complication (pneumothorax and hemoptysis) were acquired. The fluoroscopic visibility of the pulmonary lesion in the bull’s eye view was retrospectively evaluated. Two radiologists who were blinded to the biopsy results independently assessed the visibility. Only lesions that could be possibly identified by both readers were classified as visible lesions.

For the evaluation of diagnostic accuracy, pathological reports were reviewed. The lesions were classified into malignant, benign, or indeterminate. Malignant and benign from specific pathology (e.g., tuberculosis, hamartoma, or cryptococcosis) either with the PTNB or surgery were considered as the final diagnoses. Lesions with non-specific benign pathology were considered benign if the size of the lesion was decreased or remained stable for more than 6 months. Having a correct diagnosis was defined as either a malignant biopsy result followed by a final diagnosis of malignancy (true-positive) or a benign biopsy result followed by a final diagnosis of benign (true-negative). Having an incorrect diagnosis was defined as a benign pathology followed by a diagnosis of malignancy (false-negative) or a non-diagnostic specimen on the pathologic report.

Radiation dose calculation

Anthropomorphic phantom (Rando phantom, model RAN-110, Churchin Asssociates) and radiophotoluminescent glass dosimeters (GD-302M, Asahi Techno Glass Corporation) were used for the calculation of effective doses. Twenty glass dosimeters were placed in organs according to the recommendation in International Commission on Radiological Protection publication 103. ⁹ CBCT was performed three times on the phantom to calculate the average dose. To measure the absorbed radiation doses, the dosimeters were transferred to an automatic readout system (FGD-1000, Asahi Techno Glass). The effective dose of the phantom study was calculated according to weighting factors, which is described in publication 103. ⁹ The total dose area product (DAP) of the phantom study was recorded. A conversion factor between the effective dose and the DAP of the phantom study was calculated. Using the conversion factor, the effective dose for each patient was calculated in this study.

Statistical analysis

The PTNB results were categorized into two groups: the diagnostic success group (which included true-positives and true-negatives) and the diagnostic failure group (which included false-negatives). The diagnostic accuracy of the PTNB was calculated for the overall population and for fluoroscopically visible and invisible lesions. Also, the diagnostic accuracy according to lesion size was calculated. The Chi-square test was performed to compare the diagnostic accuracy of visible and invisible lesions. A p value < 0.5 was considered statistically significant. All statistical analyses were performed using MedCalc Statistical Software v 18.9 (MedCalc Software bvba, Ostend, Belgium).

Results

PTNB was successfully performed in all cases. The mean size of biopsied pulmonary lesions was 23 ± 13 mm (range: 6–75 mm). Of those, 13.9% (17/122) were unidimensional diameter ≤1 cm. The mean distance to the target lesion from the pleura was 20 ± 18 mm (range: 0–90 mm). The position was prone in 58 patients and supine in 64 patients. Overall, 68.0% (83/122) of biopsied lesions were fluoroscopically visible. The mean fluoroscopy time was 2.8 ± 2.5 min. The mean procedure time was 17.7 ± 5.6 min.

Based on the pathologic reports, there were 85 (69.7%) malignant nodules, 36 (29.5%) benign nodules, and 1 (0.8%) indeterminate nodule. Among 85 malignant nodules, 81 were confirmed as primary lung cancers and four were metastases. Among 36 benign nodules, 13 were confirmed as specific benign pathology (tuberculosis 8, hamartoma 1, cryptococcosis 1, chondroma 1, chondrosarcoma 1, pneumoconiosis 1). In the remaining 23 nodules of non-specific benign pathology at initial biopsy, repeat biopsy was performed in six nodules due to clinical suspicion of malignancy.

Diagnostic accuracy

Among 122 cases included in the diagnostic accuracy analysis, having a correct diagnosis was found in 115 cases (true positive, 85 cases; true negative, 30 cases), while having an incorrect diagnosis was found in seven cases (false negative, six cases; non-diagnostic specimen, one case). Overall sensitivity and specificity for the diagnosis of malignancy were 92.4% (85 of 92) and 100% (30 of 30), respectively. The diagnostic accuracy for the overall population was 94.3% (95% confidence interval, 88.5–97.6%). For the invisible lesions on the fluoroscopy, the diagnostic accuracy was 94.9% and this was not statistically different from that for the visible lesions (94.0%; p = 0.843). For 17 small (≤ 1 cm) pulmonary nodules, the diagnostic accuracy was 100% (true positive, 11 cases; true negative, six cases) (Table 1).

Table 1.

Diagnostic accuracy according to lesion size

Lesion size	No. of patients	Correct diagnosis	Incorrect diagnosis	Accuracy (%)
<=1 cm	17	17	0	100.0
1–2 cm	43	40	3	93.0
2–5 cm	56	54	2	96.4
>5 cm	6	4	2	-

Complications

A pneumothorax occurred in 33 of 122 (27.0%) cases. Chest tubes were inserted in eight cases (6.6%). Hemoptysis occurred in six (4.9%) cases. One patient (0.8%) required a red blood cell transfusion. Other complications resolved spontaneously with conservative management. Cerebral and cardiac air embolism occurred in one case. Intensive care unit admission was required. In sum, major complications occurred in 10 of 122 (8.2%) cases.

Radiation dose

The average effective dose and DAP for one CBCT in the phantom study were 0.44 mSv and 895.2 uGy.m^2, respectively. Therefore, the calculated conversion factor was 491.5 mSv/Gy.m^2. The average effective dose of CBCT PTNB was 1.68 mSv (range, 0.29–5.42 mSv) for 122 patients. The average effective dose of real-time fluoroscopy was 0.27 mSV (range, 0.03–1.62 mSv).

Discussion

Our results showed that CBCT VN with 3DPs for PTNB provides accurate needle targeting of pulmonary lesions. This technique is based on CBCT VN-guided PTNB and provides an outline of a lesion in addition to the needle pathway. It enables a reliable needle progression and biopsy and ensures a high puncture success rate. A high overall diagnostic accuracy of 94.3% was found in this study. The diagnostic accuracies for invisible lesions and small nodules were also high (94.9 and 100%, respectively). The major complication rate of 8.2% in this study was considerably low, with only 4.9% of hemoptysis.

CBCT-guided PTNB is a safe and effective method for the diagnosis of pulmonary lesions. It provides multiplanar imaging for needling and real-time feedback during the procedure. ^2,3,10 Especially, VN is a state-of-art guidance technique, enabling optimal and safe needle path selection under simple manipulation. ^7,8 However, VN does not delineate the pulmonary lesion from the background; thus, real-time monitoring of the lesion is not possible if the lesion is fluoroscopically invisible. ⁸

3DPs are useful for overcoming limitations of VN. By outlining the lesion, even the location of an invisible lesion can be determined in real-time fluoroscopy, as well as the monitoring of the needle position. Visualization of both the lesion and needle allows the operator to perform a biopsy with increased confidence, resulting in high diagnostic accuracy and improved clinical workflow. Further, since visualization of the outline of both the lesion and biopsy needle helps the operator adjust the needle position more accurately, we cautiously suggest this simple imaging technique using 3DPs will help inexperienced operators improve their diagnostic accuracy.

The workflow of adding 3DP guidance is simple. Generally, less than 1 min is needed for 3DP outlining. 3DPs are versatile and can be used in various situations. If the pulmonary lesion is located near the pleura, a drawing of the pleural lining can be used to be aware of needle retraction that could cause immediate pneumothorax, which would result in PTNB failure. If mispuncture occurs, manual needle tilting during the biopsy can be performed with confidence under the overlaid outline of the lesion. Also, vital internal organs could be marked with 3DPs, to avoid mispuncture. The high accuracy and low complication rate in this study might be attributed to the use of CBCT VN in conjunction with 3DP. Although the rate of pneumothorax in this study was higher than previous large case series [27.0% vs 17.0% ³ ], it is comparable or lower than recent smaller case series [27.0% vs 29.1%, ⁸ 38.8% ¹ ]. Given that only early experience was included in this study, pneumothorax rate could possibly be reduced with accumulation of experience and by adopting antipneumothorax techniques.

Although CBCT VN offers advantages in targeting the pulmonary lesion, small pulmonary lesions can easily be misdiagnosed by PTNB. Previous studies using CBCT VN-guided PTNB showed relatively low diagnostic accuracy in small lesions. Hwang et al reported diagnostic accuracy of 86.0% for subcentimeter nodules. In their study, the diagnostic accuracy was significantly lower in invisible nodules than visible nodules (76.7% vs 89.1%, respectively; p = 0.042). ⁸ Jiao et al reported 94.1% of diagnostic accuracy for lesions ≤ 2 centimeters. ¹¹ In this study, 100% of diagnostic accuracy was reached for all subcentimeter nodules. In addition, there was no significant difference in the diagnostic accuracy between the visible and the invisible lesions (94.0 and 94.9%, respectively; p = 0.843). Among seven cases of incorrect diagnosis, only two cases were related to mistargeting. Other cases were related to central necrosis in two cases and obscured margin of the lesion due to adjacent consolidation.

There are several limitations to this technique. First, the number of included patients was small. Second, the diagnostic accuracies in CBCT VN-guided PTNB and CBCT VN and 3DP-guided PTNB were not directly compared. Further prospective studies or randomized controlled trials are necessary to compare these modalities. Third, lateral projection during biopsy may increase radiation exposure to the operator. Proper radiation protection methods should be performed to avoid unnecessary radiation hazards.

In conclusion, CBCT VN and 3DP guidance provide a real-time outline of the pulmonary lesion, thus enabling a reliable and accurate PTNB. This modified technique has excellent versatility in PTNB procedures for pulmonary lesions, resulting in high diagnostic accuracy.