Vector-Field dynamic X-ray (VF-DXR) using Optical Flow Method

Abstract

Objectives:

To explore the feasibility of Vector-Field DXR (VF-DXR) using optical flow method (OFM).

Methods:

Five healthy volunteers and five COPD patients were studied. DXR was performed in the standing position using a prototype X-ray system (Konica Minolta Inc., Tokyo, Japan). During the examination, participants took several tidal breaths and one forced breath. DXR image file was converted to the videos with different frames per second (fps): 15 fps, 7.5 fps, five fps, three fps, and 1.5 fps. Pixel-value gradient was calculated by the serial change of pixel value, which was subsequently converted mathematically to motion vector using OFM. Color-coding map and vector projection into horizontal and vertical components were also tested.

Results:

Dynamic motion of lung and thorax was clearly visualized using VF-DXR with an optimal frame rate of 5 fps. Color-coding map and vector projection into horizontal and vertical components were also presented. VF-DXR technique was also applied in COPD patients.

Conclusion:

The feasibility of VF-DXR was demonstrated with small number of healthy subjects and COPD patients.

Advances in knowledge:

A new Vector-Field Dynamic X-ray (VF-DXR) technique is feasible for dynamic visualization of lung, diaphragms, thoracic cage, and cardiac contour.

Introduction

Spontaneous respiration is associated with cooperative motion of respiratory muscles and bones consisting of thorax. ¹ Diaphragm is one of the major inspiratory muscles separating thorax from abdomen. In the inspiratory phase, the diaphragm descends by contraction accompanying with the decrease in intrathoracic pressure. Rib cage also acts on its expansion: upper and lateral deviation of ribs occurs as muscles contract. These processes help passive air inflow from the outside of body via airway and lead to expansion of alveoli and lungs. In the expiratory phase, the diaphragm and rib cage act together; air outflow is urged by elastic contraction of lung. ^1,2 Thus, motion of thorax and lung works cooperatively and harmoniously.

Recently developed dynamic X-ray (DXR) with flat-panel detector provides sequential images of the thorax with high temporal resolution. ³ The procedure for DXR examination is simple, and the radiation exposure of DXR is within tolerant range from 0.4 to 1.0 mGy, only one and half times higher than that with conventional chest radiographs. ^3,4 Dynamic motion analysis of diaphragm or rib has been reported in the previous studies. ^4,5 Optical flow method (OFM) is defined as a visual pattern of the apparent motion of moving object. The motion was estimated by spatiotemporal differentiation of the image pixel-value gradient. ⁶

We hypothesize that the combination of high-temporal-resolution DXR and OFM enables the visualization of dynamic motion of lungs. Here, we report the feasibility of assessing dynamic motion of the lung and thorax using motion Vector-Field DXR (VF-DXR) with optical flow method (OFM).

Methods and materials

Patients imaging and its equipment

This study was approved by the institutional review board. Five healthy participants and five COPD patients were included in this exploratory feasibility study. During the examination with approximately 10 to 15 s, every participant took several resting breaths and one following forced breath in standing position.

DXR was performed in the standing position using a prototype X-ray system (Konica Minolta Inc., Tokyo, Japan) composed of a flat-panel detector (Pax Scan 4030CB, Varian Medical Systems Inc., Salt Lake City, UT, USA) and a pulsed X-ray generator (DHF-155HII with Cineradiography option, Hitachi Medical Corporation, Tokyo, Japan). Each DXR image was converted to the video with different frames per second (fps): 15 fps, 7.5 fps, five fps, three fps, and 1.5 fps. The image and video data obtained by DXR were assessed using software OpenCV. ⁷

Display methods for VF-DXR

Total variation regularization and robust L1 norm (TV-L1) optical flow estimation based on the interval change of background pixel-value was adopted for motion analysis. ⁸ The interval change of each pixel was calculated using two successive DXR images; it was converted mathematically to the motion vector. The theory was based on several preconditions compared with subsequent frame: unchanged luminance distribution, image differentiability, minute motion, and similar motion with adjacent pixels. Lung field obtained by DXR was divided into two-centimeter-length pixels. The motion vector from the center of each pixel was calculated and multiplied by the constant according to the frame rate. The motion vector field was also projected to horizontal and vertical components.

Color-coding map and visual assessment

In each frame, lung motion field was expressed with color-coding map as well as vector field: vector was replaced with some color presenting the orientation and magnitude was expressed as brightness. Two types of color schemes were adopted: the same as the previous study, ⁸ and above-mentioned color scheme with 135 degree counter-clockwise rotation. The latter represented the vertical motion in color-coding map with hue spectrum between red and green. Forced visual order ranking of image quality was performed among dynamic images with various frame rates ranging from 1.5 fps to 15 fps by five observers (three radiologists and two engineers).

Results

VF-DXR

Vector-field of dynamic motion of the lung was expressed successfully, which was superimposed upon the dynamic images. Dynamic and cooperative motions of ribs, and diaphragms were also well visualized using VF-DXR during inspiratory and expiratory phases (Figure 1). Simultaneously, pulmonary vasculature pulled together with expiration was observed. Rhythmical motions of heart boarders and thoracic aorta were also observed. Bone Suppression (BS) algorithm was effective for clearer demonstration of lung with vector-field map (Figure 1).

Figure 1.

Vector-Field Dynamic X-ray (VF-DXR) in inspiratory and expiratory phases with and without Bone Suppression (BS) algorithm. 41-year-old female with no smoking history. VF-DXR image (a), (c) with and (b), (d) without bone suppression (BS) algorithm. Motion vectors in (a), (b) deep inspiratory phase and (c), (d) deep expiratory phase were projected.

Frame rate for VF-DXR

Images with different frame rates per second ranging from 1.5 fps to 15 fps were shown in Figure 2. The lung motion vectors were better seen with relatively lower frame rates compared to those of heart boarder and aorta. The results of forced visual ranking were shown in Supplementary Material 1. The frame rate of 5 fps was the optimal.

Figure 2.

VF-DXR with different frame rates. 41-year-old female with no smoking history. Vector field images in inspiratory phase with (a) 15 fps, (b) 7.5fps, (c) five fps, (d) three fps, and (e) 1.5 fps and those in expiratory phase with (f) 15 fps, (g) 7.5fps, (h) five fps, (i) three fps, and (j) 1.5 fps.

Expression of vector field using color-coding map

Figure 3 demonstrates the color-coding expression of vector fields in healthy subjects. The difference of color represented the motion of the orientation of vector; and the brightness of color represented the magnitude of vector. The motion of diaphragm and adjacent lung was well visualized with 135 degree rotated color-coding map.

Figure 3.

Expression of vector field using color-coding map. 41-year-old female with no smoking history. Images with (a), (c) inspiratory and (b), (d) expiratory phase. Different color represents orientation of vector, and color strength represented the magnitude of vector. (c) and (d) adopted 135 degree counter-clockwise-rotated color scheme.

Projection of vectors into horizontal and vertical components

Vector-field projections into horizontal and vertical components in healthy subjects were shown in Supplementary Material 1. Horizontal vectors showed the horizontal movements of lung and thoracic cage better, and vertical vectors visualized lung, diaphragm, and ribs better.

VF-DXR application to COPD subjects

VF-DXR was successfully applied in COPD subjects (Figure 4). Both vector projection and color-coding map were feasible in COPD patients. The results of forced visual ranking were shown in Supplementary Material 1. The frame rate of 5 fps was the optimal.

Figure 4.

VF-DXR in COPD patient. 57-year-old female with COPD. (a) shows two-dimension vector field with five fps. Each vector is projected to (b) horizontal and (c) vertical component. Color-coding maps in deep (d) inspiratory and (e) expiratory phase with 135-degree clockwise-rotated color scheme are also shown.

Discussion

This study focused on the feasibility of VF-DXR with OFM. Dynamic and smooth motion of the thorax was optimally visualized and feasible using VF-DXR with five fps in both healthy and COPD subjects. Color-coding map and vertical and horizontal component vector-field map were feasible as additional approaches for visualizing dynamic motion.

In forced breathing, females have larger motion of upper thorax and smaller motion of abdomen than males. ⁹ Thoracic motion is also affected by aging and posture. ^9,10 Chest wall diseases including deformation of rib cage or spine as well as respiratory muscle disorders may change the dynamics of thorax. ^11,12 Lung motion can be influenced by these factors: change of lung compliance, adhesion of lung sometimes due to operation and lung or pleural disease, and regional lung disease. ^13–15

OFM has been applied for the recognition of moving objects or the quantification of movement in various situations. ^16–18 Lin et al have reported that hemodynamics in left ventricle was well visualized with ECD-gated cardiac CT using OFM. ¹⁹ Ijichi et al showed the potential combination of Gaussian filtering and optical flow for tumor tracking by X-ray. ²⁰ OFM were considered efficient to evaluate lung motion because DXR images provide transient pixel-value gradient during respiration. Tanaka et al ^3,21 have reported that the serial changes of lung or its pixel-value was observed by DXR, who referred to the possibility to describe lung local vector. ³ However, further studies have not been performed previously.

Previous studies have reported human lung motion imaging obtained by 4DCT or MRI. Lung motion field with vector was provided using point data of lung surface, or target lesion such as pulmonary vasculature or tumor. ^22–25 However, several issues remained unsolved. Subjects were examined in spine position; physiological movement of lung in standing position is known to be different with gravitational effect. Yamada et al have reported that bilateral CT-measured lung volumes in standing position in particular respiratory phase was larger than those in supine phase, ²⁶ which means that lung motion may also be affected by position. In addition, obtained images were vulnerable to artefact generated by body motion. As for X-ray, motion of human lung tumor was analyzed in several studies. ^20,27 Whole lung motion of mice has been assessed with dynamic phase-contrast X-ray using synchrotron. ²⁸

Compared with CT or MRI, VF-DXR is considered to have several advantages such as simpler method, standing position, lower radiation exposure, and shorter examination time. Physiological tidal or forced breathing during DXR examination in standing position also requires less burden. VF-DXR has the potential to easily visualize the whole lung motion in standing position, which is normal conditions for physiological respiration of human beings. VF-DXR may suggest a possibility to make a quantitative evaluation of not only lung and rib, but also diaphragm and heart motion by analyzing the vector. Video with 5fps was considered optimal due to less effect of body motion as well as not adequate size of vectors.

However, we have to admit that this study relied on the subjective visual assessment in a small number of healthy subjects and patients with COPD, which can be the limitation of this study. Further studies are required to make an objective evaluation to VF-DXR.

In conclusion, VF-DXR is feasible with an optimal frame rate of 5 fps for dynamic visualization of lung, diaphragms, thoracic cage, and cardiac contour with small number of healthy subjects and COPD patients. Color-coding map-expression, and horizontal and vertical component expressions of vector field were also introduced. The clinical studies with a larger cohort are necessary for further understanding of this new technique its clinical applications.