Ultrasound Obstet Gynecol 2005; 25: 586–591 Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/uog.1907
A new three-dimensional automatic bodymarker system for transvaginal ultrasonography T. KUWATA*, N. TANIGUCHI†, S. MATSUBARA*, T. ONO†, K. ITOH† and M. SUZUKI* Departments of *Obstetrics and Gynecology, and †Clinical Laboratory Medicine, Jichi Medical School, Tochigi, Japan
K E Y W O R D S: automatic; bodymarker; electromagnetic; three-dimensional; transvaginal
ABSTRACT Objective To evaluate the accuracy and usefulness of a newly developed three-dimensional automatic bodymarker system for transvaginal ultrasonography. Methods A bodymarker system which utilizes an electromagnetic field to specify the probe position was developed. Software was written which enabled the positional information of the probe and the ultrasound image to be simultaneously displayed on a personal computer. The bodymarker was displayed as a three-dimensional graphic model. The differences between the estimated and the actual position – i.e. the error – for both position (x, y and z) and angles (azimuth, elevation and roll) of the probe were measured. The movement of the probe was then evaluated in healthy female volunteers and the comparative time required for ultrasound examination was determined. Results Both the position and the angles of the probe were accurately shown in the computer display. The errors were 1.0 mm and 0.7◦ (median) for position and angle, respectively. The system was also shown to work well in healthy female volunteers. Calibration of the new system required only 5.0 seconds, compared with an average of 7.6 seconds for the conventional bodymarker. Conclusion The newly developed three-dimensional automatic transvaginal bodymarker system provides an accurate indication of probe position and its tilting angle. It works well in healthy female volunteers and speeds up the examination time. It may be clinically useful for transvaginal ultrasound examination. Copyright 2005 ISUOG. Published by John Wiley & Sons, Ltd.
INTRODUCTION In obstetrics and gynecology, ultrasonography has the advantage over other image-analyzing devices that it
enables different organs to be visualized in different planes1,2 . While image data of an ovarian tumor, for example, can usually be obtained in the horizontal and sagittal sections by X-ray computed tomography (CT) and magnetic resonance imaging (MRI), respectively, the features of the tumor viewed obliquely or in a tilted plane can be easily obtained by ultrasonography. Freedom of selection of the cross-section is one of the great merits of ultrasonography3,4 . However, in comparison with X-ray CT or MRI, determining the orientation of an ultrasound image can be difficult5 , especially when the image is viewed later by clinicians other than the original examiner. In extreme cases it is not possible to confirm whether the solid mass in the ultrasound image represents the uterus or an ovarian mass, or even the liver. To overcome this difficulty in determining what is being viewed in the recorded ultrasound image, most ultrasound machines have a bodymarker or transducer (probe) indication system. It indicates the position of the probe in the ultrasound display, usually by using a bar, and labels can be used to mark the organs in the ultrasound images. This conventional bodymarker with or without labeling can be inconvenient and time consuming. We developed a new bodymarker system for transabdominal ultrasonography, namely an ‘automatic virtual transducer location system’6 , to improve on the conventional bodymarker system. This bodymarker system, according to the movement of the transabdominal ultrasound probe, automatically shows both the position and tilting angle of the probe. It utilizes a pulsed DC electromagnetic field in order to specify the positional information of the probe in a three-dimensional (3D) manner and it is shown within a corresponding 3D body frame model. In the present study we applied this automatic probe location system to transvaginal sonography. We set out to address whether the newly introduced automatic probe location system works well and represents the actual movement of the transvaginal probe, whether
Correspondence to: Dr S. Matsubara, Department of Obstetrics and Gynecology, Jichi Medical School, 3311-1 Minamikawachi-machi, Kawachi-gun, Tochigi 329-0498, Japan (e-mail:
[email protected]) Accepted: 25 February 2005
Copyright 2005 ISUOG. Published by John Wiley & Sons, Ltd.
ORIGINAL PAPER
New bodymarker for transvaginal ultrasonography it reduces the time needed for the ultrasound examination compared to the conventional bodymarker system, and whether this system is useful in routine practice.
METHODS Definition of terms The bodymarker is indicated in the lower left corner of the ultrasound image (Figure 1). It consists of a body-frame and a transducer-mark or a probe-mark. For transvaginal ultrasonography the positional information of the probe or the view plane is shown in relation to the position of the uterus (uterus-mark), which represents the body-frame in Figure 1. The direction of the tilting angle of the probe was expressed in three ways: azimuth, elevation and roll. As indicated in Figure 2, azimuth indicates the right-toleft direction in the plane horizontal to the examination table; elevation means the ventral-to-dorsal direction in the sagittal plane vertical to the azimuth; and roll indicates the rotation of the vaginal probe itself.
Fundamental image of the system We used fundamentally the same bodymarker system that we recently developed as an ‘automatic transducer location system for abdominal ultrasound imaging’, which
Figure 1 The display on the personal computer. The bodymarker appears in the left lower corner and consists of the uterus-mark (body-frame) and the probe-mark (transducer-mark).
z Probe Roll
587
has previously been described in detail6 . In brief, this system has four characteristics: (1) the electromagnetic field is used to locate/specify the position of the probe; (2) not only the positions in the x, y and z planes but also the tilt of the probe are shown automatically according to the actual position/tilt; (3) the positional information of the probe is shown on the 3D body-frame as a probe-mark; and (4) both bodymarker (body-frame and probe-mark) and actual ultrasound image are shown simultaneously on a personal computer (PC) display.
Technical details We used a standard commercial ultrasound system (Model SSD-5500, Aloka Co. Ltd., Tokyo, Japan) with a 5.0MHz electronic convex transducer (vaginal probe). As an electromagnetic system we employed a miniBIRD 800/E3 system (Ascension Technology Corporation, Burlington, USA), which consisted of a transmitter sized 9.6 × 9.6 × 9.6 cm and a receiver sized 8 × 8 × 18 mm. The transmitter emitted pulsed DC electromagnetic waves that were detected by the receiver. The electromagnetic strength around the gynecological examination table was measured as 34 milligauss, which is less than 1/30 of the recommended public exposure limit (1000 milligauss) for the human body7 . A healthy female volunteer was placed in the lithotomy position on the gynecological examination table. The electromagnetic transmitter (emitter) was placed on the floor, approximately 60 cm below the patient’s perineum. The receiver for the electromagnetic field was attached to the middle of the shaft of the transvaginal probe (Figure 3). The distance between the tip of the probe and the receiver was 19 cm, and therefore the receiver itself did not enter the patient’s vagina. As described previously6 , our present system can specify the 3D positional information of the receiver within the electromagnetic field created. The receiver specifies its position, and not the position of the tip of the probe, within the electromagnetic field. Taking the distance between the probe-tip and the receiver into account, we modified the computer software so that the position of the probe-tip, and not that of the receiver, was shown in the bodymarker. The ultrasound images were captured using a video digitized card (Canopus DVStorm-RT, Canopus Co., Kobe, Japan). Both digitized images and positional information were fed into an IBM PC (CPU:600-MHz
y x
Elevation (xz)
Azimuth (xy)
Figure 2 The position and angle of the probe. Azimuth means the direction in the horizontal (xy) plane, and elevation means the ventral to dorsal direction of the patient in the sagittal (xz) plane. Roll means the rotation of the probe itself.
Copyright 2005 ISUOG. Published by John Wiley & Sons, Ltd.
Figure 3 The transvaginal probe with the electromagnetic receiver. The receiver (arrow) is attached to the shaft of the probe.
Ultrasound Obstet Gynecol 2005; 25: 586–591.
Kuwata et al.
588
Pentium III, RAM:128MB) which ran the software (written by ourselves) for the virtual probe locating system. Therefore, on the PC display, we could simultaneously see the ultrasound image (e.g. ovarian tumor) and the positional information of the probe in relation to the uterus (uterus-mark) via the probe-mark. Figure 4 provides an overview of the system.
The computer software system and system calibration The computer software was able to locate the position of the probe within a woman’s pelvic cavity in three ways: the location of the probe’s tip in the x, y and z planes; the tilting angle of the probe in both the coronal (xy) and sagittal (xz) planes (azimuth and elevation, respectively); and the probe’s rotation angle (roll) (Figure 2). This system, therefore, gives us three pieces of positional information regarding the probe: the position of the probe’s tip in the pelvic cavity; whether the probe, and thus the ultrasound beam, is located to the right or to the left side (azimuth), or in the ventral or the dorsal direction of the patient (elevation); and whether the ultrasound beam cuts the sagittal or coronal plane (roll). In order to reproduce these three pieces of information in the bodymarker system on the PC display, two-point calibrations were necessary. These calibrations were necessary only once at the beginning of the examination for each examinee. The first calibration was made when the probe shaft was held horizontal and the probe’s tip was placed just on the vaginal ostium of the examinee. The second calibration was done when the probe’s tip was inserted into the vagina, and the uterine cavity was discerned as clearly as possible, which is usually done at the beginning of the routine vaginal ultrasound examination. The best view of the uterine cavity is usually obtained when the probe’s tip is in contact with the anterior or posterior uterine wall in case of an anteverted or retroverted uterus, respectively. In the second calibration, therefore, the probe’s tip may be placed at the anterior vaginal
Ultrasound processing Video signal PC display
Signal processing
Bodymarker regulation
Superimposition
Computer
Figure 4 Overview of the system. The ultrasound image and positional information are fed into a personal computer. Bodymarker indication is shown on the computer display.
Copyright 2005 ISUOG. Published by John Wiley & Sons, Ltd.
fornix or posterior vaginal fornix according to anteversion/retroversion of the uterine body. Since uteri vary in terms of their anteversion/retroversion location and deviation to the right/left, the probe has to be moved to obtain the optimum image of the uterine cavity. In the second calibration the ultrasound beam was directed only sagittally without rotation (roll = 0◦ ). In other words, even if the uterine body was rotated, we did not take this rotation into account in the second calibration procedure. For the second calibration, we entered the following three pieces of information into the computer: (1) approximate position of the vaginal fornix; (2) uterine deviation to the right/left (azimuth); and (3) the uterine deviation in the ventral/dorsal direction (anteversion/retroversion; elevation). The uterus-mark reproduced the actual right or left deviation of the uterus (Figure 5). Furthermore, the system showed whether the probe (ultrasound beam) was directed anteriorly (ventrally) or posteriorly (dorsally) to the uterus. In fact, when the beam passes anterior relative to the uterus, the probe-mark reflects this and the ultrasound beam can be seen ahead of the uterus-mark (Figure 5a). When the beam passes posterior to the uterus, it can be seen behind the uterus-mark (Figure 5b).
Examination of the accuracy of the bodymarker system We evaluated the accuracy of the bodymarker system in three ways. The static system error of the electromagnetic system was specified as 1.8 mm and 0.5◦ for the distance and angle, respectively8 . First, we checked whether or not the probe-mark accurately indicated the actual position of the probe within the x, y and z axes. We constructed a box which contained 21 cross points, with each point 3 cm apart (Figure 6). This box was placed on the point of the gynecological examination table where the patient’s perineum would be placed. The electromagnetic transmitter was placed on the floor below the table. The PC indicated the positional information of the receiver as a numerical variable (for example, x:100 mm; y:120 mm; and z: 80 mm). The receiver was placed in the first cross point and then was arbitrarily moved to the second point. We measured the distances which the transmitter actually moved from the first to the second point (actual distances) and compared them with those indicated on the PC (indicated distances); the measurement was made at 30 different positions. Second, we checked whether or not the system accurately indicated the probe’s direction in terms of azimuth, elevation and roll. The probe, with the electromagnetic receiver attached, was directed to 0◦ , 30◦ , 45◦ , 90◦ to the right and the same to the left in the horizontal plane (azimuth). The same was done to the ventral and then to the dorsal directions in the sagittal plane (elevation). The angle was measured using a protractor and the indicated angles were compared with the actual ones. In order to evaluate the accuracy of the roll indication, we attached a semi-circular object to the shaft of the probe, which enabled us to read the actual rotation angles that the probe made (Figure 7). The
Ultrasound Obstet Gynecol 2005; 25: 586–591.
New bodymarker for transvaginal ultrasonography
589
Figure 6 Evaluation of the accuracy of the probe position (x, y and z axes). The figure shows the box with 21 cross points which was used.
Figure 7 Evaluation of the accuracy of ‘roll’. A semicircular object was attached to the probe and used to measure the roll.
Figure 5 Bodymarker indication in the new system. The cutting angle is shown clearly. (a) The beam goes to the right and ventral side of the uterus, which deviates to the right. (b) The beam goes to the left and dorsal side of the uterus, which deviates to the left.
indicated data (degrees) and the actual ones thus obtained were compared. We measured 30 different directions each for azimuth, elevation and roll. This examination was also performed with the equipment on the gynecological examination table to simulate the setting of a transvaginal ultrasound examination. Third, we analyzed whether the uterus-mark accurately indicated the actual probe position/direction in three healthy female volunteers with assumed normal pelvic anatomy. Informed consent was obtained from each of them. After making a two-point calibration, the examiner (T. K.) moved the probe as indicated below and the observer (N. T.) checked the probe-mark indications and
Copyright 2005 ISUOG. Published by John Wiley & Sons, Ltd.
wrote them down. The observer did not look at the actual probe movement, and therefore this was a ‘blind’ examination. The examiner moved the probe in various directions in 15-degree increments. For example, the examiner directed the probe 15◦ to the right (azimuth), and 30◦ to the ventral (elevation), with the probe being rotated 15◦ clockwise (roll). The observer read the probe indication. If the directions were reproduced accurately in all three components, we gave it three points, a full mark. If the observer could not distinguish between 15◦ and 30◦ , but could say that the probe was directed to either 15◦ or 30◦ , then we gave it two points. If the observer wrote down the wrong direction in one component or two components, we gave it one or 0 points, respectively. Ten different positions were examined per woman, and we studied three women. Thus, we measured a total of 30 points.
Measuring the time needed to calibrate the system The time needed to complete the calibration was measured in 20 patients who visited our outpatient clinic for the first time for various reasons and received transvaginal ultrasound examination as a routine procedure. Informed consent for the participation in the study was obtained
Ultrasound Obstet Gynecol 2005; 25: 586–591.
Kuwata et al.
590
from all patients. The examiners (S. M. and M. S.) made the first and second calibrations, and the observer (T. K.) measured the time taken from the first to the second calibration using a stopwatch. The time needed for making a bar in the conventional two-dimensional (2D) bodymarker system was also measured. In addition, we measured how much time was needed for the entire ultrasound examination, and how many ultrasound images were recorded per patient (thus, how many conventional bodymarkings were made). The data are presented as mean ± SD and were analyzed statistically using the unpaired t-test. P < 0.05 was considered to be significant.
RESULTS Table 1 shows the difference in the distances that the tip of the probe actually moved and those detected by the receiver and thus indicated in the PC. The median difference in distance, thus the error, was 1.0 (range, 0–7.5) mm. Table 2 shows the errors in the angles for azimuth, elevation and roll, respectively. The average error in these three angles was 0.7 (range, 0.1–3.5)◦ . The errors specified by the manufacturer for the distance and the angle were 1.8 mm and 0.5◦ , respectively8 , and the measured data cited above approximately coincided with them. In three healthy female volunteers 30 points were examined to evaluate the accuracy of the probe indication using a scoring system. The average score was 2.8, with 3.0 being the maximum score possible. Three points were obtained in 25 tests, and two points were obtained in the remaining five tests. There were no cases in which the observer judged the score as one or zero points. Two-point calibration required 5.0 ± 1.4 (mean ± SD) seconds, which was significantly shorter than 7.6 ± 1.3 (mean ± SD) seconds, the time needed to make only one bar in the conventional 2D bodymarker system (P < 0.001; Table 3). The total time needed for the ultrasound examination was 132 seconds, during which an average of 2.2 ultrasound images were recorded and thus bar-makings were done 2.2 times per patient. Therefore, the total time needed for bar-making per patient Table 1 Errors in distances in x, y and z axes Axis x y z Average
Median (range) mm
was 16.7 seconds (7.6 × 2.2 = 16.7), which was 12.7% (16.7/132) of the total time of the ultrasound examination. If image annotations were added in the conventional system, the time needed would have been considerably longer, although that was not measured.
DISCUSSION The new 3D bodymarker system enabled us to simultaneously view a transvaginally obtained ultrasound image as well as three-dimensional positional information, tilt angle and roll of the probe relative to the uterus on a PC display. It is sometimes very difficult to interpret ultrasound images when they are looked at long after the examination. In order to specify the positional information of the probe, or to specify what organ is being viewed, we have used a conventional 2D bodymarker or we have directly labeled organs on the ultrasound image. For example, if we were to find an ovarian mass in the right-posterior side of the uterus, in an anteverted position with the uterine body tilting to the left, in order to record this we must: (1) document that the uterus is anteverted and deviates to the left; (2) freeze the ultrasound image; (3) select the 2D body-frame; (4) set the appropriate bar to indicate the probe position or the plane being viewed; and (5) mark ‘ov’ for the corresponding ovary (Figure 8). Although this has long been done in ultrasound examination practice in Japan, there are, we believe, four fundamental ways in which the conventional 2D bodymarker system is inferior to the 3D system we have developed. First, the tilting angle or the rotation status of the probe cannot be shown in the conventional system. Second, the calibration procedure of the conventional bodymarker can be slow, especially to make the bar indicative of the probe-cutting plane. Making a twopoint calibration with the 3D system is simple. In both the routine transvaginal ultrasound examination and the calibration we first place the probe’s tip on the introitus of the examinee (the first calibration), insert the probe into the vagina, and then move the probe to obtain the best Table 3 Average time required for making the bodymarker (BM) in the conventional system and that needed for making the calibration in the new system, and average number of images recorded in each case
1.0 (0–3.0) 1.0 (0–2.9) 2.0 (0–7.5) 1.0 (0–7.5)
Table 2 Errors in the angles for azimuth, elevation and roll Parameter
Median (range) ◦
Azimuth Elevation Roll Average
0.5 (0.1–1.6) 1.2 (0.1–3.0) 0.7 (0.1–3.5) 0.7 (0.1–3.5)
Time needed (s) or number; mean ± SD Time needed for total ultrasound examination Time needed for making one BM (A) Number of ultrasound images recorded per patient (B) Total time spent for making BM in conventional system (A × B) Time needed for making calibration in the new system
Range
132 ± 66
38–269
7.6 ± 1.3* 2.2 ± 0.9
5.0–10.1 1–4
16.8
—
5.0 ± 1.4*
3.0–7.7
*P < 0.001, unpaired t-test.
Copyright 2005 ISUOG. Published by John Wiley & Sons, Ltd.
Ultrasound Obstet Gynecol 2005; 25: 586–591.
New bodymarker for transvaginal ultrasonography
Figure 8 The conventional two-dimensional bodymarker. The assumed cutting plane is indicated by a bar. The organ being viewed, an assumed ovarian tumor, is marked by the abbreviation ‘ov’.
view of the uterine cavity (the second calibration). The time saved will become more obvious when several images are to be taken for one patient. Third, inserting labels on the images is time-consuming and they may bias or mislead subsequent clinicians who review the image later. With the new system this potential bias is avoided. If, for example, there is a solid mass in the ultrasound image and we know that the probe was directed to the right lower side of the left-deviated uterus, this mass could be either a right ovarian tumor or a subserosal pedunculated myoma nodule or even a retroperitoneal tumor. Deviation of the uterus cannot be shown in the conventional bodymarker. Finally, the conventional bodymarker can only be used for still images whereas the new system is applicable to video-recordings of ultrasound images. The probe-mark moves quickly and automatically according to the actual probe movement. We are now trying to expand this system in two ways. First, we are trying to make the uterus-mark represent the actual uterine size. In the present system, the uterusmark shows the uterus as the same size, irrespective of the actual uterine size. Therefore, we cannot distinguish between a large uterus and a small one via the uterusmark. Theoretically, it is possible to approximately match the uterus-mark to the actual uterine size. Second, we are now trying to feed the bodymarker information into the ultrasound computer rather than into a separate PC.
Copyright 2005 ISUOG. Published by John Wiley & Sons, Ltd.
591
This will enable us to simultaneously look at both the ultrasound images and bodymarker in the ultrasound display, not in the PC display. An automatic indication system for the transvaginal probe was previously reported by Shinozuka et al.9 . They employed a piezoelectric vibratory gyroscope to obtain the positional information of the probe. Although in their preliminary study the position of the probe was automatically shown in the x, y and z planes, no bodymark (uterus-mark) was shown in it. Our present system automatically shows the position of the probe in relation to the position of the uterus. This may be one advantage over the previous system reported by Shinozuka et al.9 , unless of course the uterus is absent. However, we have not yet demonstrated the clinical usefulness of this new system in a large population which contains women with gynecological disease and atypical anatomy. These questions are now under investigation in our laboratory.
ACKNOWLEDGMENT This research was supported by Grant-in-Aid for Scientific Research by the Japan Society for the Promotion of Science (Nos. 13877402 and 15500346).
REFERENCES 1. Sivyer P. Pelvic ultrasound in women. World J Surg 2000; 24: 188–197. 2. Guariglia L, Rosati P. Transvaginal sonographic detection of embryonic-fetal abnormalities in early pregnancy. Obstet Gynecol 2000; 96: 328–332. 3. Freimanis AK. Ultrasonic imaging of neoplasms. Cancer 1976; 37: 496–502. 4. Taylor KJ, McCready VR. A clinical evaluation of grey-scale ultrasonography. Br J Radiol 1976; 49: 244–252. 5. Carter BL, Kahn PC, Wolpert SM, Hammerschlag SB, Schwartz AM, Scott RM. Unusual pelvic masses: a comparison of computed tomographic scanning and ultrasonography. Radiology 1976; 121: 383–390. 6. Taniguchi N, Kuwata T, Ono T, Itoh K, Omoto K, Fujii Y, Ootake A. Automatic virtual transducer location system to assist in interpreting ultrasound imaging. J Med Ultrasonics 2003; 30: 211–216. 7. ICNIRP, EMF guidelines. Health Physics 1998; 74: 494–522. 8. MiniBird product specification sheet [package insert]. Ascension Technology Corp: Burlington, 2001. 9. Shinozuka N, Oie Y, Yamakoshi Y, Taketani Y. Transvaginal sonographic orientation detection system using ceramic gyroscopes. J Ultrasound Med 1996; 15: 107–113.
Ultrasound Obstet Gynecol 2005; 25: 586–591.
