• Nu S-Au Găsit Rezultate

Stitching of sensor-navigated 3D ultrasound datasets for the determination of large thyroid volumes – a phantom study

N/A
N/A
Protected

Academic year: 2022

Share "Stitching of sensor-navigated 3D ultrasound datasets for the determination of large thyroid volumes – a phantom study"

Copied!
7
0
0

Text complet

(1)

DOI:

Original papers

Stitching of sensor-navigated 3D ultrasound datasets

for the determination of large thyroid volumes – a phantom study

Martin Freesmeyer, Leonard Knichel, Christian Kühnel, Thomas Winkens

Clinic of Nuclear Medicine, Jena University Hospital, Jena, Germany

Received 20.08.2018 Accepted 28.10.2018 Med Ultrason

2018, Vol. 20, No 4, 480-486

Corresponding author: PD Dr. med. Martin Freesmeyer, M.D.

Clinic of Nuclear Medicine Jena University Hospital

Am Klinikum 1, 07747 Jena, Germany Phone: +49-3641-9329801

Fax: +49-3641-9329802

E-mail: [email protected]

Introduction

The size and shape of the thyroid gland vary greatly, depending on environmental and individual factors such as iodine availability, age, gender and smoking habits [1,2]. Determination of the thyroid volume is relevant not only for the primary diagnostics of thyroid disorders, but also for the planning of radiotherapy and monitor- ing of therapeutic effects [3,4]. Ultrasound imaging, mostly two-dimensional (2D-US), is the standard ap- proach for the determination of thyroid volume [5,6].

Important advantages of US include the anatomical reso-

lution, soft-tissue contrast, cost effectiveness, and ubiq- uitous availability. Other relevant aspects include the ab- sence of contraindications or radiation exposure issues [6,7].

Volume assessment via 2D-US is based on the gen- eral assumption that the thyroid lobes have an ideal form and that a rotation ellipsoid model (em) can thus be ap- plied [8]. However, the routine use of em can lead to substantial inter- and intra-observer variability because individual anatomical variants of the thyroid, such as a thickened isthmus or a nodular goiter, can lead to volu- metric errors of up to 30% [4,9].

A possible method to overcome these limitations is the use of three-dimensional ultrasound (3D-US), which not only allows for the generation of volume datasets in three dimensions, but also permits data storage for post hoc analyses independent of the actual scanning session [4]. Furthermore, volumetric analysis via 3D-US allows the use of manual tracing (mt), a method that shows very high levels of agreement with the actual thyroid volumes.

A drawback of 3D-US, however, is that the size of the thyroid may exceed the field of view (FOV) of the 3D Abstract

Aims: Thyroid volume has to be measured in goiters prior to radioiodine treatment to calculate the needed amount of radioactivity. Modern clinical equipment for ultrasonography shows the trend to smaller probes, so that larger goiters do not fit any longer into the probe’s field of view. This study evaluated the feasibility and accuracy of stitching procedures applied to thyroid volumetric analysis performed using three-dimensional ultrasound (3D-US). Material and methods: Ad hoc thy- roid phantoms of different shapes (regular, nodular, thickened isthmus) and volumes (ranging between 50 and 400 mL) were developed. In 15 such phantoms the left and right lobes were separately scanned, and the 3D-US datasets were then assembled (stitched) using predefined landmarks and dedicated software. Volumetric analysis was then assessed via a conventional ellip- soid model (em) and manual tracing (mt). The correlation of measured and reference volumes was determined using Pearson’s correlation coefficients and Bland and Altman limits of agreement. Results: The results showed a high level of agreement (with correlation coefficients ranging from 0.974 to 0.999) for all shapes and volumes tested, including the largest volume of 400 mL. The mt method, although more time consuming, proved superior to the em. Conclusions: Stitching-mediated thyroid volumetric analysis is accurate, and its clinical performance should be investigated in future studies.

Keywords: Stitching; Volumetric analysis; Thyroid phantom; 3D-US; Extended field of view

DOI: 10.11152/mu-1687

(2)

probe [7], making it impossible to image the whole gland with a single scan [9].

The spatial limitation of the 3D-US can be potentially resolved by post hoc assembly of two or more 3D-US datasets using the stitching method, which results in an extended volume or extended FOV (3D-US EFOV). This method, however, has not yet been applied for thyroid imaging or for organ volumetric analysis.

Therefore, the aims of the present study were 1) to investigate whether stitching of two separately scanned 3D-US partial thyroid volumes is feasible and 2) to as- sess the accuracy of the resulting volumetric 3D-US EFOV. For this purpose, ad hoc thyroid phantoms with volume and shape variants were developed based on a previously established phantom [10].

Materials and methods Phantoms

Fifteen water-filled, heart-shaped balloons were used to produce phantoms with a shape resembling the thyroid

gland in 15 predefined volumes (50, 75, 100, 125, 150, 175, 200, 225, 250, 275, 300, 325, 350, 375, and 400 mL). For the measurements, the phantoms were placed into a double-walled plastic container filled with water (fig 1). The phantoms were submersed 2 cm below the water surface to enable US scanning. To maintain the phantoms in this container, four small pieces of plastic were inserted into the balloons and then ligated with two threads per pole to the wall of the inner plastic container (fig 2). To reproduce the shapes of the lobes and the typi- cally narrow isthmus of a healthy thyroid, a clamp was applied to the phantom as shown in fig 1 and 2.

Along the isthmus, with the phantom placed in the container, six landmarks were fixed onto the anterior sur- face of the right and left lobes. The landmarks consisted of echogenic hybrid polymer glue (Fix All Classic, Soudal, Leverkusen, Germany) and served as orientation points for the subsequent stitching procedures. The simulation of the isthmus and trachea was achieved with a 2-cm di- ameter polymethylmethacrylate (Perspex®, Billingham, UK) rod fixed onto styrofoam supports at a height of 2 cm from the bottom of the container. The phantoms were then stretched along the rod and fixed with threads in that position to simulate an anthropomorphic form of the thy- roid (fig 2a).

The unladen weight (tare) of each empty phantom was measured using a precision scale. Subsequently, the balloons were filled with pre-boiled water to minimize Fig 1. Experimental set-up for the 3D ultrasound scan (3D-

US) of thyroid phantoms. (a) Representative phantom (target volume 200 mL) with hybrid polymer landmarks (blue arrows) used for subsequent orientation in stitching procedures. A Per- spex® rod (white arrows) is used as a trachea dummy and as a support for the construction. (b) The probe is manually moved in the craniocaudal direction (yellow arrow) within a magnetic field generated by a transmitter (green arrow). The position of the probe is detected using sensors (red arrows). The yellow box represents the field of view (FOV). (c, d) US scanning in a water-filled container via a convex C2-9 probe (C-probe) (c) and a linear ML-6-15 probe (L-probe) (d).

Fig 2. Size and shape variants of the thyroid phantoms. (a–c) regular-shaped phantoms, with simulation of the isthmus using a median clamp (yellow arrow); target volumes of 50 mL (a), 200 mL (b), and 400 mL (c). (d–f) Shape variants with a target volume of 200 mL: (d) regular shape; (e) nodular shape, ob- tained via binding of the lobes (blue arrows); and (f) thickened isthmus obtained via removal of the medial clamp.

(3)

the formation of air bubbles. The net volumes derived from the weight measurements represented the reference standard for the assessment of the volumetric accuracy of the different US methods. The phantoms were weighed again at the end of the experiments to exclude leakage during the experiments.

After completion of the first measurement cycle with normally shaped thyroid phantoms (fig 2d), a second cy- cle was performed with deformed lobes as a model of nodular goiter (fig 2e). In the third investigation cycle, the clamp was removed to simulate a model of a thick- ened isthmus (fig 2f).

Ultrasonography scans

All scans were performed with the same US equip- ment (Logiq E9, GE Medical Systems, Milwaukee, WI, USA) with the application of two different sensor-navi- gated 3D-US probes: the convex C2-9 probe (C-probe) and the linear ML6-15 matrix probe (L-probe) (fig 1c and 1d). The phantom was placed approximately 2 cm from a magnetic transmitter. Two positional sensors were fixed to the probe to allow recording of the position and movement of the US probe in the magnetic field during scanning. The equipment reconstructed a 3D volume

dataset based on the coordinates of the probes and the acquired 2D images. Scans of the right and left lobes of the phantoms were performed separately with inclusion of the isthmus, ensuring that each scan included the three ipsilateral and three contralateral landmarks (fig 3). Dur- ing the use of the L-probe, a virtual convex mode was used to enlarge the FOV of the primary scan. The depth, focus number, and focus position were adapted to opti- mize the quality of the image of the corresponding phan- tom volume, whereas the remaining parameters of the probe setting were kept constant.

Digital Imaging and Communications in Medicine (DICOM) export

Storage of the 3D-US datasets in the US equipment was performed with a dedicated DICOM standard for 3D-US (Enhanced Ultrasound Volume Storage; EUVS) as a stack of transverse slices. The stored datasets were finally transferred to an external storage medium.

Stitching and post-stitching data modification The 3D-US datasets of the left and right phantom lobes were subsequently loaded to the stitching mod- ule of PMOD software (PMOD, Version 3.409, PMOD Technologies Ltd., Zurich, Switzerland), displayed in a half-transparent manner and oriented to each other so that all six landmarks became superimposed in the axial and sagittal planes (fig 3c and 3d). A combined extended 3D-US dataset, now including the whole phantom with both lobes, was then generated and stored (fig 3e). To reduce the time effort of the subsequent volumetric anal- yses using mt, the extended 3D-US datasets were con- verted from a transverse slice thickness of 0.21-0.4 mm to a uniform thickness of 3.0 mm using PMOD software and stored again [10].

Volumetric analysis

Volumetric analysis for each phantom was performed by an experienced investigator. All acquired datasets were assessed by both the ellipsoid formula (lengthmax x widthmax x depthmax x 0.5) and the mt procedure. The PMOD software tool “image viewing and volume of in- terest analysis” was used to evaluate the volumes.

In the case of the em, volumetric analysis was per- formed separately for each lobe starting from the sagittal standard planes to identify the slices with the maximum length. The transverse plane was then oriented perpen- dicularly to the sagittal plane to measure the maximum width and depth. For the models with a thickened isth- mus, the isthmus volume was divided and assigned to the volume of the right and left lobe so that no separate measurement was performed.

The mt method was applied via the mt of the phan- tom contours in parallel slices of transverse image slices (fig 4a). In accordance with the em, the left and right Fig 3. Separate scanning of phantom lobes and stitching of the

partial volumes. Sequential generation of datasets for the left lobe (a) and right lobe (b) of a 200-mL phantom. An L-probe was used in this case. The two 3D-US datasets (c, d) are orient- ed according to the apposite landmarks (red arrows) using the PMOD software and stitched together into a 3D-US EFOV (e).

(4)

lobes were assessed separately. Only the model with a thickened isthmus was evaluated with a whole contour covering the right lobe, left lobe, and isthmus. Volume calculation was automatically performed by the software.

Statistical analysis

The relationship between the calculated and reference volumes of the thyroid phantoms was quantified by Pear- son’s correlation coefficient. The correspondence of the calculated and reference volumes was then quantified us- ing a slightly modified version of the Bland and Altman method [11]. The difference between measured and ref- erence volumes was calculated as the relative difference, i.e., as a percentage of the reference volume. These rela- tive differences are described by the mean, standard de- viation (SD), median, and range (minimum, maximum).

To assess the magnitude of systematic differences, the 95% confidence intervals of the mean were also calcu- lated. When the confidence interval did not include a zero value, the corresponding scanning method was consid- ered to have a significant systematic error (significance level 0.05, overestimation or underestimation of the ref- erence volume). In addition, calculated and reference volumes were compared using the limits of agreement (mean + 1.96*SD) recommended by Bland and Altman [11]. This interval includes 95% of all differences, under the assumption that the differences are normally distrib- uted. All calculations were performed with Stata/IC 13.1 for Windows (StataCorp LP, College Station, TX, USA).

Results

Phantoms of different sizes and shapes proved prac- tical and suitable for the planned investigations and all phantoms could be investigated according to the intent of the study. Stitching of the obtained datasets via PMOD also proved feasible. The phantoms, fully captured using 3D-US EFOV, could be successfully subjected to volu- metric analysis with both the em and mt methods. In two

cases, a leak in the phantoms was detected because of a difference in weight measured before and after the final scan. In both cases, the experiments were repeated.

Acquisition of a sensor-navigated 3D-US dataset lasted 3 to 5 min, irrespective of the probe used. The mt measurements required approximately 15-20 min per phantom, and the em measurements required 1-2 min.

The correlation coefficients of the calculated and ref- erence volumes as well as the mean values, SDs, 95%

confidence intervals, and limits of agreement of relative differences for all three phantom shapes (regular, nodu- lar, thickened isthmus) are presented in table I. The high correlation coefficients (range: 0.974-0.999) for all meth- ods investigated – irrespective of the phantom shape – indicated a very strong positive relationship between the calculated and reference volumes.

Although the general level of correspondence was very high, some discrepancies were found between the calculated and reference volumes (fig 5). In general, volumetric analysis via mt showed markedly smaller de- viations from the reference volumes than the em meth- od. The em was associated with clearly higher SDs and wider 95% confidence intervals. The volumetric analy- Fig 4. Volumetric analysis of a 3D-US-EFOV via manual trac-

ing. Slice-wise manual contouring in a transverse orientation of a 200-mL phantom scanned with an L-probe. Transverse (a) and coronal (b) view of the contoured volumes (right lobe in light red, left lobe in magenta).

Fig 5. Graphic display of deviations of phantom volumes from reference volumes. The horizontal lines within the boxes rep- resent the mean percent deviations; the amplitude of each box represents the 95% confidence interval (CI) of the mean dif- ference, and whiskers represent the means + 1.96-times the standard deviation (limits of agreement). If the 95% CI did not include the zero value, the corresponding scanning method was considered to have a significant systematic error (significance level p = 0.05, overestimation or underestimation of the refer- ence volume). C: convex probe; L: linear probe; em: ellipsoid model; mt: manual tracing.

(5)

ses with the C-probe were associated with a significant underestimation of values obtained with the em for the nodular shape and thickened isthmus. The mt method also showed some deviations, i.e., a mild but significant underestimation of phantoms with a regular or nodular shape. volumetric analysis using the em was clearly im- pacted by the anatomical variants of the phantom, where- as the mt method was largely unaffected by the shape of the phantoms (fig. 5).

Discussions

Volumetric analysis of the thyroid by 2D-US has been largely established in clinical routine practice [9], but 3D-US approaches are also conceivable. In 3D-US, the resulting 3D datasets can be reconstructed in the three standard dimensions and evaluated post hoc, in accord- ance with computed tomography (CT) and magnetic res- onance imaging (MRI) [12]. Indeed, the interpretation of 3D-US images may be less prone to inter-observer vari- ability, given the clearer and more objective images[13].

However, 3D-US presents some limitations [12]. The limited FOV in particular allows only imaging of rela- tively small volumes compared to CT and MRI [14]. This not only reduces the interpretation of the anatomical con- text, but also limits the measurement of large volumes, such as in cases of severely enlarged thyroids.

The problem of augmenting the anatomical view is not limited to the 3D-US but also concerns 2D-US. To overcome this limitation, different techniques have been developed, one of which is the EFOV. Using 2D-US, for example, panoramic views can be created along the verti-

cal direction of the US probe, and the image information based on given characteristic structures can be assem- bled into a unified image[7]. The images obtained can be immediately seen on the US monitor, and reconstruc- tion is not necessary [7,15]. Another approach used in 3D-US is the enlargement of the detectable anatomical context through the assembly (stitching) of two or more 3D datasets. This strategy has been investigated in dif- ferent fields (gynecology, urology, gastrointestinal, car- diology, and orthopedics) [7,12,13,15-18], but not yet in thyroid studies. Additionally, the above studies did not focus on volumetric analysis but rather on distance meas- urements or comparisons between 3D-US EFOV and CT data. In contrast, the present study focused on the feasi- bility of stitching 3D-US datasets specifically for thyroid volumetric analysis using ad hoc-developed phantoms.

Different phantoms are available for US investigations [12,14,16,17,19] and another group also recently devel- oped a flexible balloon phantom of the thyroid that al- lows for the simulation of nodular shape alterations and isthmus thickening[10]. In the present study, this original model was further adapted to increase the volume range from 200 mL to 400 mL; additionally, ventral echogenic landmarks were placed to facilitate the subsequent stitch- ing procedure, and a support rod was added with the intent of maintaining the anthropomorphic shape of the thyroid phantom and retaining a trachea dummy in place.

Orientation of 3D-US datasets before stitching can be performed automatically or semi-automatically [12,13,16,17,19-21]; however, in this study, we opted for a manual approach based on the phantom landmarks, with a time span of not more than 2 min. The results show

Table I. Correlation between measured and reference volumes and percent deviation from the reference volumes Percent deviation

Probe type Method Phantom shape Rho* Mean ± SD Median Range 95% CI Mean ± 1.96*SD (LOA)

Convex

(C-probe) em regular 0.9939 -1.1 ± 5.2 -4.0 -8.2 - 5.7 -4.0, 1.8 -11.4, 9.1

nodular 0.9934 -5.9 ± 6.9 -8.1 -12.5 - 8.4 -9.8, -2.1 -19.4, 7.5 thick isthmus 0.9944 -9.6 ± 5.6 -10.2 -16.5 - 7.5 -12.6, -6.5 -20.5, 1.3

mt regular 0.9993 -2.8 ± 2.4 -2.4 -7.7 - 1.3 -4.1, -1.4 -7.5, 1.9

nodular 0.9987 -1.9 ± 3.0 -1.1 -7.0 - 4.3 -3.6, -0.2 -7.8, 4.0 thick isthmus 0.9984 -1.3 ± 3.0 -1.3 -6.8 - 5.0 -2.9, 0.4 -7.1, 4.6 Linear

(L-probe) em regular 0.9933 2.8 ± 5.5 4.9 -6.9 - 7.8 -0.3, 5.8 -8.0, 13.5

nodular 0.9811 1.3 ± 8.8 7.5 -10.0 - 12.3 -3.6, 6.2 -16.0, 18.6 thick isthmus 0.9774 -2.6 ± 9.7 -7.7 -11.8 - 13.3 -7.9, 2.8 -21.5, 16.4

mt regular 0.9982 0.5 ± 3.1 0.6 -4.7 - 5.0 -1.2, 2.2 -5.5, 6.5

nodular 0.9991 0.3 ± 2.8 0.5 -7.0 - 3.8 -1.2, 1.8 -5.1, 5.7

thick isthmus 0.9987 1.4 ± 3.1 2.3 -6.6 - 4.8 -0.3, 3.1 -4.7, 7.4

Rho*: correlation coefficient according to Bravais-Pearson analyses; 95% CI: 95% confidence interval of the mean; LOA: Limit of agree- ment; em: ellipsoid mode; mt: manual tracing

(6)

that the manual adjustment and the stitching of separately obtained 3D-US datasets acquired via sensor-navigated probe was possible and in fact accurate.

Compared to a previous phantom study performed with volumes up to 200 mL and without the use of stitch- ing [10], the present data showed even higher levels of agreement with the reference volumes – irrespective of whether the em or the mt method was used – indicating that 3D-US EFOV may be a promising option for thyroid volumetric analysis.

The present data are consistent with previous studies showing that mt volumetric analysis is associated with far smaller deviations from the expected values than the em method [4,9,10,22,23]. Systematic underestimations and overestimations are also smaller with mt. Larger deviations of the ellipsoid-based volumes occur particu- larly in the models of nodular shape alterations and isth- mus thickening. These imprecisions can be prevented by the use of mt in 3D-US EFOV [4], given that mt allows for a much better shape-adapted contouring of the target.

Moreover, the present study also confirms that these ad- vantages are achieved at the cost of substantially greater time and effort [9,24,25].

Compared to the L-probe, in the 3D-US images, the C-probe showed contours with more artifacts, leading to broader images of the phantom walls (results not shown) due to the lower frequency and geometry of the transduc- er. In this case, the exact placement of the landmarks for the ellipsoid method and the delineation of the contours are less clear. Consistent with a previous study [10], the ellipsoid landmarks and the contours for mt were placed in the middle of the sonographically depicted wall struc- ture, rather than in the inner or outer border. Consequent- ly, all plots obtained using the C-probe showed slightly lower values than the data obtained with the L-probe. In addition, the use of the C-probe was characterized by significant underestimations of the reference volumes in three groups, while this did not occur when the L-probe was used. Although it may be expected that the artifact- prone wall representation leads to larger limits of agree- ment, no such differences were found.

Additional analyses of intra- and inter-observer varia- bility were not attempted in this study, in accordance with similar studies by other groups [9,26] and because the results obtained with the mt method were unambiguous.

The present study has the following limitations:

1) the time effort of mt was substantially greater than that of em volumetric analysis,therefore, its use in the clinical setting is not routinely practicable and technical develop- ments are needed to optimize this aspect; 2) consistent with previous studies, the 3D-US datasets were interpo- lated to 3-mm slice thickness in the transverse dimension

to limit the time effort of the mt method [10]. While the image quality was slightly worsened due to the partial volume effect, a significant negative impact on the results was not expected because the wall structures were easily identifiable; 3) the clinical relevance of phantom-gener- ated results is yet to be determined, and clinical studies are warranted for this purpose. A potential problem in hu- man studies is the definition of appropriate anatomical landmarks that are easily identifiable in separate scans.

Conclusions

The generation of a 3D-US EFOV via the stitching of two separate 3D-US datasets obtained via sensor-nav- igated probes proved feasible and accurate, leading to a correct volumetric analysis of even large thyroid phan- toms (up to 400 mL). The quality of the data obtained via mt was superior (but the method was more time consum- ing) compared to the em approach. The present thyroid phantom study provides useful information for further investigations on the potential use of 3D-US in the clini- cal setting of thyroid disease.

Acknowledgements: The authors gratefully ac- knowledge Hiltrud Niggemann for statistical analyses and Dr. Ernesta Palombo-Kinne for revision of the manu- script. GE Healthcare, Milwaukee, WI, USA is gratefully acknowledged for providing the hardware (LOGIQ E9) for this study.

Conflict of interest: none References

1. Kayastha P, Paudel S, Shrestha DM, Ghimire R, Pradhan S.

Study of thyroid volume by ultrasonography in clinically euthyroid patients. J Institute Med 2010;32:36-43.

2. Turcios S, Lence-Anta JJ, Santana JL, et al. Thyroid volume and its relation to anthropometric measures in a healthy cu- ban population. Eur Thyroid J 2015;4:55-61.

3. Grussendorf M, Reiners C, Paschke R, Wegscheider K;

LISA Investigators. Reduction of thyroid nodule volume by levothyroxine and iodine alone and in combination: a ran- domized, placebo-controlled trial. J Clin Endocrinol Metab 2011;96:2786-2795.

4. Lyshchik A, Drozd V, Reiners C. Accuracy of three-dimen- sional ultrasound for thyroid volume measurement in chil- dren and adolescents. Thyroid 2004;14:113-120.

5. Ahidjo A, Tahir A, Tukur MA. Ultrasound determination of thyroid gland volume among adult Nigerians. Int J Radiol 2006;4:1-3.

6. Chaudhary V, Bano S. Thyroid ultrasound. Indian J Endo- crinol Metab 2013;17:219-227.

7. Shapiro RS. Panoramic ultrasound of the thyroid. Thyroid 2003;13:177-181.

(7)

8. Lyshchik A, Drozd V, Schloegl S, Reiners C. Three-dimen- sional ultrasonography for volume measurement of thyroid nodules in children. J Ultrasound Med 2004;23:247-254.

9. Schlogl S, Werner E, Lassmann M, et al. The use of three- dimensional ultrasound for thyroid volumetry. Thyroid 2001;11:569-574.

10. Freesmeyer M, Wiegand S, Schierz JH, Winkens T, Li- cht K. Multimodal evaluation of 2-D and 3-D ultrasound, computed tomography and magnetic resonance imaging in measurements of the thyroid volume using universally ap- plicable cross-sectional imaging software: a phantom study.

Ultrasound Med Biol 2014;40:1453-1462.

11. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement.

Lancet 1986;1:307-310.

12. Poon TC, Rohling RN. Three-dimensional extended field- of-view ultrasound. Ultrasound Med Biol 2006;32:357-369.

13. Dyer E, Zeeshan Ijaz U, Housden R, Prager R, Gee A, Treece G. A clinical system for three-dimensional extended- field-of-view ultrasound. Br J Radiol 2012;85:e919-e924.

14. Wachinger C, Wein W, Navab N. Three-dimensional ultra- sound mosaicing. Med Image Comput Comput Assist In- terv 2007;10:327-335.

15. Kim SC, Kim JH, Choi SH, et al. Off-site evaluation of three-dimensional ultrasound for the diagnosis of thyroid nodules: comparison with two-dimensional ultrasound. Eur Radiol 2016;26:3353-3360.

16. Dalvi R, Hacihaliloglu I, Abugharbieh R. 3D ultrasound volume stitching using phase symmetry and harris corner detection for orthopaedic applications. Proc SPIE 7623, Medical Imaging 2010: Image Processing, 762330 (12 March 2010); doi:10.1117/12.844608.

17. Ni D, Qu Y, Yang X, et al. Volumetric Ultrasound Panorama Based on 3D SIFT. In: Metaxas D, Axel L, Fichtinger G, Székely G. (Eds.). Medical Image Computing and Com- puter-Assisted Intervention – MICCAI 2008: 11th Interna- tional Conference, New York, NY, USA, September 6-10,

2008, Proceedings, Part II. Berlin, Heidelberg: Springer Berlin Heidelberg; 2008:52-60.

18. Turgut AT, Coşkun ZÜ, Kısmet K, et al. Comparison of Extended Field of View and Dual Image Ultrasound Techniques for the Measurement of the Longitudinal Di- mension of Enlarged Thyroid Glands. J Med Ultrasound 2008;16:150-157.

19. Rafii-Tari H, Lessoway VA, Kamani AA, Abolmaesumi P, Rohling R. Panorama Ultrasound for Navigation and Guidance of Epidural Anesthesia. Ultrasound Med Biol 2015;41:2220-2231.

20. Ni D, Chui YP, Qu Y, et al. Reconstruction of volumetric ultrasound panorama based on improved 3D SIFT. Comput Med Imag Graph 2009;33:559-566.

21. Schneider RJ, Perrin DP, Vasilyev NV, Marx GR, Del Nido PJ, Howe RD. Real-time image-based rigid regis- tration of three-dimensional ultrasound. Med Image Anal 2012;16:402-414.

22. Andermann P, Schlogl S, Mader U, Luster M, Lassmann M, Reiners C. Intra- and interobserver variability of thyroid volume measurements in healthy adults by 2D versus 3D ultrasound. Nuklearmedizin 2007;46:1-7.

23. Riccabona M, Nelson TR, Pretorius DH. Three-dimension- al ultrasound: accuracy of distance and volume measure- ments. Ultrasound Obstet Gynecol 1996;7:429-434.

24. Hermans R, Bouillon R, Laga K, et al. Estimation of thy- roid gland volume by spiral computed tomography. Eur Ra- diol 1997;7:214-216.

25. Tong S, Cardinal HN, McLoughlin RF, Downey DB, Fen- ster A. Intra- and inter-observer variability and reliability of prostate volume measurement via two-dimensional and three-dimensional ultrasound imaging. Ultrasound Med Biol 1998;24:673-681.

26. Reinartz P, Sabri O, Zimny M, et al. Thyroid volume meas- urement in patients prior to radioiodine therapy: compari- son between three-dimensional magnetic resonance imag- ing and ultrasonography. Thyroid 2002;12:713-717.

Referințe

DOCUMENTE SIMILARE

2 Referring to the constitutional regulation of Kosovo regarding the form of state regulation, we have a unitary state, but in practice the unitary state

The Constitution of the Republic of Albania regulates three situations that require extraordinary measures: war situation, state of emergency and state of natural

, Convergence of the family of the deformed Euler-Halley iterations under the H¨ older condition of the second derivative, Journal of Computational and Applied Mathematics,

Identity is thus constructed in interaction, which means that out of a whole host of potential identity features, those features become salient which permit a differentiation of

The modification of the promotion criteria in the area of medical university education triggered an increase in the number of articles published by the

According to our previous investigations it seems that tolerance, whether regarded as a political practice or a philosophical or moral principle, is a strategy (or tactics) of one

To obtain the spatial orientation of the acquired 2D US periodontal images with respect to the patient and to gen- erate 3D US model reconstruction of a larger area such as

In (Ie¸san and Ciarletta, 1993) is presented the problem of extension and bending of right cylinders made of an isotropic elastic material with voids and in a recent paper Ie¸san