| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Department of Radiology, Section of Thoracic Imaging, New York University Medical Center, 560 1st Ave, New York, NY 10016 (J.P.K., H.R., E.L.J., G.M., D.P.N.); Fox Chase Cancer Center, Philadelphia, Pa (J.S.B.); and Computer Science Department, Boston University, Mass (M.B.). From the 2001 RSNA scientific assembly. Received February 7, 2002; revision requested April 15; final revision received October 28; accepted December 19. Supported by a Scholars Award from the RSNA Research and Education Foundation and a Seed Grant from the Society of Thoracic Radiology. Address correspondence to J.P.K. (e-mail: jane.ko@med.nyu.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
© RSNA, 2003
Index terms: Computed tomography (CT), image processing, 60.12117 • Computed tomography (CT), technology, 60.12117 • Images, analysis, 60.12117 • Lung, CT, 60.12117 • Lung, nodule, 60.12117
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
As a consequence, attention has been focused on nodule growth to differentiate malignant from benign nodules. Malignant nodules typically double in volume between 30 days and 14 months (9,10). Doubling times of less than 30 days are usually associated with inflammatory or infectious causes. Although times longer than 14 months have been associated with benign nodules such as hamartomas (9), it should be emphasized that some bronchioloalveolar carcinomas, especially those appearing as a ground-glass nodule, have been reported as having doubling times with a mean of 880 days (11).
Currently, nodule measurement is typically assessed in two dimensions rather than in three dimensions. Unfortunately, measurements performed by radiologists are subject to inter- and intraobserver variations (12–14). Therefore, attention has focused on two-dimensional area and three-dimensional computer-aided quantitative volume measurement algorithms that depend on nodule segmentation by means of gray-level thresholds. Such three-dimensional approaches are promising in that they minimize variation and decrease measurement error (15,16).
Volume quantification of ground-glass nodules and non–threshold-based quantitative methods that minimize partial-volume effects have not been studied extensively (17). Volumes of ground-glass nodules are more difficult to measure secondary to their lower contrast to the lung parenchyma compared with those of solid soft-tissue attenuation nodules. The low-dose CT technique, introduced to limit radiation exposure to patients (5,18,19) and commonly used to screen high-risk individuals for lung cancer, has not been investigated extensively in terms of its effect on nodule quantification. Choices of reconstruction algorithm, nodule size, and thoracic geometry have been shown to affect measurement of nodule attenuation (20,21) and potentially may affect quantification of nodule volume.
Thus, the purpose of our study was to compare a variety of three-dimensional methods for quantifying pulmonary nodule volume at chest CT and to study the effect of imaging variables.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Forty manufactured approximately spherical nodules were studied. The nodules were composed of a plastic created by blending epoxy resins and urethanes. Twenty simulated ground-glass nodules had attenuations of approximately -360 HU, and 20 simulated solid nodules had attenuations of approximately 50 HU. There were four nodule size categories for each nodule attenuation. Each size category had five nodules. The four nodule sizes were selected so that their approximate diameters were larger than 2 mm and smaller than 5 mm. To obtain their true volumes, the nodules were weighed two times on a precision scale (model AJ100; Mettler Toledo, Columbus, Ohio) that was capable of 1-mg accuracy. For the few cases with a discrepancy greater than 0.5 mg between the two weights, the nodule was weighed a third time, and all three measurements were averaged. The true volume of a nodule was calculated by multiplying the nodule weight by the specific gravity of the material used to construct the nodule. Two values of specific gravity, one for ground-glass nodules and the other for solid-attenuation nodules, were provided by the manufacturer and verified in our laboratory. Specific gravity of each material was verified by means of precision machining of a cylinder with diameter (d) of 40 mm and height (h) of 20 mm. The exact largest dimensions were measured with precision calipers (±0.1 mm) and used to calculate volume [
h(d/2)2]. Each cylinder was weighed three times on the precision scale, and the specific gravity was calculated by dividing weight by volume. We estimated the nodule volumes to be accurate within 0.5%.
One nodule was placed into each well and surrounded by a mixture created so that its mean attenuation (-786 HU) was equivalent to that of the phantom lung parenchyma (-780 HU) (Fig 1). The loose mixture presented a realistic background with fine texture. The mixture was composed of two relatively homogeneous particulate materials, one with lower attenuation (ground marjoram, 72%) and the other with higher attenuation (ground coffee, 28%) compared with that of the phantom lung parenchyma. The materials were selected from a number of particulate materials that had been imaged with CT and measured for attenuation. The relative proportions of coffee and marjoram were determined by solving a linear equation with the desired attenuation of the final mixture and the known attenuations of the two materials. CT images showed no air pockets at the interface of the nodule and the surrounding mixture. Each phantom region (eg, anterior, posterior) had one nodule of each size category, and the ground-glass and solid-attenuation nodules were placed in the left and right lungs, respectively. We assumed no difference in measurement error between right and left lung location.
|
|
|
|
|
Nodule volume was analyzed with variations of a partial-volume method (PVM) and a binary threshold method. Two types of threshold methods, a variable-threshold method (VTM) and a fixed-threshold method (FTM), were used. The PVM was based on the fact that the mean attenuation in a region of interest (ROI) drawn around a nodule ("large ROI") was reflective of the proportions and attenuation of structures contained within. The large ROI included about 3 mm of lung tissue along the circumference of a nodule. The relationship Vr x Ar = Vn x An + (Vr - Vn) x Al was used to solve for Vn, the volume of the nodule in a transverse section. Vr was the volume of the large ROI, Ar was the mean attenuation in the large ROI, An was the pure nodule attenuation, and Al was the pure lung attenuation. When measured, An was obtained from the image on which the nodule was most conspicuous; at least 20 voxels of the central region of a nodule were measured, except for the smallest nodules (approximately 2.5 mm), in which at least 5 voxels in the center were sampled. Al was obtained by averaging three lung attenuation measurements adjacent to the nodule for each transverse section analyzed.
For the threshold methods, the voxels above a threshold value in the large ROI were counted. For the simplest strategy, which was the FTM, a threshold of -500 HU was applied to all images analyzed for each nodule. For the VTM, the attenuation threshold value was an average between An obtained for the PVM and the mean of five pure lung attenuation measurements obtained from the general lung. The nodule volume per transverse section was calculated by multiplying the number of voxels above the threshold by the voxel volume.
Since the value of the pure nodule attenuation was suspected to contribute critically to the variance in the volume estimates, variations of the PVM and threshold methods (VTM and FTM) were studied. These variants differed in terms of the definition of the pure nodule attenuation (Table 1). PVM1 and VT1 may be clinically applicable only when multiple nodules are present. With PVM2 and VTM2, pure nodule attenuation is sampled from each nodule being measured, regardless of its size. In the variant PVMA, assumed pure nodule attenuation values were used (-380 HU for ground-glass–attenuation nodules and 20 HU for solid-attenuation nodules) that were fixed but were about 20 HU less than the true attenuation values of the nodules.
|
Several weeks later, this observer repeated each of the six measurement methods with 24 randomly selected nodule image sets, which led to 144 measurements, without knowledge of the previous ROI locations and measurement results. Without knowledge of the results obtained by the first observer, a second observer (J.P.K.) performed each of the six measurements on 60 nodule image sets, which comprised 20 ground-glass–attenuation nodules imaged at 120 and 20 mAs (n = 40 nodule image sets) and 20 solid-attenuation nodules imaged at 20 mAs (n = 20 sets), which yielded 360 measurements. The 60 image sets were selected to include ground-glass–attenuation nodules with 20-mAs technique, given the likelihood that measurements could be affected by lower contrast and higher image noise in these scenarios. One observer (E.L.J.) measured the smallest distance of a nodule to the pleura by using the electronic calipers on the workstation.
Data Analysis
Let Vp,r denote the measured volume of a nodule, where the lumped index p represents the unknown characteristics of the nodule (such as size and attenuation) and the lumped index r represents the known measurement characteristics (tube current time, reconstruction algorithm, computational algorithm). Owing to the imaging limitations, such as CT collimation (imperfect section profile) and in-plane spatial resolution, we expect a relationship Vp,r = KrTp, where Kr 
1 is a multiplicative bias of the method and Tp is the true volume of a nodule. In the first stage of data analysis, we calibrated the measurements to eliminate the multiplicative bias due to the imaging imperfections. For each value of index r, we used the least-squares fit of the subset of measured values to the true data to derive the factors Kr. These factors will be used in our future clinical measurements of nodule volume and growth. We calibrated the measured volumes and computed the absolute error (AE) of each technique as AEp,r = |Vp,r/Kr - Tp| . The absolute residual difference between the true and bias-corrected observed volumes, or absolute error, was interpreted as a measure of precision.
The absolute errors were analyzed statistically by means of repeated measures analysis of variance that incorporated correlations introduced through the bias adjustment. Since a single least squares estimate of the slope parameter was used to adjust all volume assessments in a bias group, the absolute errors in a bias group are correlated. Initially, the model included the following as fixed effects: nodule location, nodule size, nodule attenuation, reconstruction algorithm, tube current times, nodule distance from the pleura, and measurement method, as well as all two-factor interactions among these factors. Bias group was included as a random factor. Nodule size and distance from the pleura were treated as numeric rather than ordinal categoric factors. This implies that the effect of each factor on absolute error is linear, and there was no indication in the data to refute this assumption. P values reflected the effect of a given factor after adjustment for the effects of all other factors. Methods were combined into groups and compared in terms of mean absolute error. Inter- and intraexaminer reliability of each method was assessed by means of the fixed effects interclass correlation coefficient (24). The coefficients were calculated with 95% CIs. Differences in bias were tested by means of the Student t test. Statistical analysis was performed (SAS; SAS Institute, Cary, NC).
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
By using the 960 observations (960 = six methods x two tube current times x two reconstruction algorithms x two nodule densities x five locations x four sizes), the tube current time, reconstruction algorithm, quantitative method, nodule attenuation, and nodule size significantly (P < .001) affected the volume error. No significant effect on volume error was noted for nodule location (P = .18) or distance from the pleura (P = .83).
Nodule volume measured on images reconstructed with a high-frequency reconstruction algorithm (mean absolute error = 3.0 mm3) was more precise than that measured on images reconstructed with a low-frequency algorithm (mean absolute error = 3.7 mm3) (P = .002). A high-frequency reconstruction algorithm significantly interacted with both method (P < .001) and attenuation (P < .001). This implies that in terms of minimizing the expected mean absolute error, the choice of optimal algorithm might depend on the method of measurement and/or nodule attenuation. However, a high-frequency algorithm had a lower mean absolute error with all methods of measurement, irrespective of nodule attenuation. Hence, the high-frequency algorithm was determined to be superior to the low-frequency algorithm under the conditions investigated in this study. Consequently, the analyses were redone, and all subsequent results pertain only to the data obtained with the high-frequency algorithm. Factors with significant effects on mean absolute error with this analysis are listed in Table 2.
|
|
|
|
|
Systematic multiplicative measurement bias (true volume divided by calculated volume) of the volume measurements with all variants of the PVMs was 0.91 ± 0.04. Bias of measurements with the threshold methods was 1.06 ± 0.09 for VTM1, 0.99 ± 0.18 for VTM2, and 0.99 ± 0.07 for FTM. Bias of measurements with the low-dose technique was 0.94 ± 0.07, and that with the high-dose technique was 0.99 ± 0.08 (P = .001), which was closer to the value 1.0.
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
One-dimensional (26) or two-dimensional (12) perpendicular measurements are used in oncology practice to measure lesion size and any change in size over time; however, they are suboptimal. The extent of a nodule in the craniocaudal direction at transverse imaging is overlooked. Results of measurements of lesions by human observers are affected by interobserver variability and are reduced when radiologists are assisted by a semiautomated autocontour technique for measuring two-dimensional perpendicular diameters (12). Measurements of nodules are affected by partial-volume averaging, which makes size estimates susceptible to window display settings and section thickness (13).
Three-dimensional methods applied to thin-section CT images enable accurate assessment of nodule size by means of quantification of nodule volume (15,16). Yankelevitz et al (16) reported use of two-dimensional and three-dimensional semiautomated thresholding techniques for volume measurement. For small solid spherical nodules that were imaged in air with 1.0- and 0.5-mm-thick sections with a single–detector row helical CT scanner, volume errors of 2.88% and 1.05%, respectively, were demonstrated for 3–6-mm nodules (16). When the spheres were deformed, an increase in volume error to 3% on 1.0-mm-thick sections and a significant advantage of three-dimensional over two-dimensional methods were demonstrated for volume measurement.
Despite high-resolution imaging with 1-mm-thick sections, three-dimensional measurement of nodule volume for small nodules approximately 3–5 mm in largest dimension is affected by partial-volume averaging (16). Previous methods that addressed quantification of structures and nodule volume were typically based on thresholding algorithms (12,15,16,27). Thresholding algorithms, which apply attenuation criteria to identify voxels as belonging either to a nodule or background, may not detect or may classify as 100% the peripheral voxels of a nodule, which are most susceptible to partial-volume effect.
Results in our phantom experiment demonstrate that the PVMs have higher precision for measuring volumes, possibly because the effect of partial-volume averaging is reduced. The lower mean absolute error with the PVMs was significant, particularly for 20 mAs, which is typically used with low-dose CT. Moreover, results with the variant PVMA had an advantage over those with the other two methods in the group of PVMs. Therefore, we conclude that for precise measurement of small lung nodules, PVMs should use assumed rather than sampled nodule attenuation for respective solid- and ground-glass–attenuation nodules. Difficulty in measuring nodule attenuation, particularly of small nodules, was demonstrated in previous studies on nodule densitometry (20,21,28). Multiple variables consistently influenced the measurement of nodule attenuation, particularly the reconstruction algorithm (20,21), nodule size (20), and partial-volume effect (20,28). Nodules with mixed ground-glass– and solid-attenuation components have been described with adenocarcinoma (11). Monitoring for the growth of mixed attenuation nodules is feasible, provided that a pure nodule attenuation measurement is maintained for calculating volumes on initial and follow-up studies and provided that the proportions of solid and ground-glass attenuations do not change. However, one should be aware that if PVMs are used, an increase in the solid component of a subsolid nodule may be interpreted falsely as an increase in volume.
When all methods were included, we demonstrated a significantly higher volume error for measurement of ground-glass–attenuation nodules as opposed to that for measurement of solid-attenuation nodules. While more work is needed to assess the error as a function of nodule attenuation, we hypothesize that any differences between methods for ground-glass–attenuation nodules are related to the reduced contrast between the ground-glass nodule and the lung parenchyma. Decreased contrast makes binary threshold segmentation of the nodule from surrounding lung tissue more difficult. The realistic lung background in our phantom and the use of a 30-cm field of view might explain the larger volume errors (1.2–10.0 mm3 for 8–60-mm3 nodules) compared with those in previous studies on solid-attenuation nodules in air (3% error) (16).
The primary clinical use of volume measurements is monitoring of the growth rate of pulmonary nodules. Results in the current study can be used with the probability theory to estimate errors in growth rate. With the precision of 2.1 mm3 that we obtained in our phantom experiment, the SD for assessment of the 20% growth of a 50-mm3 nodule would be 6.5%, which would translate to a 95% CI of 7%, 33%. In the case of the 20% growth of a smaller 12-mm3 nodule measured with 1.2-mm3 precision, the 95% CI would be -12%, 52%. Therefore, knowledge of the precision for volume measurements is essential for correct interpretation of nodule growth.
Results in the current study demonstrate good correlation between two independent observers. Despite the need for manual sampling, all methods yielded acceptable and high inter- and intraexaminer reliability. In the near future, the manual sampling of nodule attenuation and lung background can be readily automated, although implementation of placement of the large ROI and full automation may be difficult secondary to the presence of vessels in the human lung.
In the current study, we demonstrated an increase of at least 0.48 mm3 and a mean increase of 0.86 mm3 in precision of volume measurement when the technique was changed from 20 mAs (tube current of 40 mA with our parameters), which is used to perform low-dose chest CT, to 120 mAs, which is used to perform diagnostic CT. Low-dose chest CT is typically performed with tube currents between 20 and 50 mA (5,29–31). Given the small difference in precision and the benefit of lower radiation exposure to patients, the use of low-dose technique for the follow-up of nodules detected at screening CT is a consideration and a topic for future study.
In our study, a high-frequency reconstruction algorithm provided more precise quantification of pulmonary nodules, and we propose reconstruction of image data by means of a high-frequency algorithm when volume quantification is planned. Greater precision may be related to the higher spatial resolution that a high-frequency algorithm provides (32), which facilitates the sampling of small 2–3-mm nodules. Nodule location and distance to pleura did not significantly affect volume measurement. Results of prior work on nodule densitometry in phantoms demonstrate that beam-hardening artifact associated with chest-wall structures lowered attenuation by approximately 20 HU (20). These differences were likely too small to influence the quantitative methods used in our study. Additionally, in prior studies, inhomogeneity associated with object position in the field of view created attenuation differences of 50 HU (20). Such differences may have been minimized in our study if the CT scanner had equally responsive detectors and a homogeneous radiation beam. Sampling of pure nodule and pure lung measurements from the same vicinity may have overcome both artifact and inhomogeneities related to position.
We emphasize that the present study was performed in a phantom model, which likely results in underestimation of the errors that occur when quantitative methods are applied in vivo. The presence of normal structures, such as vessels and bronchi, or heterogeneity in the surrounding lung, as may result from emphysema, pneumonia, or infiltrative lung disease, may negatively affect volume quantification. In this study, we attempted to simulate variation in the lung parenchyma attenuation by surrounding the nodules with a heterogeneous material. The influence of intervening discrete structures and other factors on volume measurement and detection of growth will need to be clarified in future studies. Use of a PVM, which does not rely on precise binary segmentation of a nodule, may overcome these obstacles.
We did not address the measurement of spiculated and grossly nonspherical nodules or subsolid nodules with mixed ground-glass and solid attenuation. Additionally, we did not use images reconstructed on the basis of transverse intervals of less than 1.0 mm. While this may be clinically feasible in the future, our goal was to investigate imaging protocols that were being used with the available multi–detector row CT technology, which provides high-resolution 1.0-mm-thick sections for nodule analysis that were obtained during the same breath hold. Rather than use a smaller field of view that increased spatial resolution, we chose to use a field of view that encompassed the entire thorax so that we could evaluate the volume errors of nodules on high-resolution data without needing to selectively target reconstructions to specific nodules. Use of a smaller field of view would likely lead to more accurate volume estimates. On the other hand, respiratory motion would likely decrease the precision for all methods, given the creation of artifacts, and in the future could be minimized by using CT scanners with faster rotation speeds.
In summary, a PVM approach is a promising method for quantification of nodule volume and follow-up of nodules discovered at low-dose and diagnostic CT. Results in our study suggest that use of a PVM and a high-frequency reconstruction algorithm will yield improved precision in volume quantification.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, J.P.K., H.R.; study concepts and design, J.P.K., H.R.; literature research, J.P.K.; experimental studies, J.P.K., E.L.J., H.R.; data acquisition and analysis/interpretation, J.P.K., E.L.J., H.R.; statistical analysis, J.S.B.; manuscript preparation, J.P.K., H.R.; manuscript definition of intellectual content, J.P.K., H.R., D.P.N.; manuscript editing, revision/review, and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  F. Girvin and J. P. Ko Pulmonary Nodules: Detection, Assessment, and CAD Am. J. Roentgenol., October 1, 2008; 191(4): 1057 - 1069. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Wang, R. J. van Klaveren, H. J. van der Zaag-Loonen, G. H. de Bock, H. A. Gietema, D. M. Xu, A. L. M. Leusveld, H. J. de Koning, E. T. Scholten, J. Verschakelen, et al. Effect of Nodule Characteristics on Variability of Semiautomated Volume Measurements in Pulmonary Nodules Detected in a Lung Cancer Screening Program Radiology, August 1, 2008; 248(2): 625 - 631. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. G. Ravenel, W. M. Leue, P. J. Nietert, J. V. Miller, K. K. Taylor, and G. A. Silvestri Pulmonary Nodule Volume: Effects of Reconstruction Parameters on Automated Measurements--A Phantom Study Radiology, May 1, 2008; 247(2): 400 - 408. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. Honda, T. Johkoh, H. Sumikawa, A. Inoue, N. Tomiyama, N. Mihara, Y. Fujita, M. Tsubamoto, M. Yanagawa, T. Daimon, et al. Pulmonary Nodules: 3D Volumetric Measurement with Multidetector CT Effect of Intravenous Contrast Medium Radiology, December 1, 2007; 245(3): 881 - 887. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H Bolte, C Riede, S Muller-Hulsbeck, S Freitag-Wolf, G Kohl, T Drews, M Heller, and J Bieder Precision of computer-aided volumetry of artificial small solid pulmonary nodules in ex vivo porcine lungs Br. J. Radiol., June 1, 2007; 80(954): 414 - 421. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. J. Jeong, C. A. Yi, and K. S. Lee Solitary Pulmonary Nodules: Detection, Characterization, and Guidance for Further Diagnostic Workup and Treatment Am. J. Roentgenol., January 1, 2007; 188(1): 57 - 68. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. P. Ko, R. Marcus, E. Bomsztyk, J. S. Babb, C. Stefanescu, M. Kaur, D. P. Naidich, and H. Rusinek Effect of blood vessels on measurement of nodule volume in a chest phantom. Radiology, April 1, 2006; 239(1): 79 - 85. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. R. Goodman, M. Gulsun, L. Washington, P. G. Nagy, and K. L. Piacsek Inherent Variability of CT Lung Nodule Measurements In Vivo Using Semiautomated Volumetric Measurements. Am. J. Roentgenol., April 1, 2006; 186(4): 989 - 994. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Awai, K. Murao, A. Ozawa, Y. Nakayama, T. Nakaura, D. Liu, K. Kawanaka, Y. Funama, S. Morishita, and Y. Yamashita Pulmonary Nodules: Estimation of Malignancy at Thin-Section Helical CT--Effect of Computer-aided Diagnosis on Performance of Radiologists Radiology, April 1, 2006; 239(1): 276 - 284. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. P. Ko, J. Chang, E. Bomsztyk, J. S. Babb, D. P. Naidich, and H. Rusinek Effect of CT Image Compression on Computer-assisted Lung Nodule Volume Measurement Radiology, October 1, 2005; 237(1): 83 - 88. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Goo, T. Tongdee, R. Tongdee, K. Yeo, C. F. Hildebolt, and K. T. Bae Volumetric Measurement of Synthetic Lung Nodules with Multi-Detector Row CT: Effect of Various Image Reconstruction Parameters and Segmentation Thresholds on Measurement Accuracy Radiology, June 1, 2005; 235(3): 850 - 856. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Bornemann, J.-M. Kuhnigk, V. Dicken, S. Zidowitz, B. Kuemmerlen, S. Krass, H.-O. Peitgen, B. B. Wein, H. Schubert, H.-o. Shin, et al. Informatics in Radiology (infoRAD): New Tools for Computer Assistance in Thoracic CT Part 2. Therapy Monitoring of Pulmonary Metastases RadioGraphics, May 1, 2005; 25(3): 841 - 848. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. T. Boll, R. C. Gilkeson, T. R. Fleiter, K. A. Blackham, J. L. Duerk, and J. S. Lewin Volumetric Assessment of Pulmonary Nodules with ECG-Gated MDCT Am. J. Roentgenol., November 1, 2004; 183(5): 1217 - 1223. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
