Chest ACCP Education Calendar
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Guest Access | Sign In via User Name/Password
First published online on June 15, 2007
Chest, doi:10.1378/chest.07-0793
doi:10.1378/chest.07-0793
(Chest. 2007; 132:984-990)
© 2007 American College of Chest Physicians
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow All Versions of this Article:
chest.07-0793v1
chest.07-0793v2
132/3/984    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (3)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ikeda, K.
Right arrow Articles by Nomori, H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Ikeda, K.
Right arrow Articles by Nomori, H.

Differential Diagnosis of Ground-Glass Opacity Nodules*

CT Number Analysis by Three-Dimensional Computerized Quantification

Koei Ikeda, MD, PhD; Kazuo Awai, MD, PhD; Takeshi Mori, MD, PhD; Koichi Kawanaka, MD; Yasuyuki Yamashita, MD, PhD and Hiroaki Nomori, MD, PhD

* From the Departments of Thoracic Surgery (Drs. Ikeda, Mori, and Nomori) and Diagnostic Radiology (Drs. Awai, Kawanaka, and Yamashita), Graduate School of Medical Sciences, Kumamoto University, Honjo, Kumamoto, Japan.

Correspondence to: Hiroaki Nomori, MD, PhD, Department of Thoracic Surgery, Graduate School of Medical Sciences, Kumamoto University, 1–1-1 Honjo, Kumamoto 860-8556, Japan; e-mail: hnomori{at}qk9.so-net.ne.jp

Abstract

Objectives: To differentiate among atypical adenomatous hyperplasia (AAH), bronchioloalveolar carcinoma (BAC), and adenocarcinoma showing ground-glass opacity (GGO) on CT scans, we conducted a study to determine the optimal parameter on CT number analysis using three-dimensional (3D) computerized quantification.

Methods: From the CT numbers of GGO lesions obtained by 3D computerized quantification, CT number histogram pattern, peak CT number on the histogram, mean CT number, and the 5th to 95th percentile CT numbers were analyzed to determine the optimal parameter for differentiation among AAH (n = 10), BAC (n = 21), and adenocarcinoma (n = 12).

Results: While the CT number histogram showed one peak in all 10 of the AAH lesions (100%), it showed two peaks in 8 of 21 BAC lesions (38%), and in 5 of 12 adenocarcinoma lesions (42%). For differentiation between AAH and BAC, the 75th percentile CT number with a cutoff value of –584 Hounsfield units (HU) was optimal, with a sensitivity of 0.90 and a specificity of 0.81. For differentiation between BAC and adenocarcinoma, a mean CT number with a cutoff value of –472 HU was optimal, with a sensitivity of 0.75 and a specificity of 0.81.

Conclusions: From the analysis of CT numbers of GGO lesions obtained by 3D computerized quantification, we conclude the following: (1) two peaks on the CT number histogram can rule out AAH; (2) the 75th percentile is the optimal CT number for differentiating between AAH and BAC; and (3) the mean CT number is the optimal CT number for differentiating between BAC and adenocarcinoma.

Key Words: lung cancer • pathology lung cancer • radiology lung cancer

While advances in high-resolution CT (HRCT) scanning have increased the detection of small ground-glass opacity (GGO) nodules and also allowed such images to be investigated in detail, it is often difficult to differentiate among atypical adenomatous hyperplasia (AAH), bronchioloalveolar carcinoma (BAC), and well-differentiated adenocarcinoma. In addition, while the term GGO is used to describe a hazy increased attenuation of the lung seen on a CT scan with preservation of bronchiole and vascular margins, the criteria for defining a GGO are as "hazy" as the CT scan image. Several authors1234 have used the ratio between GGO and solid areas to classify the malignant grade of adenocarcinomas, but this method of differentiation is also partly subjective. In contrast, other authors5678 have reported methods for the objective evaluation of GGO lesions, such as CT number histogram and some types of software. We previously reported56 that the CT number histogram was useful for differentiating among AAH, BAC, and adenocarcinoma, although our previous CT number histogram was constructed from only one slice at the largest cut surface of tumors, which did not allow the evaluation of three-dimensional (3D) CT numbers.

On the other hand, multidetector CT scanning, which is now widely used in routine clinical practice, offers high scan speed and enables volume data for pulmonary nodules to be obtained with thin sections. In this study, we performed a 3D analysis of the CT number from volume data obtained with multidetector CT scanning and investigated the optimal parameters for differentiating among AAH, BAC, and adenocarcinoma.

Materials and Methods

Eligibility
The study protocol for examining CT numbers on HRCT scans in patients with GGO nodules was approved by the ethics committee of Kumamoto University Hospital in May 2005. Informed consent was obtained from all patients after they had a discussion of the risks and benefits of the study with their surgeons.

Patients
Between June 2005 and December 2006, 38 patients with GGO nodules prospectively underwent HRCT scans, and the CT number obtained by 3D computerized quantification was examined. IV contrast material was not used for this examination. Of these, 43 GGO nodules in 33 patients were resected and examined pathologically, and were included in this study (Table 1 ). The number of the lesions resected per patient was one in 27 patients, two in 3 patients, three in 2 patients, and four in 1 patient. All of the lesions were < 3 cm in size and showed GGO with a small solid component. The histologic diagnosis was AAH in 10 lesions, BAC in 21 lesions, and adenocarcinoma in 12 lesions. The mean size was 0.8 cm (range, 0.6 to 1.0 cm) in AAH lesions, 1.4 cm (range, 0.5 to 2.4 cm) in BAC lesions, and 1.8 cm (range, 1.2 to 2.5 cm) in adenocarcinoma lesions. All of the 12 adenocarcinoma lesions were histologically well differentiated. All 21 BAC lesions and all 12 adenocarcinoma lesions were T1N0M0. The histologic criteria for AAH, BAC, and adenocarcinoma were based on the 1999 World Health Organization histologic classification.910 To check stromal, vascular, and pleural invasion, we routinely performed Victoria blue staining.


View this table:
[in this window]
[in a new window]

  		Table 1.. Clinicopathologic Characteristics of Patients With GGO Nodules

 
Segmentation of GGO Nodules by a Computerized Automated Diagnosis System
Volume scans of the nodules were performed using a four-detector CT scanner (LightSpeed QX/I; GE Medical Systems; Milwaukee, WI). The scanning parameters were as follows: detector collimation, 4 x 1.25 mm; helical pitch, 0.75; section thickness, 1.25 mm: section interval, 1.25 mm; rotation time, 0.8 s; tube voltage, 120 kVp; and tube current, 160 to 200 mA. The segmentation of GGO nodules was conducted using a computerized automated diagnosis (CAD) system developed in-house, which has been reported previously.1112 First, one radiologist (K.A.) specified the region of interest on a section that included the target GGO nodule (Fig 1 ). The threshold CT number between GGO nodules and surrounding normal lung tissue was set at –700 to –800 Hounsfield units (HU). The CAD system automatically identified the GGO nodules in all x-axis, y-axis, and z-axis directions from the surrounding normal lung tissue. The elimination of normal structures within or around the nodule, such as vessels and bronchioli, was performed using several image-processing techniques,.1314 Therefore, the nodule was identified as the lesion area without vessels and bronchioli.


Figure 1
View larger version (67K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1.. Top left: a GGO nodule segmented by a CAD system. Top right: 3D image of the GGO lesion segmented by the CAD system. Bottom left: coronal view of the lesion. Bottom right: sagittal view of the lesion.

 
Analysis of CT Number of GGO Nodules
CT numbers of nodules were tallied up three-dimensionally with a computer workstation (CELSIUS; Fujitsu; Tokyo, Japan) with dual 3.0-GHz processors (Xeon; Intel; Santa Clara, CA). A CT number histogram was generated with the class interval of 8 HU. We quantified the following variables: (1) the CT number histogram pattern; (2) the CT number at the highest peak on the histogram; (3) the mean CT number; and (4) the CT numbers at the 5th, 10th, 25th, 50th, 75th, 90th, and 95th percentiles (eg, a 25th percentile value means a CT number of the 25% of the pixels calculated from the pixel with the minimum CT number, and a 50th percentile signifies the median CT number).

Evaluation by Receiver Operating Characteristic Curve
Optimum variables to differentiate among AAH, BAC, and adenocarcinoma were evaluated on receiver operating characteristic (ROC) curves using a statistical software package (SPSS; SPSS Inc; Chicago, IL).

Statistical Analysis
All values in the text and tables are given as the mean ± SD. Category data were compared using the Fisher exact test. The data for the CT number were analyzed for significance using the two-tailed Student t test. A Bonferroni test was used to determine significance for comparisons among the three parts. p Values of < 0.05 were accepted as significant.

Results

The CT number histograms showed the following two patterns: one in which there was a peak at a low CT number zone (n = 30); and the other with two peaks, one each at the low and high CT number zones (n = 13) [Fig 2 ]. The low CT number zone ranged from –440 to –800 HU, and the high zone ranged from –24 to –96 HU. In the pattern with two peaks, the peak at the low CT number zone was usually higher than that at the high zone. While all of the 10 AAH lesions showed the one-peak pattern, 8 of the 21 BAC lesions (38%) and 5 of the 12 adenocarcinoma lesions (42%) showed the two-peak pattern (Table 2 ). There was no significant difference in histogram pattern between BAC and adenocarcinoma lesions.


Figure 2
View larger version (54K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2.. CT number histogram and CT image. The vertical axis in each histogram shows the number of pixels in the lesion. The horizontal axis shows the CT numbers, for each of which the class interval was 8 HU. Top, a: AAH with one peak on the histogram. Middle, b: BAC with one peak on the histogram. Bottom, c: adenocarcinoma with two peaks on the histogram, showing a mixed solid component in a GGO lesion.

 

View this table:
[in this window]
[in a new window]

  		Table 2.. Patterns of CT Number Histogram in AAH, BAC, and Adenocarcinoma*

 
Table 3 shows the area under the curve (AUC) on the ROC curve for each variable for differentiating among AAH, BAC, and adenocarcinoma lesions. For differentiating AAH from BAC or differentiating AAH from BAC and adenocarcinoma, a 75th percentile CT number showed the largest AUC, indicating the highest sensitivity and specificity. For differentiating adenocarcinoma from BAC or differentiating adenocarcinoma from BAC and AAH, the mean CT number showed the largest AUC. The ROC curve to differentiate between AAH and BAC with the 75th percentile CT number showed the cutoff value to be –584 HU (Fig 3 ). The ROC curve to differentiate between adenocarcinoma and BAC with the mean CT number showed the cutoff value to be –472 HU (Fig 4 ). Figure 5 shows the distribution of the 75th percentile CT number in AAH, BAC, and adenocarcinoma. The mean values of the 75th percentile of AAH, BAC, and adenocarcinoma were –609 ± 45, –450 ± 147, and –319 ± 97 HU, respectively, which shows a significant difference between AAH and BAC and between BAC and adenocarcinoma (p < 0.05). Figure 6 shows the distribution of mean CT numbers for AAH, BAC, and adenocarcinoma. The mean values of the mean CT number for AAH, BAC, and adenocarcinoma were –660 ± 35, –556 ± 95, and –442 ± 99 HU, respectively, with a significant difference between AAH and BAC and between BAC and adenocarcinoma (p < 0.01).


View this table:
[in this window]
[in a new window]

  		Table 3.. AUC of ROC Curve for Differentiating Among AAH, BAC, and Adenocarcinoma*

 

Figure 3
View larger version (9K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3.. ROC curve of the CT number of the 75th percentile for differentiation between AAH and BAC, showing the cutoff value to be –584 HU.

 

Figure 4
View larger version (8K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4.. ROC curve of the mean CT number for differentiation between BAC and adenocarcinoma, showing the cutoff value to be –472 HU.

 

Figure 5
View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5.. The plots of the CT number of the 75th percentile in AAH, BAC, and adenocarcinoma lesions. The dotted line indicates a –584-HU cutoff value between AAH and BAC. The differences between AAH and BAC, and between BAC and adenocarcinoma were significant (p < 0.05). Bars = mean ± SD.

 

Figure 6
View larger version (19K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 6.. The plots of the mean CT number in AAH, BAC, and adenocarcinoma. The dotted line indicates a –472-HU cutoff value between BAC and adenocarcinoma. The differences between AAH and BAC, and between BAC and adenocarcinoma were significant (p < 0.01). Bars = mean ± SD.

 
Table 4 shows the result of differentiation between AAH and BAC with a cutoff value of –584 HU at the 75th percentile, for which the sensitivity was 0.9, the specificity was 0.76, and the accuracy was 0.81. Table 5 shows the result of differentiation between adenocarcinoma and BAC with a cutoff value of –472 HU at the mean CT number, with a sensitivity of 0.75, a specificity of 0.81, and an accuracy of 0.79.


View this table:
[in this window]
[in a new window]

  		Table 4.. Differentiation Between AAH and BAC With the 75th Percentile CT Number*

 

View this table:
[in this window]
[in a new window]

  		Table 5.. Differentiation Between Adenocarcinoma and BAC With the Mean CT Numbers*

 
Discussion

This study showed that the 75th percentile CT number of GGO lesions was the optimal CT number variable for differentiating between AAH and BAC or AAH and another histology (ie, BAC and adenocarcinoma). We previously reported that, by examining the CT numbers on two-dimensional images, the peak CT histogram number was superior to the mean CT number for differentiating between AAH and BAC.5 Because this study examined the CT numbers from 3D computerized quantification and investigated a many more parameters, including several percentile CT numbers, we believe that the data from this study is more precise than that from our previous study. The reason why the 75th percentile was superior to the peak CT number on the histogram could be the following: (1) because BAC frequently shows an AAH-like component at the peripheral site, its peak CT number on the histogram may sometimes be similar to that of AAH, especially in BAC with a broad AAH-like component; (2) because the 75th percentile indicates the high CT number zone within the lesions, differences in the thickening of alveolar septa and cellularity between AAH and BAC may be revealed.

However, 1 of the 10 AAH lesions and 5 of the 21 BAC lesions could not be distinguished from each other by the 75th percentile CT number. Histologic findings of one indistinguishable AAH lesion showed reasonable cellularity and nuclear atypia as typically seen in AAH, but had thickened alveolar septa like BAC, resulting in a higher 75th percentile value than the cutoff value. Five indistinguishable BAC lesions showed reasonable nuclear atypia as typically seen in BAC, but they had less thickened alveolar septa than usually seen in BAC, resulting in lower 75th percentile values than the cutoff value. It is well known that histologic differentiation between AAH and BAC is sometimes difficult. Kitamura et al classified AAH lesions into low grade and high grade, of which the later was histologically similar to BAC.15 Therefore, we consider that the complete differentiation between AAH and BAC is difficult even using the 3D analysis of CT number. However, this study showed that all of the 10 AAH lesions showed one peak on the CT number histogram, whereas 38% of BAC lesions and 42% of adenocarcinoma lesions showed two peaks. It is also well known that AAH lesions usually appear as pure GGOs, whereas some BAC and adenocarcinoma lesions appear as GGOs mixed with a solid component. The CT number histogram pattern can clearly reveal a mixed solid component within GGO lesions as the second peak; therefore, AAH can be ruled out and recommended for surgical resection when the histogram shows a two-peak pattern.

In contrast, the CT number histogram pattern (ie, one or two peaks) could not be used to differentiate between BAC and adenocarcinoma. While the peak at the high CT number zone on the histogram in cases of adenocarcinoma usually showed an increased amount of central fibrosis, that in BAC showed an area of alveolar structural collapse, both of which can be easily distinguished from pathologic findings.16 However, HRCT imaging has a very limited ability to distinguish between tumor fibrosis and alveolar structural collapse, resulting in a similar frequency of two peak patterns between BAC and adenocarcinoma. In contrast, the lesions with higher than –472 HU of mean CT numbers were more frequently adenocarcinoma than BAC. Therefore, mean CT number is more useful in differentiating between BAC and adenocarcinoma.

While a BAC lesion of < 2 cm has been reported1718 to be able to be cured by wedge resection, adenocarcinoma should be treated by lobectomy or segmentectomy even for stage p-T1N0M0 disease, because a wedge resection can cause a local relapse due to the spread of tumor cells into lymphatic vessels outside the primary tumor.19 We believe that mean CT number could determine operative procedures for GGO lesions.

How often and how long do the GGO lesions suspected of AAH need to be followed by CT scanning? By analogy with multistep carcinogenesis, AAH could progress to adenocarcinoma through BAC. Even if AAH subsequently develops into BAC, follow-up with CT scanning would not miss the chance of surgical cure, because the lesion would remain within the T1N0M0 stage during the follow-up period until it grows to a diameter of 2 cm.216 Aoki et al3 reported that the mean tumor doubling time of BAC was 880 days. Therefore, careful follow-up by CT scanning every 2 years would be enough for GGO lesions suspected to be AAH.

We conclude the following from the 3D analysis of the CT number of GGO lesions: (1) two peaks on the CT number histogram can rule out AAH; (2) the 75th percentile is the optimal CT number for differentiating between AAH and BAC; and (3) the mean CT number is the optimal CT number to differentiate between BAC and adenocarcinoma. While it has been time-consuming to obtain information on 3D imaging from CT scans, we believe that further advances in software and CAD systems will make the procedure reported in this study easier.

Footnotes

Abbreviations: AAH = atypical adenomatous hyperplasia; AUC = area under the curve; BAC = bronchioloalveolar carcinoma; CAD = computerized automated diagnosis; GGO = ground-glass opacity; HRCT = high resolution CT; HU = Hounsfield units; ROC = receiver operating characteristic; 3D = three-dimensional

The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

Received for publication March 29, 2007. Accepted for publication May 17, 2007.

References

  1. Kuriyama, K, Seto, M, Kasugai, T, et al (1999) Ground-glass opacity on thin-section CT: value in differentiating subtypes of adenocarcinoma of the lung. AJR Am J Roentgenol 173,465-469[Abstract/Free Full Text]
  2. Higashiyama, M, Kodama, K, Yokouchi, H, et al Prognostic value of bronchiolo-alveolar carcinoma component of small lung adenocarcinoma. Ann Thorac Surg 1999;68,2069-2073[Abstract/Free Full Text]
  3. Aoki, T, Nakata, H, Watanabe, H, et al Evolution of peripheral lung adenocarcinomas: CT findings correlated with histology and tumor doubling time. AJR Am J Roentgenol 2000;174,763-768[Abstract/Free Full Text]
  4. Kodama, K, Higashiyama, M, Yokouchi, H, et al Prognostic value of ground-glass opacity found in small lung adenocarcinoma on high-resolution CT scanning. Lung Cancer 2001;33,17-25[CrossRef][ISI][Medline]
  5. Nomori, H, Ohtsuka, T, Naruke, T, et al Differentiating between atypical adenomatous hyperplasia and bronchioloalveolar carcinoma using the computed tomography number histogram. Ann Thorac Surg 2003;76,867-871[Abstract/Free Full Text]
  6. Nomori, H, Ohtsuka, T, Naruke, T, et al Histogram analysis of computed tomography numbers of clinical T1N0M0 lung adenocarcinoma with special reference to lymph node metastasis and tumor invasiveness. J Thorac Cardiovasc Surg 2003;126,1584-1589[Abstract/Free Full Text]
  7. Matsuguma, H, Nakahara, R, Anraku, M, et al Objective definition and measurement method of ground-glass opacity for planning limited resection in patients with clinical IA adenocarcinoma of the lung. Lung Cancer 2004;25,1102-1106
  8. Nakata, M, Sawada, S, Yamashita, M, et al Objective radiological analysis of ground-glass opacity aimed at curative limited resection for small peripheral non-small cell lung cancer. J Thorac Cardiovasc Surg 2005;129,1226-1231[Abstract/Free Full Text]
  9. Travis, MD, Colby, TV, Corrin, B, et al World Health Organization international histological classification of tumors: histological typing of lung and pleural tumors 3rd ed. 1999 Springer-Verlag. Berlin, Germany:
  10. Brambilla, E, Travis, WD, Colby, TV, et al The new World Health Organization classification of lung tumors. Eur Respir J 2002;18,1059-1068[CrossRef][ISI]
  11. Awai, K, Murao, K, Ozawa, A, et al Pulmonary nodules at chest CT: effect of computer-aided diagnosis on radiologists detection performance. Radiology 2004;230,347-352[Abstract/Free Full Text]
  12. Awai, K, Murao, K, Ozawa, A, et al Pulmonary nodules: estimation of malignancy at thin-section helical CT-effect of computer-aided diagnosis on performance of radiologists. Radiology 2006;239,276-284[Abstract/Free Full Text]
  13. Kass, M, Witkin, A, Terzopoulos, D Snakes: active contour models. Int J Comput Vis 1988;1,321-331[CrossRef]
  14. Dougherty, E Digital image processing method. Thompson, BJ eds. Optical engineering 1994,77-85 Dekker. New York, NY:
  15. Kitamura, H, Kameda, Y, Nakamura, N, et al Proliferative potential and p53 overexpression in precursor and early stage lesions of bronchioloalveolar lung carcinoma. Am J Pathol 1995;146,876-887[Abstract]
  16. Noguchi, M, Morikawa, A, Kawasaki, M, et al Small adenocarcinoma of the lung: histologic characteristics and prognosis. Cancer 1995;75,2844-2852[CrossRef][ISI][Medline]
  17. Nakata, M, Sawada, S, Saeki, H, et al Prospective study of thoracoscopic limited resection for ground-glass opacity selected by computed tomography. Ann Thorac Surg 2003;75,1601-1606[Abstract/Free Full Text]
  18. Yoshida, J, Nagai, K, Yokose, T, et al Limited resection trial for pulmonary ground-glass opacity nodules: fifty-case experience. J Thorac Cardiovasc Surg 2005;129,991-996[Abstract/Free Full Text]
  19. Ichinose, Y, Yano, T, Yokoyama, H, et al The correlation between tumor size and lymphatic vessel invasion in resected peripheral stage I non-small cell lung cancer: a potential risk of limited resection. J Thorac Cardiovasc Surg 1994;108,684-686[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 ChestHome page
F. Detterbeck, H. Nomori, K. Ikeda, K. Awai, T. Mori, K. Kawanaka, and Y. Yamashita
Great Data, But How Do We Use It?
Chest, March 1, 2008; 133(3): 827 - 828.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow All Versions of this Article:
chest.07-0793v1
chest.07-0793v2
132/3/984    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (3)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ikeda, K.
Right arrow Articles by Nomori, H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Ikeda, K.
Right arrow Articles by Nomori, H.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS