Computer-aided differentiation of malignant from benign solitary pulmonary nodules imaged by high-resolution CT

doi:10.1016/j.compmedimag.2008.04.001

Computerized Medical Imaging and Graphics

Volume 32, Issue 5, July 2008, Pages 416-422

https://doi.org/10.1016/j.compmedimag.2008.04.001 Get rights and content

Abstract

We investigated the possibility of using computer analysis of high-resolution CT images to radiologically classify the shape of pulmonary nodules. From a total of 107 HRCT images of solid, solitary pulmonary nodules with prior differentiation as benign (n = 55) or malignant (n = 52), we extracted the desired pulmonary nodules and calculated two quantitative parameters for characterizing nodules: circularity and second central moment. Using discriminant analysis for two thresholds in differentiating malignant from benign states resulted in a sensitivity of 76.9%, a specificity of 80%, a positive predictive value of 78.4%, and a negative predictive value of 78.6%.

Introduction

The incidence and mortality rate of lung cancer have been increasing worldwide. The early detection of small lung cancers that are still at an early stage is crucial for improvement of the prognosis. Several studies have demonstrated the detection of small pulmonary nodules (SPNs) by low-dose helical computed tomography (CT) [3], [5], [6], [9], [19]. However, for diagnosis of lung cancer at early stage, the differentiation of pulmonary nodules as benign or malignant is as important as the detection of the nodules. However, the increased rate of detection of pulmonary nodules afforded by helical CT screening has resulted in increased labor (to differentiate malignant from benign forms) on the part of the radiologist. Therefore, radiologists have an immediate need for computer-aided diagnosis (CAD) in the differentiation of malignant and benign nodules.

Both the shape and the internal structure of a nodule are important in the differentiation between benignity and malignancy [1], [2]. Owing to its enhanced spatial resolution, high-resolution CT (HRCT) provides superior characterization of pulmonary tissue as well as depiction of the shape and internal characteristics of pulmonary nodules [2], [7], [18], [20], [21]. Radiologists usually classify pulmonary nodules as benign or malignant based on the shape of their HRCT images.

Several CAD studies have shown that morphologic analysis of pulmonary nodules on HRCT is useful within a CAD scheme [9], [10], [11], [12], [14]. However, few of these CAD studies reflected the scheme employed by radiologists for shape classification of nodules.

We have previously reported that computer image analysis based on just two quantitative shape features (circularity and second central moment) was capable of reproducing the radiological classification of nodules [8]. In this study, we had already compared quantitative measures of shape by computer image analysis with conventional classification by radiologists. In the present study, we examined whether this shape-classification method actually enabled a computer-aided diagnostic differentiation between malignant and benign solitary, solid pulmonary nodules.

Section snippets

Study group

A total of 107 HRCT images of solitary pulmonary nodules from 106 patients were used, with prior differentiation as benign or malignant. There were 48 females and 58 males with a mean age of 60.9 ± 11.3 years. HRCT images were obtained from patients who underwent HRCT examinations at Nagoya University Hospital and Toyota Memorial Hospital between September 1999 and December 2004. Nodules > 30 mm in diameter, those with ground-glass opacity (GGO), and those adhering extensively to pleura or large

Experimental results

Fig. 1 is a scatter diagram plotting second moment against circularity. Nodules having low circularity and low second moment could be malignant.

The average circularity of malignant nodules was 0.513 ± 0.156 and that of benign nodules was 0.619 ± 0.219. The circularity of malignant nodules was significantly lower (p < 0.05). The average second moment of malignant nodules was 0.176 ± 0.018 and that of benign nodules was 0.184 ± 0.021. The second moment of malignant nodules was lower than of benign nodules

Discussion

Of several studies on clinical chest radiology, ours presents well differentiation of malignant from benign SPNs by computer analysis of high-resolution CT images. Kido et al. reported that the fractal dimensions of hamartomas were smaller than those of bronchogenic carcinomas; but, there were no significant differences among the bronchogenic carcinomas, organizing pneumoniae, and tuberculomas upon fractal analysis of the surface of 117 pulmonary nodules [10]. McNitt Gray et al. reported that

Shingo Iwano graduated from the Nagoya University School of Medicine in March 1993, with a bachelor of medicine degree. Further, he graduated from the Graduate School Division of Medical Research, Nagoya University in 1999, and received a PhD degree in 2001 from the Nagoya University. At present, he is an associate professor at the Department of Radiology, Nagoya University Hospital. His current research interest is chest radiology, especially in computer-aided diagnosis. He is a member of the

References (21)

S. Iwano et al.
Solitary pulmonary nodules: optimal slice thickness of high-resolution CT in differentiating malignant from benign
Clin Imaging
(2004)
S. Iwano et al.
Computer-aided diagnosis: a shape classification of pulmonary nodules imaged by high-resolution CT
Comput Med Imaging Graph
(2005)
K. Kanazawa et al.
Computer-aided diagnosis for pulmonary nodules based on helical CT images
Comput Med Imaging Graph
(1998)
M.D. Seemann et al.
Differentiation of malignant from benign solitary pulmonary lesions using chest radiography, spiral CT and HRCT
Lung Cancer
(2000)
S. Sone et al.
Mass screening for lung cancer with mobile spiral computed tomography scanner
Lancet
(1998)
J.J. Erasmus et al.
Solitary pulmonary nodules. Part I. Morphologic evaluation for differentiation of benign and malignant lesions
Radiographics
(2000)
K. Furuya et al.
New classification of small pulmonary nodules by margin characteristics on high-resolution CT
Acta Radiol
(1999)
C.I. Henschke
Early lung cancer action project: overall design and findings from baseline screening
Cancer
(2000)
T. Hyodo et al.
Intrapulmonary lymph nodes: thin-section CT findings, pathological findings, and CT differential diagnosis from pulmonary metastatic nodules
Acta Medica Okayama
(2004)
S. Itoh et al.
Lung: feasibility of a method for changing tube current during low-dose helical CT
Radiology
(2002)

There are more references available in the full text version of this article.

Cited by (43)

On the performance of lung nodule detection, segmentation and classification
2021, Computerized Medical Imaging and Graphics
Citation Excerpt :
The AUC they obtained is up to 0.89. Iwano et al. (2008) designed a simple LDA classifier to differentiate malignant from benign solitary nodules in CT images by using the circularity and the second central moment as nodule features. For a dataset with 55 benign and 52 malignant nodules, they got a sensitivity of 76.9 % and a specificity of 80 %.
Computed tomography (CT) screening is an effective way for early detection of lung cancer in order to improve the survival rate of such a deadly disease. For more than two decades, image processing techniques such as nodule detection, segmentation, and classification have been extensively studied to assist physicians in identifying nodules from hundreds of CT slices to measure shapes and HU distributions of nodules automatically and to distinguish their malignancy. Thanks to new parallel computation, multi-layer convolution, nonlinear pooling operation, and the big data learning strategy, recent development of deep-learning algorithms has shown great progress in lung nodule screening and computer-assisted diagnosis (CADx) applications due to their high sensitivity and low false positive rates. This paper presents a survey of state-of-the-art deep-learning-based lung nodule screening and analysis techniques focusing on their performance and clinical applications, aiming to help better understand the current performance, the limitation, and the future trends of lung nodule analysis.
Quantitative Contour Analysis as an Image-based Discriminator Between Benign and Malignant Renal Tumors
2018, Urology
To investigate whether morphologic analysis can differentiate between benign and malignant renal tumors on clinically acquired imaging.
Between 2009 and 2014, 3-dimensional tumor volumes were manually segmented from contrast-enhanced computerized tomography (CT) images from 150 patients with predominantly solid, nonmacroscopic fat-containing renal tumors: 100 renal cell carcinomas and 50 benign lesions (eg, oncocytoma and lipid-poor angiomyolipoma). Tessellated 3-dimensional tumor models were created from segmented voxels using MATLAB code. Eleven shape descriptors were calculated: sphericity, compactness, mean radial distance, standard deviation of the radial distance, radial distance area ratio, zero crossing, entropy, Feret ratio, convex hull area and convex hull perimeter ratios, and elliptic compactness. Morphometric parameters were compared using the Wilcoxon rank-sum test to investigate whether malignant renal masses demonstrate more morphologic irregularity than benign ones.
Only CHP in sagittal orientation (median 0.96 vs 0.97) and EC in coronal orientation (median 0.92 vs 0.93) differed significantly between malignant and benign masses (P = .04). When comparing these 2 metrics between coronal and sagittal orientations, similar but nonsignificant trends emerged (P = .07). Other metrics tested were not significantly different in any imaging plane.
Computerized image analysis is feasible using shape descriptors that otherwise cannot be visually assessed and used without quantification. Shape analysis via the transverse orientation may be reasonable, but encompassing all 3 planar dimensions to characterize tumor contour can achieve a more comprehensive evaluation. Two shape metrics (CHP and EC) may help distinguish benign from malignant renal tumors, an often challenging goal to achieve on imaging and biopsy.
Vasculature surrounding a nodule: A novel lung cancer biomarker
2017, Lung Cancer
Citation Excerpt :
The high prevalence of indeterminate nodules detected during LDCT screening is one challenging aspect of screening and will need to be addressed as the practice of lung screening grows [4]. Although there has been significantly more research effort in image processing focused on lung nodule detection compared to nodule diagnosis, investigators have worked on nodule diagnosis for more than 20 years using a variety of image processing and evaluation approaches [8–23]. Image features developed to discriminate benign and malignant nodules have included nodule morphology, nodule density, image texture, deep learning, and many others that were used in a variety of classifiers.
To investigate whether the vessels surrounding a nodule depicted on non-contrast, low-dose computed tomography (LDCT) can discriminate benign and malignant screen detected nodules.
We collected a dataset consisting of LDCT scans acquired on 100 subjects from the Pittsburgh Lung Screening study (PLuSS). Fifty subjects were diagnosed with lung cancer and 50 subjects had suspicious nodules later proven benign. For the lung cancer cases, the location of the malignant nodule in the LDCT scans was known; while for the benign cases, the largest nodule in the LDCT scan was used in the analysis. A computer algorithm was developed to identify surrounding vessels and quantify the number and volume of vessels that were connected or near the nodule. A nonparametric receiver operating characteristic (ROC) analysis was performed based on a single nodule per subject to assess the discriminability of the surrounding vessels to provide a lung cancer diagnosis. Odds ratio (OR) were computed to determine the probability of a nodule being lung cancer based on the vessel features.
The areas under the ROC curves (AUCs) for vessel count and vessel volume were 0.722 (95% CI = 0.616-0.811, p < 0.01) and 0.676 (95% CI = 0.565-0.772), respectively. The number of vessels attached to a nodule was significantly higher in the lung cancer group 9.7 (±9.6) compared to the non-lung cancer group 4.0 (±4.3)
Our preliminary results showed that malignant nodules are often surrounded by more vessels compared to benign nodules, suggesting that the surrounding vessel characteristics could serve as lung cancer biomarker for indeterminate nodules detected during LDCT lung cancer screening using only the information collected during the initial visit.
Assessing diagnostic complexity: An image feature-based strategy to reduce annotation costs
2015, Computers in Biology and Medicine
Citation Excerpt :
In the computer-aided diagnosis (CAD) literature, consensus image interpretation is used as a standard of reference to provide the target class label to which the CAD method is compared [26]. For diagnosis of lung nodules, the application domain of this paper, most of the CAD systems use traditional classification techniques such as linear discriminant analysis [27–33], decision trees [34,35], and neural networks [36–39] to learn the class label from nodules׳ appearance, size, and shape image features. The CAD performance is generally evaluated using receiver operator characteristic (ROC) analysis and area under the ROC is used as a performance index [40,41].
Computer-aided diagnosis systems can play an important role in lowering the workload of clinical radiologists and reducing costs by automatically analyzing vast amounts of image data and providing meaningful and timely insights during the decision making process. In this paper, we present strategies on how to better manage the limited time of clinical radiologists in conjunction with predictive model diagnosis. We first introduce a metric for discriminating between the different categories of diagnostic complexity (such as easy versus hard) encountered when interpreting CT scans. Second, we propose to learn the diagnostic complexity using a classification approach based on low-level image features automatically extracted from pixel data. We then show how this classification can be used to decide how to best allocate additional radiologists to interpret a case based on its diagnosis category. Using a lung nodule image dataset, we determined that, by a simple division of cases into hard and easy to diagnose, the number of interpretations can be distributed to significantly lower the cost with limited loss in prediction accuracy. Furthermore, we show that with just a few low-level image features (18% of the original set) we are able to determine the easy from hard cases for a significant subset (66%) of the lung nodule image data.
Clinical Characteristics of Human Pulmonary Dirofilariasis in Japan: An Uncommon Differential Diagnosis of a Solitary Pulmonary Nodule
2023, Japanese Journal of Infectious Diseases
Lung Image Classification Based on Long-Short Term Memory recurrent neural network
2023, Journal of Physics: Conference Series

View all citing articles on Scopus

Tatsuya Nakamura graduated from the Hamamatsu University School of Medicine, in March 1999, with a bachelor of medicine degree. He received a PhD degree in March 2004 from Nagoya University. He is currently on the staff in the Department of Radiation Oncology, Aichi Cancer Center. He is a member of the Japan Radiological Society and the Japanese Society of Therapeutic Radiology and Oncology.

Yuko Kamioka graduated from the Mie University, School of Medicine in March 2000, with a bachelor of medicine degree. At present, she is on the staff of Department of radiology in Anjo Kosei Hospital, Japan. She is a member of the Japan Radiological Society.

Mitsuru Ikeda graduated from the Tokyo Kougyo University in March 1977, with a bachelor of engineering degree. Further, he graduated from the Nagoya University School of Medicine in March 1982, with a bachelor of medicine degree. He received a PhD degree in March 1986 from the Nagoya University. At present, he is an associate professor in the Department of Radiological Technology, Nagoya University School of Health Sciences. Dr. Ikeda is a member of the Radiological Society of North America and the Japan Radiological Society.

Takeo Ishigaki was born in Tokyo on April 17, 1942. He graduated from the University of Tokyo, Faculty of Medicine in 1968. He received a PhD degree in 1976 with research on TV monitor diagnosis for the upper GI tract from the Nagoya City University, Japan. At present, he is a emeritus professor of the Department of Radiology, Nagoya University School of Medicine. He is a member of the Radiological Society of North America. His present research includes digital radiography, PACS, and telemedicine.

View full text

Computer-aided differentiation of malignant from benign solitary pulmonary nodules imaged by high-resolution CT

Abstract

Introduction

Section snippets

Study group

Experimental results

Discussion

Clin Imaging

Comput Med Imaging Graph

Comput Med Imaging Graph

Lung Cancer

Lancet

Solitary pulmonary nodules. Part I. Morphologic evaluation for differentiation of benign and malignant lesions

Radiographics

New classification of small pulmonary nodules by margin characteristics on high-resolution CT

Acta Radiol

Early lung cancer action project: overall design and findings from baseline screening

Cancer

Intrapulmonary lymph nodes: thin-section CT findings, pathological findings, and CT differential diagnosis from pulmonary metastatic nodules

Acta Medica Okayama

Lung: feasibility of a method for changing tube current during low-dose helical CT

Radiology