Automated detection of lung nodules in CT scans: preliminary results

Hierarchical classification of early microscopic lung nodule based on cascade network

Purpose

Early-stage lung cancer is typically characterized clinically by the presence of isolated lung nodules. Thousands of cases are examined each year, and one case usually contains numerous lung CT slices. Detecting and classifying early microscopic lung nodules is demanding due to their diminutive dimensions and restricted characterization capabilities. Therefore, a lung nodule classification model that performs well and is sensitive to microscopic lung nodules is needed to accurately classify lung nodules.

Methods

This paper uses the Resnet34 network as a basic classification model. A new cascade lung nodule classification method is proposed to classify lung nodules into 6 classes instead of the traditional 2 or 4 classes. It can effectively classify six different nodule types including ground-glass and solid nodules, benign and malignant nodules, and nodules with predominantly ground-glass or solid components.

Results

In this paper, the traditional multi-classification method and the cascade classification method proposed in this paper were tested using real lung nodule data collected in the clinic. The test results demonstrate that the cascade classification method in this study achieves an accuracy of 80.04% , outperforming the conventional multi-classification approach.

Conclusions

Different from the existing methods for categorizing the benign and malignant nature of lung nodules, the approach presented in this paper can classify lung nodules into 6 categories more accurately. At the same time, This paper proposes a rapid, precise, and dependable approach for classifying six distinct categories of lung nodules, which increases the accuracy categorization compared with the traditional multivariate categorization method.

Deep Cascade-Learning Model via Recurrent Attention for Immunofixation Electrophoresis Image Analysis

Immunofixation Electrophoresis (IFE) analysis has been an indispensable prerequisite for the diagnosis of M-protein, which is an important criterion to recognize diversified plasma cell diseases. Existing intelligent methods of IFE diagnosis commonly employ a single unified classifier to directly classify whether M-protein exists and which isotype of M-protein is. However, this unified classification is not optimal because the two tasks have different characteristics and require different feature extraction techniques. Classifying the M-protein existence depends on the presence or absence of dense bands in IFE data, while classifying the M-protein isotype depends on the location of dense bands. Consequently, a cascading two-classifier framework suitable to the two tasks respectively may achieve better performance. In this paper, we propose a novel deep cascade-learning model, which sequentially integrates a positive-negative classifier based on deep collocative learning and an isotype classifier based on recurrent attention model to address these two tasks respectively. Specifically, the attention mechanism can mimic the visual perception of clinicians, where only the most informative local regions are extracted through sequential partial observations. This not only avoids the interference of redundant regions but also saves computational power. Further, domain knowledge about SP lane and heavy-light-chain lanes is also introduced to assist our attention location. Extensive numerical experiments show that our deep cascade-learning outperforms state-of-the-art methods on recognized evaluation metrics and can effectively capture the co-location of dense bands in different lanes.

A radiomics approach for lung nodule detection in thoracic CT images based on the dynamic patterns of morphological variation

OBJECTIVES

To propose and evaluate a set of radiomic features, called morphological dynamics features, for pulmonary nodule detection, which were rooted in the dynamic patterns of morphological variation and needless precise lesion segmentation.

MATERIALS AND METHODS

Two datasets were involved, namely, university hospital (UH) and LIDC datasets, comprising 72 CT scans (360 nodules) and 888 CT scans (2230 nodules), respectively. Each nodule was annotated by multiple radiologists. Denoted the category of nodules identified by at least k radiologists as ALk. A nodule detection algorithm, called CAD-MD algorithm, was proposed based on the morphological dynamics radiomic features, characterizing a lesion by ten sets of the same features with different values extracted from ten different thresholding results. Each nodule candidate was classified by a two-level classifier, including ten decision trees and a random forest, respectively. The CAD-MD algorithm was compared with a deep learning approach, the N-Net, using the UH dataset.

RESULTS

On the AL1 and AL2 of the UH dataset, the AUC of the AFROC curves were 0.777 and 0.851 for the CAD-MD algorithm and 0.478 and 0.472 for the N-Net, respectively. The CAD-MD algorithm achieved the sensitivities of 84.4% and 91.4% with 2.98 and 3.69 FPs/scan and the N-Net 74.4% and 80.7% with 3.90 and 4.49 FPs/scan, respectively. On the LIDC dataset, the CAD-MD algorithm attained the sensitivities of 87.6%, 89.2%, 92.2%, and 95.0% with 4 FPs/scan for AL1-AL4, respectively.

CONCLUSION

The morphological dynamics radiomic features might serve as an effective set of radiomic features for lung nodule detection.

KEY POINTS

• Texture features varied with such CT system settings as reconstruction kernels of CT images, CT scanner models, and parameter settings, and so on. • Shape and first-order statistics were shown to be the most robust features against variation in CT imaging parameters. • The morphological dynamics radiomic features, which mainly characterized the dynamic patterns of morphological variation, were shown to be effective for lung nodule detection.

Deep Collocative Learning for Immunofixation Electrophoresis Image Analysis

Immunofixation Electrophoresis (IFE) analysis is of great importance to the diagnosis of Multiple Myeloma, which is among the top-9 cancer killers in the United States, but has rarely been studied in the context of deep learning. Two possible reasons are: 1) the recognition of IFE patterns is dependent on the co-location of bands that forms a binary relation, different from the unary relation (visual features to label) that deep learning is good at modeling; 2) deep classification models may perform with high accuracy for IFE recognition but is not able to provide firm evidence (where the co-location patterns are) for its predictions, rendering difficulty for technicians to validate the results. We propose to address these issues with collocative learning, in which a collocative tensor has been constructed to transform the binary relations into unary relations that are compatible with conventional deep networks, and a location-label-free method that utilizes the Grad-CAM saliency map for evidence backtracking has been proposed for accurate localization. In addition, we have proposed Coached Attention Gates that can regulate the inference of the learning to be more consistent with human logic and thus support the evidence backtracking. The experimental results show that the proposed method has obtained a performance gain over its base model ResNet18 by 741.30% in IoU and also outperformed popular deep networks of DenseNet, CBAM, and Inception-v3.

A practical and adaptive approach to lung cancer screening: a review of international evidence and position on CT lung cancer screening in the Singaporean population by the College of Radiologists Singapore

Lung cancer is the leading cause of cancer-related death around the world, being the top cause of cancer-related deaths among men and the second most common cause of cancer-related deaths among women in Singapore. Currently, no screening programme for lung cancer exists in Singapore. Since there is mounting evidence indicating a different epidemiology of lung cancer in Asian countries, including Singapore, compared to the rest of the world, a unique and adaptive approach must be taken for a screening programme to be successful at reducing mortality while maintaining cost-effectiveness and a favourable risk-benefit ratio. This review article promotes the use of low-dose computed tomography of the chest and explores the radiological challenges and future directions.

Automatic lung nodule detection using multi-scale dot nodule-enhancement filter and weighted support vector machines in chest computed tomography

A novel CAD scheme for automated lung nodule detection is proposed to assist radiologists with the detection of lung cancer on CT scans. The proposed scheme is composed of four major steps: (1) lung volume segmentation, (2) nodule candidate extraction and grouping, (3) false positives reduction for the non-vessel tree group, and (4) classification for the vessel tree group. Lung segmentation is performed first. Then, 3D labeling technology is used to divide nodule candidates into two groups. For the non-vessel tree group, nodule candidates are classified as true nodules at the false positive reduction stage if the candidates survive the rule-based classifier and are not screened out by the dot filter. For the vessel tree group, nodule candidates are extracted using dot filter. Next, RSFS feature selection is used to select the most discriminating features for classification. Finally, WSVM with an undersampling approach is adopted to discriminate true nodules from vessel bifurcations in vessel tree group. The proposed method was evaluated on 154 thin-slice scans with 204 nodules in the LIDC database. The performance of the proposed CAD scheme yielded a high sensitivity (87.81%) while maintaining a low false rate (1.057 FPs/scan). The experimental results indicate the performance of our method may be better than the existing methods.

Automatic lung nodule detection using a 3D deep convolutional neural network combined with a multi-scale prediction strategy in chest CTs

OBJECTIVE

A novel computer-aided detection (CAD) scheme for lung nodule detection using a 3D deep convolutional neural network combined with a multi-scale prediction strategy is proposed to assist radiologists by providing a second opinion on accurate lung nodule detection, which is a crucial step in early diagnosis of lung cancer.

METHOD

A 3D deep convolutional neural network (CNN) with multi-scale prediction was used to detect lung nodules after the lungs were segmented from chest CT scans, with a comprehensive method utilized. Compared with a 2D CNN, a 3D CNN can utilize richer spatial 3D contextual information and generate more discriminative features after being trained with 3D samples to fully represent lung nodules. Furthermore, a multi-scale lung nodule prediction strategy, including multi-scale cube prediction and cube clustering, is also proposed to detect extremely small nodules.

RESULT

The proposed method was evaluated on 888 thin-slice scans with 1186 nodules in the LUNA16 database. All results were obtained via 10-fold cross-validation. Three options of the proposed scheme are provided for selection according to the actual needs. The sensitivity of the proposed scheme with the primary option reached 87.94% and 92.93% at one and four false positives per scan, respectively. Meanwhile, the competition performance metric (CPM) score is very satisfying (0.7967).

CONCLUSION

The experimental results demonstrate the outstanding detection performance of the proposed nodule detection scheme. In addition, the proposed scheme can be extended to other medical image recognition fields.

Virtual monoenergetic imaging in rapid kVp-switching dual-energy CT (DECT) of the abdomen: impact on CT texture analysis

PURPOSE

To study the impact of keV levels of virtual monoenergetic images generated from rapid kVp-switching dual-energy CT (rsDECT) on CT texture analysis (CTTA).

METHODS

This study included 30 consecutive patients (59.3 ± 12 years; range 34-77 years; 17M:13F) who underwent portal venous phase abdominal CT on a rsDECT scanner. Axial 5-mm monoenergetic images at 5 energy levels (40/50/60/70/80 keV) were created and CTTA of liver was performed. CTTA comprised a filtration-histogram technique with different spatial scale filter (SSF) values (0-6). CTTA quantification at each SSF value included histogram-based statistical parameters such as mean intensity, standard deviation (SD), entropy, mean of positive pixels (MPP), skewness, and kurtosis. The values were compared using repeated measures ANOVA.

RESULTS

Among the different CTTA metrics, mean intensity (at SSF > 0), skewness, and kurtosis did not show variability whereas entropy, MPP, and SD varied with different keV levels. There was no change in skewness and kurtosis values for all 6 filters (p > 0.05). Mean intensity showed no change for filters 2-6 (p > 0.05). Mean intensity at SSF = 0 i.e., mean attenuations were 91.2 ± 2.9, 108.7 ± 3.6, 136.1 ± 4.7, 179.8 ± 6.9, and 250.5 ± 10.1 HU for 80, 70, 60, 50, and 40 keV images, respectively demonstrating significant variability (decrease) with increasing keV levels (p < 0.001). Entropy, MPP, and SD values showed a statistically significant decrease with increasing keV of monoenergetic images on all 6 filters (p < 0.001).

CONCLUSION

The energy levels of monoenergetic images have variable impact on the different CTTA parameters, with no significant change in skewness, kurtosis, and filtered mean intensity whereas significant decrease in mean attenuation, entropy, MPP, and SD values with increasing energy levels.

Performance analysis of a computer-aided detection system for lung nodules in CT at different slice thicknesses

We study the performance of a computer-aided detection (CAD) system for lung nodules in computed tomography (CT) as a function of slice thickness. In addition, we propose and compare three different training methodologies for utilizing nonhomogeneous thickness training data (i.e., composed of cases with different slice thicknesses). These methods are (1) aggregate training using the entire suite of data at their native thickness, (2) homogeneous subset training that uses only the subset of training data that matches each testing case, and (3) resampling all training and testing cases to a common thickness. We believe this study has important implications for how CT is acquired, processed, and stored. We make use of 192 CT cases acquired at a thickness of 1.25 mm and 283 cases at 2.5 mm. These data are from the publicly available Lung Nodule Analysis 2016 dataset. In our study, CAD performance at 2.5 mm is comparable with that at 1.25 mm and is much better than at higher thicknesses. Also, resampling all training and testing cases to 2.5 mm provides the best performance among the three training methods compared in terms of accuracy, memory consumption, and computational time.

Overview of deep learning in medical imaging

The use of machine learning (ML) has been increasing rapidly in the medical imaging field, including computer-aided diagnosis (CAD), radiomics, and medical image analysis. Recently, an ML area called deep learning emerged in the computer vision field and became very popular in many fields. It started from an event in late 2012, when a deep-learning approach based on a convolutional neural network (CNN) won an overwhelming victory in the best-known worldwide computer vision competition, ImageNet Classification. Since then, researchers in virtually all fields, including medical imaging, have started actively participating in the explosively growing field of deep learning. In this paper, the area of deep learning in medical imaging is overviewed, including (1) what was changed in machine learning before and after the introduction of deep learning, (2) what is the source of the power of deep learning, (3) two major deep-learning models: a massive-training artificial neural network (MTANN) and a convolutional neural network (CNN), (4) similarities and differences between the two models, and (5) their applications to medical imaging. This review shows that ML with feature input (or feature-based ML) was dominant before the introduction of deep learning, and that the major and essential difference between ML before and after deep learning is the learning of image data directly without object segmentation or feature extraction; thus, it is the source of the power of deep learning, although the depth of the model is an important attribute. The class of ML with image input (or image-based ML) including deep learning has a long history, but recently gained popularity due to the use of the new terminology, deep learning. There are two major models in this class of ML in medical imaging, MTANN and CNN, which have similarities as well as several differences. In our experience, MTANNs were substantially more efficient in their development, had a higher performance, and required a lesser number of training cases than did CNNs. "Deep learning", or ML with image input, in medical imaging is an explosively growing, promising field. It is expected that ML with image input will be the mainstream area in the field of medical imaging in the next few decades.

The impact of trained radiographers as concurrent readers on performance and reading time of experienced radiologists in the UK Lung Cancer Screening (UKLS) trial

Objectives

To compare radiologists’ performance reading CTs independently with their performance using radiographers as concurrent readers in lung cancer screening.

Methods

369 consecutive baseline CTs performed for the UK Lung Cancer Screening (UKLS) trial were double-read by radiologists reading either independently or concurrently with a radiographer. In concurrent reading, the radiologist reviewed radiographer-identified nodules and then detected any additional nodules. Radiologists recorded their independent and concurrent reading times. For each radiologist, sensitivity, average false-positive detections (FPs) per case and mean reading times for each method were calculated.

Results

694 nodules in 246/369 (66.7%) studies comprised the reference standard. Radiologists’ mean sensitivity and average FPs per case both increased with concurrent reading compared to independent reading (90.8 ± 5.6% vs. 77.5 ± 11.2%, and 0.60 ± 0.53 vs. 0.33 ± 0.20, respectively; p < 0.05 for 3/4 and 2/4 radiologists, respectively). The mean reading times per case decreased from 9.1 ± 2.3 min with independent reading to 7.2 ± 1.0 min with concurrent reading, decreasing significantly for 3/4 radiologists (p < 0.05).

Conclusions

The majority of radiologists demonstrated improved sensitivity, a small increase in FP detections and a statistically significantly reduced reading time using radiographers as concurrent readers.

Key Points

• Radiographers as concurrent readers could improve radiologists’ sensitivity in lung nodule detection.

• An increase in false-positive detections with radiographer-assisted concurrent reading occurred.

• The false-positive detection rate was still lower than reported for computer-aided detection.

• Concurrent reading with radiographers was also faster than single reading.

• The time saved per case using concurrently reading radiographers was relatively modest.

Electronic supplementary material

The online version of this article (doi:10.1007/s00330-017-4903-z) contains supplementary material, which is available to authorized users.

Lung Lesion Extraction Using a Toboggan Based Growing Automatic Segmentation Approach

The accurate segmentation of lung lesions from computed tomography (CT) scans is important for lung cancer research and can offer valuable information for clinical diagnosis and treatment. However, it is challenging to achieve a fully automatic lesion detection and segmentation with acceptable accuracy due to the heterogeneity of lung lesions. Here, we propose a novel toboggan based growing automatic segmentation approach (TBGA) with a three-step framework, which are automatic initial seed point selection, multi-constraints 3D lesion extraction and the final lesion refinement. The new approach does not require any human interaction or training dataset for lesion detection, yet it can provide a high lesion detection sensitivity (96.35%) and a comparable segmentation accuracy with manual segmentation (P > 0.05), which was proved by a series assessments using the LIDC-IDRI dataset (850 lesions) and in-house clinical dataset (121 lesions). We also compared TBGA with commonly used level set and skeleton graph cut methods, respectively. The results indicated a significant improvement of segmentation accuracy . Furthermore, the average time consumption for one lesion segmentation was under 8 s using our new method. In conclusion, we believe that the novel TBGA can achieve robust, efficient and accurate lung lesion segmentation in CT images automatically.

Three-dimensional lung nodule segmentation and shape variance analysis to detect lung cancer with reduced false positives

The three-dimensional analysis on lung computed tomography scan was carried out in this study to detect the malignant lung nodules. An automatic three-dimensional segmentation algorithm proposed here efficiently segmented the tissue clusters (nodules) inside the lung. However, an automatic morphological region-grow segmentation algorithm that was implemented to segment the well-circumscribed nodules present inside the lung did not segment the juxta-pleural nodule present on the inner surface of wall of the lung. A novel edge bridge and fill technique is proposed in this article to segment the juxta-pleural and pleural-tail nodules accurately. The centroid shift of each candidate nodule was computed. The nodules with more centroid shift in the consecutive slices were eliminated since malignant nodule’s resultant position did not usually deviate. The three-dimensional shape variation and edge sharp analyses were performed to reduce the false positives and to classify the malignant nodules. The change in area and equivalent diameter was more for malignant nodules in the consecutive slices and the malignant nodules showed a sharp edge. Segmentation was followed by three-dimensional centroid, shape and edge analysis which was carried out on a lung computed tomography database of 20 patient with 25 malignant nodules. The algorithms proposed in this article precisely detected 22 malignant nodules and failed to detect 3 with a sensitivity of 88%. Furthermore, this algorithm correctly eliminated 216 tissue clusters that were initially segmented as nodules; however, 41 non-malignant tissue clusters were detected as malignant nodules. Therefore, the false positive of this algorithm was 2.05 per patient.

Shape and texture based novel features for automated juxtapleural nodule detection in lung CTs

Lung cancer is one of the types of cancer with highest mortality rate in the world. In case of early detection and diagnosis, the survival rate of patients significantly increases. In this study, a novel method and system that provides automatic detection of juxtapleural nodule pattern have been developed from cross-sectional images of lung CT (Computerized Tomography). Shape-based and both shape and texture based 7 features are contributed to the literature for lung nodules. System that we developed consists of six main stages called preprocessing, lung segmentation, detection of nodule candidate regions, feature extraction, feature selection (with five feature ranking criteria) and classification. LIDC dataset containing cross-sectional images of lung CT has been utilized, 1410 nodule candidate regions and 40 features have been extracted from 138 cross-sectional images for 24 patients. Experimental results for 10 classifiers are obtained and presented. Adding our derived features to known 33 features has increased nodule recognition performance from 0.9639 to 0.9679 AUC value on generalized linear model regression (GLMR) for 22 selected features and being reached one of the most successful results in the literature.

Lung nodule classification with multilevel patch-based context analysis

In this paper, we propose a novel classification method for the four types of lung nodules, i.e., well-circumscribed, vascularized, juxta-pleural, and pleural-tail, in low dose computed tomography scans. The proposed method is based on contextual analysis by combining the lung nodule and surrounding anatomical structures, and has three main stages: an adaptive patch-based division is used to construct concentric multilevel partition; then, a new feature set is designed to incorporate intensity, texture, and gradient information for image patch feature description, and then a contextual latent semantic analysis-based classifier is designed to calculate the probabilistic estimations for the relevant images. Our proposed method was evaluated on a publicly available dataset and clearly demonstrated promising classification performance.

Impact of image quality, radiologists, lung segments, and Gunnar eyewear on detectability of lung nodules in chest CT

BACKGROUND

Despite the increasingly higher spatial and contrast resolution of CT, nodular lesions are prone to be missed on chest CT. Tinted lenses increase visual acuity and contrast sensitivity by filtering short wavelength light of solar and artificial origin.

PURPOSE

To test the impact of Gunnar eyewear, image quality (standard versus low dose CT) and nodule location on detectability of lung nodules in CT and to compare their individual influence.

MATERIAL AND METHODS

A pre-existing database of CT images of patients with lung nodules >5 mm, scanned with standard does image quality (150 ref mAs/120 kVp) and lower dose/quality (40 ref mAs/120 kVp), was used. Five radiologists read 60 chest CTs twice: once with Gunnar glasses and once without glasses with a 1 month break between. At both read-outs the cases were shown at lower dose or standard dose level to quantify the influence of both variables (eyewear vs. image quality) on nodule sensitivity.

RESULTS

The sensitivity of CT for lung nodules increased significantly using Gunnar eyewear for two readers and insignificantly for two other readers. Over all, the mean sensitivity of all radiologist raised significantly from 50% to 53%, using the glasses (P value = 0.034). In contrast, sensitivity for lung nodules was not significantly affected by lowering the image quality from 150 to 40 ref mAs. The average sensitivity was 52% at low dose level, that was even 0.7% higher than at standard dose level (P value = 0.40). The strongest impact on sensitivity had the factors readers and nodule location (lung segments).

CONCLUSION

Sensitivity for lung nodules was significantly enhanced by Gunnar eyewear (+3%), while lower image quality (40 ref mAs) had no impact on nodule sensitivity. Not using the glasses had a bigger impact on sensitivity than lowering the image quality.

Machine Learning in Computer-aided Diagnosis of the Thorax and Colon in CT: A Survey

Computer-aided detection (CADe) and diagnosis (CAD) has been a rapidly growing, active area of research in medical imaging. Machine leaning (ML) plays an essential role in CAD, because objects such as lesions and organs may not be represented accurately by a simple equation; thus, medical pattern recognition essentially require "learning from examples." One of the most popular uses of ML is the classification of objects such as lesion candidates into certain classes (e.g., abnormal or normal, and lesions or non-lesions) based on input features (e.g., contrast and area) obtained from segmented lesion candidates. The task of ML is to determine "optimal" boundaries for separating classes in the multidimensional feature space which is formed by the input features. ML algorithms for classification include linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), multilayer perceptrons, and support vector machines (SVM). Recently, pixel/voxel-based ML (PML) emerged in medical image processing/analysis, which uses pixel/voxel values in images directly, instead of features calculated from segmented lesions, as input information; thus, feature calculation or segmentation is not required. In this paper, ML techniques used in CAD schemes for detection and diagnosis of lung nodules in thoracic CT and for detection of polyps in CT colonography (CTC) are surveyed and reviewed.

Quantifying tumour heterogeneity with CT

Heterogeneity is a key feature of malignancy associated with adverse tumour biology. Quantifying heterogeneity could provide a useful non-invasive imaging biomarker. Heterogeneity on computed tomography (CT) can be quantified using texture analysis which extracts spatial information from CT images (unenhanced, contrast-enhanced and derived images such as CT perfusion) that may not be perceptible to the naked eye. The main components of texture analysis can be categorized into image transformation and quantification. Image transformation filters the conventional image into its basic components (spatial, frequency, etc.) to produce derived subimages. Texture quantification techniques include structural-, model- (fractal dimensions), statistical- and frequency-based methods. The underlying tumour biology that CT texture analysis may reflect includes (but is not limited to) tumour hypoxia and angiogenesis. Emerging studies show that CT texture analysis has the potential to be a useful adjunct in clinical oncologic imaging, providing important information about tumour characterization, prognosis and treatment prediction and response.

Multiple diagnostic task performance in CT examination of the chest

Objectives In three experiments, we studied the detection of multiple abnormality types using the satisfaction of search (SOS) paradigm, the provision of a computer-aided detection (CAD) of pulmonary nodules and a focused nodule detection task. Methods 51 chest CT examinations (24 that demonstrated subtle pulmonary nodules and 27 that demonstrated no pulmonary nodules) were read by 15 radiology residents and fellows under two experimental conditions: (1) when there were no other abnormalities present except test abnormalities in the exams (non-SOS condition), and (2) when other abnormalities were present in the exams (SOS condition). Trials from the two conditions were intermixed. Readers were invited to return for two sessions: one in which the SOS condition was repeated with a simulated CAD; another in which only the non-SOS condition was presented. Detection accuracy was measured using receiver operating characteristic (ROC) analysis. Results An SOS effect (reduced detection accuracy for the test nodules in the presence of the diverse added abnormalities) was not found. Average accuracy was much higher when the CAD prompt was provided, without cost in the detection of the added abnormalities. Accuracy for detecting nodules appearing without intermixed SOS trials was also substantially improved. Conclusions CT interpretation was highly task dependent. Nodule detection was poor in the general search task. Therefore, CAD may offer a greater performance improvement than demonstrated in experiments assessing CAD using focused search. The absence of SOS may be due to limited nodule detection even without other abnormalities. Advances in knowledge CAD prompts of nodules increase the detection accuracy of nodules and decrease the time to detection-without impairing the detection accuracy-of non-nodule abnormalities.

Computer-aided diagnosis systems for lung cancer: challenges and methodologies

This paper overviews one of the most important, interesting, and challenging problems in oncology, the problem of lung cancer diagnosis. Developing an effective computer-aided diagnosis (CAD) system for lung cancer is of great clinical importance and can increase the patient's chance of survival. For this reason, CAD systems for lung cancer have been investigated in a huge number of research studies. A typical CAD system for lung cancer diagnosis is composed of four main processing steps: segmentation of the lung fields, detection of nodules inside the lung fields, segmentation of the detected nodules, and diagnosis of the nodules as benign or malignant. This paper overviews the current state-of-the-art techniques that have been developed to implement each of these CAD processing steps. For each technique, various aspects of technical issues, implemented methodologies, training and testing databases, and validation methods, as well as achieved performances, are described. In addition, the paper addresses several challenges that researchers face in each implementation step and outlines the strengths and drawbacks of the existing approaches for lung cancer CAD systems.

A review of computer-aided diagnosis in thoracic and colonic imaging

Medical imaging has been indispensable in medicine since the discovery of x-rays. Medical imaging offers useful information on patients' medical conditions and on the causes of their symptoms and diseases. As imaging technologies advance, a large number of medical images are produced which physicians/radiologists must interpret. Thus, computer aids are demanded and become indispensable in physicians' decision making based on medical images. Consequently, computer-aided detection and diagnosis (CAD) has been investigated and has been an active research area in medical imaging. CAD is defined as detection and/or diagnosis made by a radiologist/physician who takes into account the computer output as a "second opinion". In CAD research, detection and diagnosis of lung and colorectal cancer in thoracic and colonic imaging constitute major areas, because lung and colorectal cancers are the leading and second leading causes, respectively, of cancer deaths in the U.S. and also in other countries. In this review, CAD of the thorax and colon, including CAD for detection and diagnosis of lung nodules in thoracic CT, and that for detection of polyps in CT colonography, are reviewed.

A semiautomatic CT-based ensemble segmentation of lung tumors: comparison with oncologists' delineations and with the surgical specimen

PURPOSE

To assess the clinical relevance of a semiautomatic CT-based ensemble segmentation method, by comparing it to pathology and to CT/PET manual delineations by five independent radiation oncologists in non-small cell lung cancer (NSCLC).

MATERIALS AND METHODS

For 20 NSCLC patients (stages Ib-IIIb) the primary tumor was delineated manually on CT/PET scans by five independent radiation oncologists and segmented using a CT based semi-automatic tool. Tumor volume and overlap fractions between manual and semiautomatic-segmented volumes were compared. All measurements were correlated with the maximal diameter on macroscopic examination of the surgical specimen. Imaging data are available on www.cancerdata.org.

RESULTS

High overlap fractions were observed between the semi-automatically segmented volumes and the intersection (92.5±9.0, mean±SD) and union (94.2±6.8) of the manual delineations. No statistically significant differences in tumor volume were observed between the semiautomatic segmentation (71.4±83.2 cm(3), mean±SD) and manual delineations (81.9±94.1 cm(3); p=0.57). The maximal tumor diameter of the semiautomatic-segmented tumor correlated strongly with the macroscopic diameter of the primary tumor (r=0.96).

CONCLUSIONS

Semiautomatic segmentation of the primary tumor on CT demonstrated high agreement with CT/PET manual delineations and strongly correlated with the macroscopic diameter considered as the "gold standard". This method may be used routinely in clinical practice and could be employed as a starting point for treatment planning, target definition in multi-center clinical trials or for high throughput data mining research. This method is particularly suitable for peripherally located tumors.

Automatic detection of lung nodules in CT datasets based on stable 3D mass-spring models

We propose a computer-aided detection (CAD) system which can detect small-sized (from 3mm) pulmonary nodules in spiral CT scans. A pulmonary nodule is a small lesion in the lungs, round-shaped (parenchymal nodule) or worm-shaped (juxtapleural nodule). Both kinds of lesions have a radio-density greater than lung parenchyma, thus appearing white on the images. Lung nodules might indicate a lung cancer and their early stage detection arguably improves the patient survival rate. CT is considered to be the most accurate imaging modality for nodule detection. However, the large amount of data per examination makes the full analysis difficult, leading to omission of nodules by the radiologist. We developed an advanced computerized method for the automatic detection of internal and juxtapleural nodules on low-dose and thin-slice lung CT scan. This method consists of an initial selection of nodule candidates list, the segmentation of each candidate nodule and the classification of the features computed for each segmented nodule candidate.The presented CAD system is aimed to reduce the number of omissions and to decrease the radiologist scan examination time. Our system locates with the same scheme both internal and juxtapleural nodules. For a correct volume segmentation of the lung parenchyma, the system uses a Region Growing (RG) algorithm and an opening process for including the juxtapleural nodules. The segmentation and the extraction of the suspected nodular lesions from CT images by a lung CAD system constitutes a hard task. In order to solve this key problem, we use a new Stable 3D Mass-Spring Model (MSM) combined with a spline curves reconstruction process. Our model represents concurrently the characteristic gray value range, the directed contour information as well as shape knowledge, which leads to a much more robust and efficient segmentation process. For distinguishing the real nodules among nodule candidates, an additional classification step is applied; furthermore, a neural network is applied to reduce the false positives (FPs) after a double-threshold cut. The system performance was tested on a set of 84 scans made available by the Lung Image Database Consortium (LIDC) annotated by four expert radiologists. The detection rate of the system is 97% with 6.1 FPs/CT. A reduction to 2.5 FPs/CT is achieved at 88% sensitivity. We presented a new 3D segmentation technique for lung nodules in CT datasets, using deformable MSMs. The result is a efficient segmentation process able to converge, identifying the shape of the generic ROI, after a few iterations. Our suitable results show that the use of the 3D AC model and the feature analysis based FPs reduction process constitutes an accurate approach to the segmentation and the classification of lung nodules.

Multiple diagnostic task performance in CT examination of the chest

OBJECTIVES

In three experiments, we studied the detection of multiple abnormality types using the satisfaction of search (SOS) paradigm, the provision of a computer-aided detection (CAD) of pulmonary nodules and a focused nodule detection task.

METHODS

51 chest CT examinations (24 that demonstrated subtle pulmonary nodules and 27 that demonstrated no pulmonary nodules) were read by 15 radiology residents and fellows under two experimental conditions: (1) when there were no other abnormalities present except test abnormalities in the exams (non-SOS condition), and (2) when other abnormalities were present in the exams (SOS condition). Trials from the two conditions were intermixed. Readers were invited to return for two sessions: one in which the SOS condition was repeated with a simulated CAD; another in which only the non-SOS condition was presented. Detection accuracy was measured using receiver operating characteristic (ROC) analysis.

RESULTS

An SOS effect (reduced detection accuracy for the test nodules in the presence of the diverse added abnormalities) was not found. Average accuracy was much higher when the CAD prompt was provided, without cost in the detection of the added abnormalities. Accuracy for detecting nodules appearing without intermixed SOS trials was also substantially improved.

CONCLUSIONS

CT interpretation was highly task dependent. Nodule detection was poor in the general search task. Therefore, CAD may offer a greater performance improvement than demonstrated in experiments assessing CAD using focused search. The absence of SOS may be due to limited nodule detection even without other abnormalities. Advances in knowledge CAD prompts of nodules increase the detection accuracy of nodules and decrease the time to detection-without impairing the detection accuracy-of non-nodule abnormalities.

Potential of computer-aided diagnosis to improve CT lung cancer screening

The development of low-dose spiral computed tomography (CT) has rekindled hope that effective lung cancer screening might yet be found. Screening is justified when there is evidence that it will extend lives at reasonable cost and acceptable levels of risk. A screening test should detect all extant cancers while avoiding unnecessary workups. Thus optimal screening modalities have both high sensitivity and specificity. Due to the present state of technology, radiologists must opt to increase sensitivity and rely on follow-up diagnostic procedures to rule out the incurred false positives. There is evidence in published reports that computer-aided diagnosis technology may help radiologists alter the benefit-cost calculus of CT sensitivity and specificity in lung cancer screening protocols. This review will provide insight into the current discussion of the effectiveness of lung cancer screening and assesses the potential of state-of-the-art computer-aided design developments.

Pixel-based machine learning in medical imaging

Machine learning (ML) plays an important role in the medical imaging field, including medical image analysis and computer-aided diagnosis, because objects such as lesions and organs may not be represented accurately by a simple equation; thus, medical pattern recognition essentially require “learning from examples.” One of the most popular uses of ML is classification of objects such as lesions into certain classes (e.g., abnormal or normal, or lesions or nonlesions) based on input features (e.g., contrast and circularity) obtained from segmented object candidates. Recently, pixel/voxel-based ML (PML) emerged in medical image processing/analysis, which use pixel/voxel values in images directly instead of features calculated from segmented objects as input information; thus, feature calculation or segmentation is not required. Because the PML can avoid errors caused by inaccurate feature calculation and segmentation which often occur for subtle or complex objects, the performance of the PML can potentially be higher for such objects than that of common classifiers (i.e., feature-based MLs). In this paper, PMLs are surveyed to make clear (a) classes of PMLs, (b) similarities and differences within (among) different PMLs and those between PMLs and feature-based MLs, (c) advantages and limitations of PMLs, and (d) their applications in medical imaging.

Usefulness of the CAD system for detecting pulmonary nodule in real clinical practice

Objective

We wanted to evaluate the usefulness of the computer-aided detection (CAD) system for detecting pulmonary nodules in real clinical practice by using the CT images.

Materials and Methods

Our Institutional Review Board approved our retrospective study with a waiver of informed consent. This study included 166 CT examinations that were performed for the evaluation of pulmonary metastasis in 166 patients with colorectal cancer. All the CT examinations were interpreted by radiologists and they were also evaluated by the CAD system. All the nodules detected by the CAD system were evaluated with regard to whether or not they were true nodules, and they were classified into micronodules (MN, diameter < 4 mm) and significant nodules (SN, 4 ≤ diameter ≤ 10 mm). The radiologic reports and CAD results were compared.

Results

The CAD system helped detect 426 nodules; 115 (27%) of the 426 nodules were classified as true nodules and 35 (30%) of the 115 nodules were SNs, and 83 (72%) of the 115 were not mentioned in the radiologists' reports and three (4%) of the 83 nodules were non-calcified SNs. One of three non-calcified SNs was confirmed as a metastatic nodule. According to the radiologists' reports, 60 true nodules were detected, and 28 of the 60 were not detected by the CAD system.

Conclusion

Although the CAD system missed many SNs that are detected by radiologists, it helps detect additional nodules that are missed by the radiologists in real clinical practice. Therefore, the CAD system can be useful to support a radiologist's detection performance.

Texture analysis of non-small cell lung cancer on unenhanced computed tomography: initial evidence for a relationship with tumour glucose metabolism and stage

The aim was to undertake an initial study of the relationship between texture features in computed tomography (CT) images of non-small cell lung cancer (NSCLC) and tumour glucose metabolism and stage. This retrospective pilot study comprised 17 patients with 18 pathologically confirmed NSCLC. Non-contrast-enhanced CT images of the primary pulmonary lesions underwent texture analysis in 2 stages as follows: (a) image filtration using Laplacian of Gaussian filter to differentially highlight fine to coarse textures, followed by (b) texture quantification using mean grey intensity (MGI), entropy (E) and uniformity (U) parameters. Texture parameters were compared with tumour fluorodeoxyglucose (FDG) uptake (standardised uptake value (SUV)) and stage as determined by the clinical report of the CT and FDG-positron emission tomography imaging. Tumour SUVs ranged between 2.8 and 10.4. The number of NSCLC with tumour stages I, II, III and IV were 4, 4, 4 and 6, respectively. Coarse texture features correlated with tumour SUV (E: r = 0.51, p = 0.03; U: r = −0.52, p = 0.03), whereas fine texture features correlated with tumour stage (MGI: r_s = 0.71, p = 0.001; E: r_s = 0.55, p = 0.02; U: r_s = −0.49, p = 0.04). Fine texture predicted tumour stage with a kappa of 0.7, demonstrating 100% sensitivity and 87.5% specificity for detecting tumours above stage II ( p = 0.0001). This study provides initial evidence for a relationship between texture features in NSCLC on non-contrast-enhanced CT and tumour metabolism and stage. Texture analysis warrants further investigation as a potential method for obtaining prognostic information for patients with NSCLC undergoing CT.

Comparing and combining algorithms for computer-aided detection of pulmonary nodules in computed tomography scans: The ANODE09 study

Numerous publications and commercial systems are available that deal with automatic detection of pulmonary nodules in thoracic computed tomography scans, but a comparative study where many systems are applied to the same data set has not yet been performed. This paper introduces ANODE09 ( http://anode09.isi.uu.nl), a database of 55 scans from a lung cancer screening program and a web-based framework for objective evaluation of nodule detection algorithms. Any team can upload results to facilitate benchmarking. The performance of six algorithms for which results are available are compared; five from academic groups and one commercially available system. A method to combine the output of multiple systems is proposed. Results show a substantial performance difference between algorithms, and demonstrate that combining the output of algorithms leads to marked performance improvements.

Methodology for automatic detection of lung nodules in computerized tomography images

Lung cancer is a disease with significant prevalence in several countries around the world. Its difficult treatment and rapid progression make the mortality rates among people affected by this illness to be very high. Aiming to offer a computational alternative for helping in detection of nodules, serving as a second opinion to the specialists, this work proposes a totally automatic methodology based on successive detection refining stages. The automated lung nodules detection scheme consists of six stages: thorax extraction, lung extraction, lung reconstruction, structures extraction, tubular structures elimination, and false positive reduction. In the thorax extraction stage all the artifacts external to the patient's body are discarded. Lung extraction stage is responsible for the identification of the lung parenchyma. The objective of the lung reconstruction stage is to prevent incorrect elimination of portions belonging to the parenchyma. Structures extraction stage comprises the selection of dense structures from inside the lung parenchyma. The next stage, tubular structures elimination eliminates a great part of the pulmonary trees. Finally, the false positive stage selects only structures with great probability to be nodule. Each of the several stages has very specific objectives in detection of particular cases of lung nodules, ensuring good matching rates even in difficult detection situations. We use 33 exams with diversified diagnosis and slices numbers for validating the methodology. We obtained a false positive per exam rate of 0.42 and false negative rate of 0.15. The total classification sensitivity obtained, measured out of the nodule candidates, was 84.84%. The specificity achieved was 96.15% and the total accuracy of the method was 95.21%.

A new computationally efficient CAD system for pulmonary nodule detection in CT imagery

Early detection of lung nodules is extremely important for the diagnosis and clinical management of lung cancer. In this paper, a novel computer aided detection (CAD) system for the detection of pulmonary nodules in thoracic computed tomography (CT) imagery is presented. The paper describes the architecture of the CAD system and assesses its performance on a publicly available database to serve as a benchmark for future research efforts. Training and tuning of all modules in our CAD system is done using a separate and independent dataset provided courtesy of the University of Texas Medical Branch (UTMB). The publicly available testing dataset is that created by the Lung Image Database Consortium (LIDC). The LIDC data used here is comprised of 84 CT scans containing 143 nodules ranging from 3 to 30mm in effective size that are manually segmented at least by one of the four radiologists. The CAD system uses a fully automated lung segmentation algorithm to define the boundaries of the lung regions. It combines intensity thresholding with morphological processing to detect and segment nodule candidates simultaneously. A set of 245 features is computed for each segmented nodule candidate. A sequential forward selection process is used to determine the optimum subset of features for two distinct classifiers, a Fisher Linear Discriminant (FLD) classifier and a quadratic classifier. A performance comparison between the two classifiers is presented, and based on this, the FLD classifier is selected for the CAD system. With an average of 517.5 nodule candidates per case/scan (517.5+/-72.9), the proposed front-end detector/segmentor is able to detect 92.8% of all the nodules in the LIDC/testing dataset (based on merged ground truth). The mean overlap between the nodule regions delineated by three or more radiologists and the ones segmented by the proposed segmentation algorithm is approximately 63%. Overall, with a specificity of 3 false positives (FPs) per case/patient on average, the CAD system is able to correctly identify 80.4% of the nodules (115/143) using 40 selected features. A 7-fold cross-validation performance analysis using the LIDC database only shows CAD sensitivity of 82.66% with an average of 3 FPs per CT scan/case.

Automatic lung segmentation in CT images with accurate handling of the hilar region

A fully automated and three-dimensional (3D) segmentation method for the identification of the pulmonary parenchyma in thorax X-ray computed tomography (CT) datasets is proposed. It is meant to be used as pre-processing step in the computer-assisted detection (CAD) system for malignant lung nodule detection that is being developed by the Medical Applications in a Grid Infrastructure Connection (MAGIC-5) Project. In this new approach the segmentation of the external airways (trachea and bronchi), is obtained by 3D region growing with wavefront simulation and suitable stop conditions, thus allowing an accurate handling of the hilar region, notoriously difficult to be segmented. Particular attention was also devoted to checking and solving the problem of the apparent 'fusion' between the lungs, caused by partial-volume effects, while 3D morphology operations ensure the accurate inclusion of all the nodules (internal, pleural, and vascular) in the segmented volume. The new algorithm was initially developed and tested on a dataset of 130 CT scans from the Italung-CT trial, and was then applied to the ANODE09-competition images (55 scans) and to the LIDC database (84 scans), giving very satisfactory results. In particular, the lung contour was adequately located in 96% of the CT scans, with incorrect segmentation of the external airways in the remaining cases. Segmentation metrics were calculated that quantitatively express the consistency between automatic and manual segmentations: the mean overlap degree of the segmentation masks is 0.96 ± 0.02, and the mean and the maximum distance between the mask borders (averaged on the whole dataset) are 0.74 ± 0.05 and 4.5 ± 1.5, respectively, which confirms that the automatic segmentations quite correctly reproduce the borders traced by the radiologist. Moreover, no tissue containing internal and pleural nodules was removed in the segmentation process, so that this method proved to be fit for the use in the framework of a CAD system. Finally, in the comparison with a two-dimensional segmentation procedure, inter-slice smoothness was calculated, showing that the masks created by the 3D algorithm are significantly smoother than those calculated by the 2D-only procedure.

Automated segmentation of lungs with severe interstitial lung disease in CT

PURPOSE

Accurate segmentation of lungs with severe interstitial lung disease (ILD) in thoracic computed tomography (CT) is an important and difficult task in the development of computer-aided diagnosis (CAD) systems. Therefore, we developed in this study a texture analysis-based method for accurate segmentation of lungs with severe ILD in multidetector CT scans.

METHODS

Our database consisted of 76 CT scans, including 31 normal cases and 45 abnormal cases with moderate or severe ILD. The lungs in three selected slices for each CT scan were first manually delineated by a medical physicist, and then confirmed or revised by an expert chest radiologist, and they were used as the reference standard for lung segmentation. To segment the lungs, we first employed a CT value thresholding technique to obtain an initial lung estimate, including normal and mild ILD lung parenchyma. We then used texture-feature images derived from the co-occurrence matrix to further identify abnormal lung regions with severe ILD. Finally, we combined the identified abnormal lung regions with the initial lungs to generate the final lung segmentation result. The overlap rate, volume agreement, mean absolute distance (MAD), and maximum absolute distance (dmax) between the automatically segmented lungs and the reference lungs were employed to evaluate the performance of the segmentation method.

RESULTS

Our segmentation method achieved a mean overlap rate of 96.7%, a mean volume agreement of 98.5%, a mean MAD of 0.84 mm, and a mean dmax of 10.84 mm for all the cases in our database; a mean overlap rate of 97.7%, a mean volume agreement of 99.0%, a mean MAD of 0.66 mm, and a mean dmax of 9.59 mm for the 31 normal cases; and a mean overlap rate of 96.1%, a mean volume agreement of 98.1%, a mean MAD of 0.96 mm, and a mean dmax of 11.71 mm for the 45 abnormal cases with ILD.

CONCLUSIONS

Our lung segmentation method provided accurate segmentation results for abnormal CT scans with severe ILD and would be useful for developing CAD systems for quantification, detection, and diagnosis of ILD.

A novel multithreshold method for nodule detection in lung CT

Multislice computed tomography (MSCT) is a valuable tool for lung cancer detection, thanks to its ability to identify noncalcified nodules of small size (from about 3 mm). Due to the large number of images generated by MSCT, there is much interest in developing computer-aided detection (CAD) systems that could assist radiologists in the lung nodule detection task. A complete multistage CAD system, including lung boundary segmentation, regions of interest (ROIs) selection, feature extraction, and false positive reduction is presented. The selection of ROIs is based on a multithreshold surface-triangulation approach. Surface triangulation is performed at different threshold values, varying from a minimum to a maximum value in a wide range. At a given threshold value, a ROI is defined as the volume inside a connected component of the triangulated isosurface. The evolution of a ROI as a function of the threshold can be represented by a treelike structure. A multithreshold ROI is defined as a path on this tree, which starts from a terminal ROI and ends on the root ROI. For each ROI, the volume, surface area, roundness, density, and moments of inertia are computed as functions of the threshold and used as input to a classification system based on artificial neural networks. The method is suitable to detect different types of nodules, including juxta-pleural nodules and nodules connected to blood vessels. A training set of 109 low-dose MSCT scans made available by the Pisa center of the Italung-CT trial and annotated by expert radiologists was used for the algorithm design and optimization. The system performance was tested on an independent set of 23 low-dose MSCT scans coming from the Pisa Italung-CT center and on 83 scans made available by the Lung Image Database Consortium (LIDC) annotated by four expert radiologists. On the Italung-CT test set, for nodules having a diameter greater than or equal to 3 mm, the system achieved 84% and 71% sensitivity at false positive/scan rates of 10 and 4, respectively. For nodules having a diameter greater than or equal to 4 mm, the sensitivities were 97% and 80% at false positive/scan rates of 10 and 4, respectively. On the LIDC data set, the system achieved a 79% sensitivity at a false positive/scan rate of 4 in the detection of nodules with a diameter greater than or equal to 3 mm that have been annotated by all four radiologists.

A supervised 'lesion-enhancement' filter by use of a massive-training artificial neural network (MTANN) in computer-aided diagnosis (CAD)

Computer-aided diagnosis (CAD) has been an active area of study in medical image analysis. A filter for the enhancement of lesions plays an important role for improving the sensitivity and specificity in CAD schemes. The filter enhances objects similar to a model employed in the filter; e.g. a blob-enhancement filter based on the Hessian matrix enhances sphere-like objects. Actual lesions, however, often differ from a simple model; e.g. a lung nodule is generally modeled as a solid sphere, but there are nodules of various shapes and with internal inhomogeneities such as a nodule with spiculations and ground-glass opacity. Thus, conventional filters often fail to enhance actual lesions. Our purpose in this study was to develop a supervised filter for the enhancement of actual lesions (as opposed to a lesion model) by use of a massive-training artificial neural network (MTANN) in a CAD scheme for detection of lung nodules in CT. The MTANN filter was trained with actual nodules in CT images to enhance actual patterns of nodules. By use of the MTANN filter, the sensitivity and specificity of our CAD scheme were improved substantially. With a database of 69 lung cancers, nodule candidate detection by the MTANN filter achieved a 97% sensitivity with 6.7 false positives (FPs) per section, whereas nodule candidate detection by a difference-image technique achieved a 96% sensitivity with 19.3 FPs per section. Classification-MTANNs were applied for further reduction of the FPs. The classification-MTANNs removed 60% of the FPs with a loss of one true positive; thus, it achieved a 96% sensitivity with 2.7 FPs per section. Overall, with our CAD scheme based on the MTANN filter and classification-MTANNs, an 84% sensitivity with 0.5 FPs per section was achieved.

Commercially available computer-aided detection system for pulmonary nodules on thin-section images using 64 detectors-row CT: preliminary study of 48 cases

RATIONALE AND OBJECTIVES

Most studies of computer-aided detection (CAD) for pulmonary nodules have focused on solid nodule detection. The aim of this study was to evaluate the performance of a commercially available CAD system in the detection of pulmonary nodules with or without ground-glass opacity (GGO) using 64-detector-row computed tomography compared to visual interpretation.

MATERIALS AND METHODS

Computed tomographic examinations were performed on 48 patients with existing or suspicious pulmonary nodules on chest radiography. Three radiologists independently reported the location and pattern (GGO, solid, or part solid) of each nodule candidate on computed tomographic scans, assigned each a confidence score, and then analyzed all scans using the CAD system. A reference standard was established by a consensus panel of different radiologists, who found 229 noncalcified nodules with diameters > or = 4 mm. True-positive and false-positive results and confidence levels were used to generate jackknife alternative free-response receiver-operating characteristic plots.

RESULTS

The sensitivity of GGO for 3 radiologists (60%-80%) was significantly higher than that for the CAD system (21%) (McNemar's test, P < .0001). For overall and solid nodules, the figure-of-merit values without and with the CAD system were significantly different (P = .005-.04) on jackknife alternative free-response receiver-operating characteristic analysis. For GGO and part-solid nodules, the figure-of-merit values with the CAD system were greater than those without the CAD system, indicating no significant differences.

CONCLUSION

Radiologists are significantly superior to this CAD system in the detection of GGO, but the CAD system can still play a complementary role in detecting nodules with or without GGO.

Shape-based computer-aided detection of lung nodules in thoracic CT images

In this paper, a new computer tomography (CT) lung nodule computer-aided detection (CAD) method is proposed for detecting both solid nodules and ground-glass opacity (GGO) nodules (part solid and nonsolid). This method consists of several steps. First, the lung region is segmented from the CT data using a fuzzy thresholding method. Then, the volumetric shape index map, which is based on local Gaussian and mean curvatures, and the "dot" map, which is based on the eigenvalues of a Hessian matrix, are calculated for each voxel within the lungs to enhance objects of a specific shape with high spherical elements (such as nodule objects). The combination of the shape index (local shape information) and "dot" features (local intensity dispersion information) provides a good structure descriptor for the initial nodule candidates generation. Antigeometric diffusion, which diffuses across the image edges, is used as a preprocessing step. The smoothness of image edges enables the accurate calculation of voxel-based geometric features. Adaptive thresholding and modified expectation-maximization methods are employed to segment potential nodule objects. Rule-based filtering is first used to remove easily dismissible nonnodule objects. This is followed by a weighted support vector machine (SVM) classification to further reduce the number of false positive (FP) objects. The proposed method has been trained and validated on a clinical dataset of 108 thoracic CT scans using a wide range of tube dose levels that contain 220 nodules (185 solid nodules and 35 GGO nodules) determined by a ground truth reading process. The data were randomly split into training and testing datasets. The experimental results using the independent dataset indicate an average detection rate of 90.2%, with approximately 8.2 FP/scan. Some challenging nodules such as nonspherical nodules and low-contrast part-solid and nonsolid nodules were identified, while most tissues such as blood vessels were excluded. The method's high detection rate, fast computation, and applicability to different imaging conditions and nodule types shows much promise for clinical applications.

[Automatic detection of pulmonary nodules on computed tomography: a preliminary study]

OBJECTIVES

Recent years have seen growing interest in the development of algorithms for computer-assisted diagnosis (CAD) for the detection of pulmonary nodules on both plain-film radiographs and computed tomography (CT) studies. The purpose of CAD algorithms in this context is to alert radiologists to suspicious radioopacities that might represent cancer in the images. We are developing a CAD system for the detection of pulmonary nodules on helical CT images.

MATERIAL AND METHODS

We collected cases of patients with pulmonary nodules examined with helical CT. A total of 64 nodules, including both calcified and noncalcified lesions, ranging from 3 to 30 mm in diameter were included in the study. Studies were acquired on one 4-slice and one 64-slice CT scanners. Three chest radiologists at two institutions interpreted the studies to determine whether pulmonary nodules were present. We calculated the sensitivity and the number of false positives per image to evaluate the CAD system.

RESULTS

We have developed and evaluated an algorithm for the automatic detection of pulmonary nodules on CT images. For a sensitivity of 76%, the false-positive rate was 1.3 per image.

CONCLUSIONS

Our preliminary results suggest that the system might be useful for radiologists in the detection of pulmonary nodules on helical CT images.