Accurate Lungs Segmentation on CT Chest Images by Adaptive Appearance-Guided Shape Modeling

Soliman, Ahmed; Khalifa, Fahmi; Elnakib, Ahmed; El-Ghar, Mohamed Abou; Dunlap, Neal E.; Wang, Brian; Gimel’farb, Georgy; Keynton, Robert; El-Baz, Ayman

doi:10.1109/tmi.2016.2606370

Cited by 88 publications

(62 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Therefore, many attempts have been made to develop computer-aided diagnosis (CAD) systems for automatic discrimination. [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] Conventional CAD systems first use classical image processing techniques, such as morphologic operators, 3-5 region growing, 6 energy optimization, 7,8 and statistical learning, 9,10 to segment a region of interest (ROI) that includes the nodule. Then, handcrafted features are extracted from the ROI, which are then fed to a classifier for nodule classification.…”

Section: Introductionmentioning

confidence: 99%

Feature‐shared adaptive‐boost deep learning for invasiveness classification of pulmonary subsolid nodules in CT images

Wang

Chen

et al. 2020

Medical Physics

View full text Add to dashboard Cite

Purpose: In clinical practice, invasiveness is an important reference indicator for differentiating the malignant degree of subsolid pulmonary nodules. These nodules can be classified as atypical adenomatous hyperplasia (AAH), adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), or invasive adenocarcinoma (IAC). The automatic determination of a nodule's invasiveness based on chest CT scans can guide treatment planning. However, it is challenging, owing to the insufficiency of training data and their interclass similarity and intraclass variation. To address these challenges, we propose a two-stage deep learning strategy for this task: prior-feature learning followed by adaptive-boost deep learning. Methods: The adaptive-boost deep learning is proposed to train a strong classifier for invasiveness classification of subsolid nodules in chest CT images, using multiple 3D convolutional neural network (CNN)-based weak classifiers. Because ensembles of multiple deep 3D CNN models have a huge number of parameters and require large computing resources along with more training and testing time, the prior-feature learning is proposed to reduce the computations by sharing the CNN layers between all weak classifiers. Using this strategy, all weak classifiers can be integrated into a single network. Results: Tenfold cross validation of binary classification was conducted on a total of 1357 nodules, including 765 noninvasive (AAH and AIS) and 592 invasive nodules (MIA and IAC). Ablation experimental results indicated that the proposed binary classifier achieved an accuracy of 73:4% AE 1:4 with an AUC of 81.3% AE 2:2. These results are superior compared to those achieved by three experienced chest imaging specialists who achieved an accuracy of 69:1% , 69:3% , and 67:9% , respectively. About 200 additional nodules were also collected. These nodules covered 50 cases for each category (AAH, AIS, MIA, and IAC, respectively). Both binary and multiple classifications were performed on these data and the results demonstrated that the proposed method definitely achieves better performance than the performance achieved by nonensemble deep learning methods. Conclusions: It can be concluded that the proposed adaptive-boost deep learning can significantly improve the performance of invasiveness classification of pulmonary subsolid nodules in CT images, while the prior-feature learning can significantly reduce the total size of deep models. The promising results on clinical data show that the trained models can be used as an effective lung cancer screening tool in hospitals. Moreover, the proposed strategy can be easily extended to other similar classification tasks in 3D medical images.

show abstract

Section: Introductionmentioning

confidence: 99%

Feature‐shared adaptive‐boost deep learning for invasiveness classification of pulmonary subsolid nodules in CT images

Wang

Chen

et al. 2020

Medical Physics

View full text Add to dashboard Cite

show abstract

“…The method in [22] models the stack of chest CT scans in connection with 3D Markov-Gibbs Random Field of voxel wise lungs and CT image intensities. 3D chest scans are pre-processed by region growing and connected components to identify the background voxels.…”

Section: Introductionmentioning

confidence: 99%

Efficient Segmentation of Lung Using FIDRM Filter and Successive Peak Quantization Transform

Firdouse¹,

Balasubramanian²

2018

IJIES

View full text Add to dashboard Cite

Abstract:Computed Tomography (CT) is the main source for analyzing the lung diseases and the preplanning of pulmonary surgeries. The accurate interpretation of chest CT scan makes a challenge to radiologists due to its complex visual nature. This paper proposes a novel method to segment the left and right lungs from the background of CT images to assist the pathologists for the easy and accurate diagnosis of pulmonary diseases. The segmentation of lung is a challenging task due to the anatomical difference of different people. This paper develops a novel algorithm namely successive peak based quantization transform (SPQT) to segment the lungs from the background. This method extracts 8 bit patterns from the input lung image and at last a BYTE data which is formed from 8 bit patterns is produced. Prior to SPQT lung segmentation, the lung image which may contain impulse noise is reduced into noise free one by using 'Fuzzy Impulse Detection and Reduction Method' (FIDRM). The novel lung image hikes the accuracy to a significant level than the existing algorithms.

show abstract

“…Imaging [384], the International Symposium on Computational Models for Life Sciences (CMLS) [385], the International Symposium on Biomedical Imaging (ISBI) [386], and the International Conference on Image Processing (ICIP) [387,388].…”

Section: E Summary and Discussionmentioning

confidence: 99%

“…To segment the 4D-CT scans, some modifications has been added to the 3D accurate lungs segmentation framework, that has been introduced in Chapter III which has the ability to segment lungs with wide range of pathologies, to be able to segment the 4D-CT scans. After segmenting the Exhale phase, using the 3D segmentation [384], the segmentation labels is propagated to the subsequent phases using the modified adaptive shape prior component only without the need to any other components which leads to a an accurate and faster segmentation. Visual appearances of different respiratory phases images and the exhale images guide the shape prior adaptation as follows (outlined in Algorithm 8).…”

Section: D-ct Lung Segmentationmentioning

confidence: 99%

Computational methods for the analysis of functional 4D-CT chest images.

Soliman¹

Self Cite

View full text Add to dashboard Cite

DEDICATIONThis dissertation is dedicated to my father's soul.iii ACKNOWLEDGMENTS In the name of Allah the most merciful, the most compassionate. All deepest thanks are due to Almighty Allah for the uncountable gifts given to me. Medical imaging is an important emerging technology that has been intensively used in the last few decades for disease diagnosis and monitoring as well as for the assessment of treatment effectiveness. Medical images provide a very large amount of valuable information that is too huge to be exploited by radiologists and physicians. Therefore, the design of computer-aided diagnostic (CAD) system, which can be used as an assistive tool for the medical community, is of a great importance. This dissertation deals with the development of a complete CAD system for lung cancer patients, which remains the leading cause of cancer-related death in the USA. In 2014, there were approximately 224,210 new cases of lung cancer and 159,260 related deaths. The process begins with the detection of lung cancer which is detected through the diagnosis of lung nodules (a manifestation of lung cancer). These nodules are approximately spherical regions of primarily high density tissue that are visible in computed tomography (CT) images of the lung. The treatment of these lung cancer nodules is complex, nearly 70% of lung cancer patients require radiation therapy as part of their treatment. Radiation-induced lung injury is a limiting toxicity that may decrease cure rates and increase morbidity and mortality treatment. By finding ways to accurately detect, at early stage, and hence prevent lung injury, it will have significant positive consequences for lung cancer patients.v The ultimate goal of this dissertation is to develop a clinically usable CAD system that can improve the sensitivity and specificity of early detection of radiation-induced lung injury based on the hypotheses that radiated lung tissues may get affected and suffer decrease of their functionality as a side effect of radiation therapy treatment. These hypotheses have been validated by demonstrating that automatic segmentation of the lung regions and registration of consecutive respiratory phases to estimate their elasticity, ventilation, and texture features to provide discriminatory descriptors that can be used for early detection of radiation-induced lung injury. The proposed methodologies will lead to novel indexes for distinguishing normal/healthy and injured lung tissues in clinical decisionmaking. To achieve this goal, a CAD system for accurate detection of radiation-induced lung injury that requires three basic components has been developed. These components are the lung fields segmentation, lung registration, and features extraction and tissue classification.This dissertation starts with an exploration of the available medical imaging modalities to present the importance of medical imaging in today's clinical applications. Secondly, the methodologies, challenges, and limitations of recent CAD systems for lung cancer detection are covered...

show abstract

Accurate Lungs Segmentation on CT Chest Images by Adaptive Appearance-Guided Shape Modeling

Cited by 88 publications

References 40 publications

Feature‐shared adaptive‐boost deep learning for invasiveness classification of pulmonary subsolid nodules in CT images

Feature‐shared adaptive‐boost deep learning for invasiveness classification of pulmonary subsolid nodules in CT images

Efficient Segmentation of Lung Using FIDRM Filter and Successive Peak Quantization Transform

Computational methods for the analysis of functional 4D-CT chest images.

Contact Info

Product

Resources

About