Convolutional neural networks (CNNs) have been increasingly applied to computer-aided diagnosis (CADx) for lung nodule malignancy prediction, which usually is a binary classification task. However, CNNs were often difficult to capture optimal features, thereby affect the classification performance. This study developed a CADx system based on a CNN model with auxiliary task learning to predict lung nodule malignancy in chest computed tomography (CT) scans. Our CADx system took raw CT image cubes centering at nodules as input and generated one main output and eight auxiliary outputs. The main output predicted lung nodule malignancy; the auxiliary outputs predicted lesion size and characteristics. The auxiliary tasks offered assistance for predicting the final nodule malignancy. The performance of the developed lung nodule CADx system was verified by use of the LIDC dataset. Results showed that our CADx system achieved improved performance for lung nodule malignancy prediction.
The purpose of this study was to evaluate radiologists’ performance in detecting lung nodules using chest computed tomography (CT) scans when assisted by a computer-aided detection (CAD) system with a vessel suppression function. Three radiologists participated in this preliminary observer study. The observer study was conducted on 80 CT scans including 94 nodules. The ratio of nodule-free scans to with-nodule scans was 1:1. CAD systems with (CAD-VS) and without (CAD-nVS) a vessel suppression function were developed to assist radiologists in reading chest CT scans. The radiologists read the CT scans in a two-session process, which had at least a one-month interval in between. Freeresponse receiver operating characteristic (FROC) curves and localization receiver operating characteristic (LROC) curves were utilized to analyze the nodule detection results. The CAD-VS and the CAD-nVS detected 96.8% and 93.6% of nodules, respectively, at 0.5 false positive per scan. For the observer study, the mean area under the LROC curve (LROC-AUC) for nodule detection improved from 0.877 by use of the CAD-nVS to 0.942 by use of the CAD-VS. Radiologists averagely detected 94.0% and 96.5% of nodules with the CAD-nVS and CAD-VS, respectively; average specificity increased from 71.7% to 81.7%. The CAD-VS improved radiologists’ performance for lung nodule detection, compared to the general CAD-nVS. This suggests that the CAD-VS technique is feasible to help radiologists further improve the clinical detection accuracy of lung nodules in chest CT scans.
Abdominal MRI is susceptible to respiratory motion artifacts. The existing clinical solution is using breathing belt to track the movement of the abdomen and trigger MRI acquisition during the end-expiration phase. Attaching respiratory belt to patients often slows down clinical workflow and affects patient comfort especially for those with surgical wounds and respiratory disorders. Herein we, for the first, propose a novel MRI compatible frequency modulated continuous wave (FMCW) radar to track respiratory motion within MRI bore in a non-contact fashion. The electromagnetic wave from FMCW radar can penetrate clothing and MRI RF coils to achieve continuous monitoring of patient’s vital signs. The system consists of a front-end FMCW radar sensor and a FPGA based power management/communication board that interface with a clinical MRI scanner. This design fully integrates the FMCW radar signal with MRI control console to enable real time respiratory triggered MRI acquisition. Consistent respiratory waveform was validated by comparing FMCW signal with traditional breathing belt measurement. Superior image quality from clinical MRI pulse sequence was achieved using the developed system in healthy volunteers.
The suppression of lung vessels in chest computed tomography (CT) scans can enhance the conspicuity of lung nodules, thereby may improve the detection rate of early lung cancer. This study aimed to verify the effect of lung vessel suppression on the performance of the lung nodule detector. Firstly, a lung vessel suppression technique was developed to remove the vessels while preserving the nodules. Then, a lung nodule detector was developed with two stages: nodule candidate generation and false positive reduction. The vessel suppression and nodule detection methods were validated respectively in 50 three-dimensional (3D) chest CT images with manually-labeled vessel trees and 888 3D chest CT images with manually-located nodules (LUNA16). The lung vessel suppression results were quantitatively evaluated by using the Dice coefficient (DICE) and the contrast-to-noise ratio (CNR), and the lung nodule detection results were quantitatively evaluated by using the sensitivity under two conditions: “without” and “with vessel suppression”. The lung vessel suppression accurately removed vessels with a DICE of 0.943 and improved the CNR for nodules from 4.24 (6.27 dB) to 7.02 (8.46 dB), which subsequently improved the average sensitivity from 0.948 to 0.969 under 7 specified false positives for lung nodule detection.
Temporal subtraction of sequential chest radiographs based on image registration technique has been developed for decades to assist radiologists in the detection of interval changes. Although the performance of current methods is good, the computation cost of these methods is generally high. The high computation cost is mainly due to the iterative optimization problem of non-learning-based deformable registration. In this work we present a fast unsupervised learning-based algorithm for deformable registration of chest radiographs. Based on a convolutional neural network, the proposed model learns to directly estimate spatial transformations from pairs of moving images and fixed images, and uses the transformations to warp the moving images. We apply a regularization term to constrain the model to learn local matching. The model is trained by optimizing a pair-wise similarity metric between the warped moving image and the fixed image, with no need for any supervised information such as ground truth deformation fields. The trained model can be used to predict the warped moving images in one shot, and is thus very fast. The subtraction images of the warped images and the fixed images are able to enhance various interval changes. The preliminary results showed that for approximately 98.55% cases, the learning-based method could obtain improved or comparable registration in comparison with the baseline method.
Lung cancer is the leading cause of cancer deaths worldwide. Early diagnosis is critical in increasing the 5-year survival rate of lung cancer, so the efficient and accurate detection of lung nodules, potential precursors to lung cancer, is evermore important. In this paper, a computer-aided lung nodule detection system using convolution neural networks (CNN) and handcrafted features for false positive reduction is developed. The CNNs were trained with three types of images: lung CT images, their nodule-enhanced images, and their blood vessel-enhanced images. For each nodule candidate, nine 2D patches from differently oriented planes were extracted from each type of images. Patches of the same orientation from the same type of image across different candidates were used to train the CNNs independently, which were used to extract 864 features. 88 handcrafted features including intensity, shape, and texture features were also obtained from the lung CT images. The CNN features and handcrafted features were then combined to train a classifier, and a support vector machine was adopted to achieve the final classification results. The proposed method was evaluated on 1004 CT scans from the LIDC-IDRI database using 10-fold cross-validation. Compared with the traditional CNN method using only lung CT images, the proposed method boosted the sensitivity of nodule detection from 89.0% to 90.9% at 4 FPs/scan and from 71.6% to 78.2% at 1 FP/scan. This indicates that a combination of handcrafted features and CNN features from both lung CT images and enhanced images is a promising method for lung nodule detection.
Lung cancer is the first killer among the cancer deaths. Malignant lung nodules have extremely high mortality while
some of the benign nodules don't need any treatment .Thus, the accuracy of diagnosis between benign or malignant
nodules diagnosis is necessary. Notably, although currently additional invasive biopsy or second CT scan in 3 months
later may help radiologists to make judgments, easier diagnosis approaches are imminently needed. In this paper, we
propose a novel CAD method to distinguish the benign and malignant lung cancer from CT images directly, which can
not only improve the efficiency of rumor diagnosis but also greatly decrease the pain and risk of patients in biopsy
collecting process. Briefly, according to the state-of-the-art radiomics approach, 583 features were used at the first step
for measurement of nodules' intensity, shape, heterogeneity and information in multi-frequencies. Further, with Random
Forest method, we distinguish the benign nodules from malignant nodules by analyzing all these features. Notably, our
proposed scheme was tested on all 79 CT scans with diagnosis data available in The Cancer Imaging Archive (TCIA)
which contain 127 nodules and each nodule is annotated by at least one of four radiologists participating in the project.
Satisfactorily, this method achieved 82.7% accuracy in classification of malignant primary lung nodules and benign
nodules. We believe it would bring much value for routine lung cancer diagnosis in CT imaging and provide
improvement in decision-support with much lower cost.
In order to enable multiple disciplines of medical researchers, clinical physicians and biomedical engineers working together in a secured, efficient, and transparent cooperative environment, we had designed an e-Science platform for biomedical imaging research and application cross multiple academic institutions and hospitals in Shanghai and presented this work in SPIE Medical Imaging conference held in San Diego in 2012. In past the two-years, we implemented a biomedical image chain including communication, storage, cooperation and computing based on this e-Science platform. In this presentation, we presented the operating status of this system in supporting biomedical imaging research, analyzed and discussed results of this system in supporting multi-disciplines collaboration cross-multiple institutions.
In this paper, we extend the spectrography method to visualize 3D structures of complex samples from only one spectral view. Utilizing a weighted difference map and the Fourier central slice theorem, a number of Fourier planes are reconstructed, which go through the origin of the 3D Fourier space and interact with a region formed by the Ewald spheres. Thus, the complex x-ray wave fronts can be recovered at small tilting angles from the incident x-ray beam. Patterns from various computed projections can generate perception of 3D structure features inside the sample. To demonstrate the feasibility of the proposed spectrographic imaging method, numerical simulations are performed and analyzed. The results suggest that spectrography is an effective method for 3D structure studies by a single spectral exposure.
KEYWORDS: X-ray computed tomography, Monte Carlo methods, X-rays, Tissues, Photons, Sensors, Temporal resolution, Data modeling, Chest, Imaging systems
Multiple-source structure is promising in the development of computed tomography, for it could effectively eliminate
motion artifacts in the cardiac scanning and other time-critical implementations with high temporal resolution. However,
concerns about the dose performance shade this technique, as few reports on the evaluation of dose performance of
multiple-source CT have been proposed for judgment. Our experiments focus on the dose performance of one specific
multiple-source CT geometry, the triple-source CT scanner, whose theories and implementations have already been
well-established and testified by our previous work. We have modeled the triple-source CT geometry with the help of
EGSnrc Monte Carlo radiation transport code system, and simulated the CT examinations of a digital chest phantom with
our modified version of the software, using x-ray spectrum according to the data of physical tube. Single-source CT
geometry is also estimated and tested for evaluation and comparison. Absorbed dose of each organ is calculated
according to its real physics characteristics. Results show that the absorbed radiation dose of organs with the
triple-source CT is almost equal to that with the single-source CT system. As the advantage of temporal resolution, the
triple-source CT would be a better choice in the x-ray cardiac examination.
To utilize the synergy between CT and MR datasets from an object at the same time, a unified dual-modality image reconstruction approach is proposed using a dual-dictionary learning technique. The key is to establish a knowledgebased connection between these two datasets for a tight fusion of different imaging modalities. Our scheme consists of three inter-related elements: dual-dictionary learning, CT image reconstruction, and MR image reconstruction. Our experiments show that even with highly under-sampled MR data and few x-ray projections, we can still satisfactorily reconstruct both MR and CT images. This approach can be potentially useful for a CT-MRI system.
Grating-based phase-contrast imaging has been a hot topic for several years owing to its excellent ability to discern lowdensity
materials. Compared with commercial X-ray computed tomography (CT) systems, the ‘phase-stepping’ motion
in grating-based phase-contrast CT (GPC-CT) takes too long to implement in a CT scan. Additionally, the radiation dose
delivered to the sample is several times that delivered by a conventional CT scan. In this paper, we compared four kinds
of scanning schemes for GPC-CT and pointed out their advantages and limitations for clinical applications. Following
the interlaced phase stepping method, we proposed a novel reconstruction method, namely the inner-focusing
reconstruction method. With the proposed method, the gratings can be easily integrated into a conventional CT system
implementing phase-contrast imaging with a complete field-of-view. Numerical experiments verified the effectiveness of
the proposed method in achieving fast and low-dose GPC-CT.
We study reconstruction methods and imaging properties of the circular and helical interlaced source-detector array
(ISDA) computed tomography (CT) system. The system uses carbon nanotube (CNT) based field emission X-ray source
arrays and detector arrays. Distributed sources and detectors allow projection data been acquired from different direction by switching on the X-ray sources sequentially, requiring no rotation of gantry or object. Thus the system enables high temporal resolution and eliminates motion artifacts caused by gantry rotation as is common in the conventional CT systems. Interpolation is implemented to patch the absent data in the projection image, and tilted plane Feldkamp type reconstruction algorithm (TPFR) is used to reduce the cone beam artifacts for helical ISDA CT. We analyze the distribution of artifacts in the reconstruction, as well as the influence of the detector array gap and the helix pitch on reconstruction quality. Simulation studies demonstrate that the gap ratio is the key factor on the artifacts due to the gap, and increasing the pitch will reduced gap ratio. Choosing the helix pitch appropriately by getting a balance between cone beam artifacts and gap induced artifacts can get a better reconstruction.
A scheme is proposed here for spectrography that is for 3D analysis of an object from a single spectral view. In this scheme, we first reconstruct some planes that go through the origin of the Fourier space and interact significantly with Ewald spheres. Since only amplitude data are measured, the corresponding phase information is estimated using a hybrid-input-output algorithm. Finally, each of the specified planes is reconstructed from the highly under-sampled Fourier data using a dictionary learning technique. According to the Fourier central slice theorem, the reconstructed image is equivalent to a parallel-beam projection of the object. Hence, the reconstructed projections can be used to analyze internal structures of the object via stereo imaging. The numerical experiments suggest that our method is promising for studies on 3D structures from one spectral view.
Multiple-source cone-beam scanning is a promising mode for dynamic volumetric CT/micro-CT. The
first dynamic CT system is the Dynamic Spatial Reconstructor (DSR) built in 1979. The pursuance for
higher temporal resolution has largely driven the development of CT technology, and recently led to
the emergence of Siemens dual-source CT scanner. Given the impact and limitation of dual-source
cardiac CT, triple-source cone-beam CT seems a natural extension for future cardiac CT. Our work
shows that trinity (triple-source architecture) is superior to duality (dual-source architecture) for helical
cone-beam CT in terms of exact reconstruction. In particular, a triple-source helical scan allows a
perfect mosaic of longitudinally truncated cone-beam data to satisfy the Orlov condition and yields
better noise performance than the dual-source counterpart. In the (2N+1)-source helical CT case, the
more sources, the higher temporal resolution. In the N-source saddle CT case, a triple-source scan
offers the best temporal resolution for continuous dynamic exact reconstruction of a central volume.
The recently developed multi-source cone-beam algorithms include an exact backprojection-filtration
(BPF) approach and a "slow" exact filtered-backprojection (FBP) algorithm for (2N+1)-source helical
CT, two fast quasi-exact FBP algorithms for triple-source helical CT, as well as a fast exact FBP
algorithm for triple-source saddle CT. Some latest ideas will be also discussed, such as multi-source
interior tomography and multi-beam field-emission x-ray CT.
In this paper, a hybrid helix-saddle trajectory scanning mode is proposed for bolus-chasing CT angiography. By combining the conventional helical trajectory and saddle trajectory appropriately, an optimal curve can be obtained with a capability of localized volumetric imaging at desirable locations. In this context, a condition for the PI-line existence is determined. Then, both filtered-backprojection (FBP), backprojection filtration (BPF) and reduced-scan FBP algorithms are developed. Numerical studies with the 3D Shepp-Logan phantom support the validity and merits of the
proposed trajectories and associated algorithms.
In this paper, we propose an exact shift-invariant filtered backprojection (FBP) algorithm for triple-source saddle-curve
cone-beam CT. In this imaging geometry, the sources are symmetrically positioned along a circle, and the trajectory of each x-ray source is a saddle. Then, we extend Yang's formula from the single-source case to the triple-source case. The saddles can be divided into four parts to specify four datasets. Each of them contains three data segments associated with different saddles. Then, images can be reconstructed on the planes orthogonal to the z-axis. Each plane intersects the
trajectories at six pointes which can be used to define the filtering directions. With our triple-source approach, the scanning time is only one-third of that with the single-source saddle trajectory, and the reconstructed image quality is excellent in our numerical studies. These new features are important for cardiac imaging and small animal imaging.
In a previous study, we proposed a helical scanning configuration with triple X-ray sources symmetrically positioned
and established its reconstruction algorithm. Although symmetrically positioned sources are convenient in practice,
artifacts can be produced in a reconstructed image if the physical sources are not equally spaced. In this work, we
develop an exact backprojection filtration (BPF) type algorithm for the configuration with unequally spaced triple
sources to reduce the artifacts. Similar to the Tam-Danielsson window, we define the minimum detection window as the
region bounded by the most adjacent turns of two helices. The sum of the heights of the three consequent minimum
detection windows is equal to that of the traditional Tam-Danielsson window for a single source. Furthermore, we prove
that inter-helix PI-lines satisfy the existence and uniqueness properties (i.e., through any point inside the triple helices,
there exists one and only one inter-helix PI-line for any pair of helices). The proposed algorithm is of the
backprojection-filtration (BPF) type and can be implemented in three steps: 1) differentiation of the cone-beam
projection from each source; 2) weighted backprojection of the derivates on the inter-helix PI-arcs; 3) inverse Hilbert
transformation along one of the three inter-helix PI-lines. Numerical simulations with 3D Shepp-Logan phantoms are
performed to validate the algorithm. We also demonstrate that artifacts are generated when the algorithm for the
symmetric configuration is applied to the unequally spaced helices setting.
Multiple source helical cone-beam scanning is a promising technique for dynamic volumetric CT/micro-CT. In the previous studies, we had proposed a helical cone-beam scanning mode with triple x-ray source and detector assemblies that are symmetrically arranged, and proved the property of minimum detection windows under this configuration. Moreover, we had established an exact backprojection filtration (BFP) reconstruction algorithm for this setting. In this paper, we perform simulation studies for this reconstruction algorithm with 3D Shepp-Logan and Defrise phantoms. The implementation of the BFP algorithm in the planar detector geometry consists of three steps. First, the cone-beam projection from each of the three sources is differentiated respectively. Second, the derivates on the three inter-helix PI-arcs are summed up with weights to form the backprojection. Third, inverse Hilbert transformations are performed along each of the three inter-helix PI-lines. The reconstructed images validate the proposed algorithm. Furthermore, this work can be generalized to the case of multiple source helical cone-beam CT.
In this paper, we propose a helical cone-beam scanning configuration of triple symmetrically located X-ray sources, and study minimum detection windows to extend the traditional Tam-Danielsson window for exact image reconstruction. For three longitudinally displaced scanning helices of the same radius and a source location on any helix, the corresponding minimum detection window is bounded by the most adjacent turns respectively selected from the other two helices. The height of our proposed minimum detector window is only 1/3 of that in the single helix case. Associated with proposed minimum detection windows, we define the inter-helix PI-line and establish its existence and uniqueness property: through any point inside the triple helices, there exists one and only one inter-helix PI-line for any pair of the helices. Furthermore, we prove that cone-beam projection data from such a triple-source helical scan are sufficient for exact image reconstruction. Although there are certain redundancies among those projection data, the redundant part cannot be removed by shrinking the detector window without violating the data sufficiency condition. Those results are important components for development of exact or quasi-exact image reconstruction algorithms in the case of triple-source helical cone-beam scanning in the future.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.