ABSTRACT
We present the use of mean Hounsfield units within lungs as a metric of disease severity for the comparison of image analysis models in patients with COPD and COVID. We used this metric to assess the performance of a novel 3D global context attention network for image segmentation that produces lung masks from thoracic HRCT scans. Results showed that the mean Hounsfield units enable a detailed comparison of our 3D implementation of the GC-Net model to the V-Net segmentation algorithm. We implemented a biomimetic data augmentation strategy and used a quantitative severity metric to assess its performance. Framing our investigation around lung segmentation for patients with respiratory diseases allows analysis of the strengths and weaknesses of the implemented models in this context.Clinical Relevance - Mean Hounsfield units within the lung volume can be used as an objective measure of respiratory disease severity for the comparison of CT scan analysis algorithms.
Subject(s)
COVID-19 , Humans , Lung/diagnostic imaging , Algorithms , Tomography, X-Ray Computed/methods , ThoraxABSTRACT
Background: Identifying systemic sclerosis (SSc) and idiopathic pulmonary fibrosis (IPF) patients at risk of more rapid forced vital capacity (FVC) decline could improve trial design. The purpose of the present study was to explore the prognostic value of quantitative high-resolution computed tomography (HRCT) metrics derived by Imbio lung texture analysis (LTA) tool in predicting FVC slope. Methods: This retrospective study used data from patients who were not treated with investigational drugs with and without background antifibrotic therapies in tocilizumab phase 3 SSc, lebrikizumab phase 2 IPF, and zinpentraxin alfa phase 2 IPF studies conducted from 2015 to 2021. Controlled HRCT axial volumetric multidetector computed tomography scans were evaluated using the Imbio LTA tool. Associations between HRCT metrics and FVC slope were assessed through the Spearman correlation coefficient and adjusted R2 in a linear regression model adjusted by demographics and baseline clinical characteristics. Results: A total of 271 SSc and IPF patients were analysed. Correlation coefficients of highest magnitude were observed in the SSc study between the extent of ground glass, normal volume, quantification of interstitial lung disease, reticular pattern, and FVC slope (-0.25, 0.28, -0.28, and -0.33, respectively), while the correlation coefficients observed in IPF studies were in general <0.2. The incremental prognostic value of the baseline HRCT metrics was marginal after adjusting baseline characteristics and was inconsistent across study arms. Conclusion: Data from the SSc and IPF studies suggested weak to no and inconsistent correlation between quantitative HRCT metrics derived by the Imbio LTA tool and FVC slope in the studied SSc and IPF population.
ABSTRACT
Speech and language changes occur in Alzheimer's disease (AD), but few studies have characterized their longitudinal course. We analyzed open-ended speech samples from a prodromal-to-mild AD cohort to develop a novel composite score to characterize progressive speech changes. Participant speech from the Clinical Dementia Rating (CDR) interview was analyzed to compute metrics reflecting speech and language characteristics. We determined the aspects of speech and language that exhibited significant longitudinal change over 18 months. Nine acoustic and linguistic measures were combined to create a novel composite score. The speech composite exhibited significant correlations with primary and secondary clinical endpoints and a similar effect size for detecting longitudinal change. Our results demonstrate the feasibility of using automated speech processing to characterize longitudinal change in early AD. Speech-based composite scores could be used to monitor change and detect response to treatment in future research. HIGHLIGHTS: Longitudinal speech samples were analyzed to characterize speech changes in early AD.Acoustic and linguistic measures showed significant change over 18 months.A novel speech composite score was computed to characterize longitudinal change.The speech composite correlated with primary and secondary trial endpoints.Automated speech analysis could facilitate remote, high frequency monitoring in AD.
ABSTRACT
RATIONALE AND OBJECTIVES: The objective of this study was to assess the feasibility of single-inhalation xenon-enhanced computed tomography (XeCT) to provide clinically practical, high-resolution pulmonary ventilation imaging to clinics with access to only a single-energy computed tomography scanner, and to reduce the subject's overall exposure to xenon by utilizing a higher (70%) concentration for a much shorter time than has been employed in prior studies. MATERIALS AND METHODS: We conducted an institutional review board-approved prospective feasibility study of XeCT for 15 patients undergoing thoracic radiotherapy. For XeCT, we acquired two breath-hold single-energy computed tomography images of the entire lung with a single inhalation each of 100% oxygen and a mixture of 70% xenon and 30% oxygen, respectively. A video biofeedback system for coached patient breathing was used to achieve reproducible breath holds. We assessed the technical success of XeCT acquisition and side effects. We then used deformable image registration to align the breath-hold images with each other to accurately subtract them, producing a map of lung xenon distribution. Additionally, we acquired ventilation single-photon emission computed tomography-computed tomography (V-SPECT-CT) images for 11 of the 15 patients. For a comparative analysis, we partitioned each lung into 12 sectors, calculated the xenon concentration from the Hounsfield unit enhancement in each sector, and then correlated this with the corresponding V-SPECT-CT counts. RESULTS: XeCT scans were tolerated well overall, with a mild (grade 1) dizziness as the only side effect in 5 of the 15 patients. Technical failures in five patients occurred because of inaccurate breathing synchronization with xenon gas delivery, leaving seven patients analyzable for XeCT and single-photon emission computed tomography correlation. Sector-wise correlations were strong (Spearman coefficient >0.75, Pearson coefficient >0.65, P value <.002) for two patients for whom ventilation deficits were visibly pronounced in both scans. Correlations were nonsignificant for the remaining five who had more homogeneous XeCT ventilation maps, as well as strong V-SPECT-CT imaging artifacts attributable to airway deposition of the aerosolized imaging agent. Qualitatively, XeCT demonstrated higher resolution and no central airway deposition artifacts compared to V-SPECT-CT. CONCLUSIONS: In this pilot study, single-breath XeCT ventilation imaging was generally feasible for patients undergoing thoracic radiotherapy, using an imaging protocol that is clinically practical and potentially widely available. In the future, the xenon delivery failures can be addressed by straightforward technical improvements to the patient biofeedback coaching system.
Subject(s)
Lung/diagnostic imaging , Pulmonary Ventilation , Tomography, X-Ray Computed/methods , Administration, Inhalation , Aged , Algorithms , Breath Holding , Feasibility Studies , Female , Humans , Image Processing, Computer-Assisted , Male , Middle Aged , Pilot Projects , Prospective Studies , Tomography, Emission-Computed, Single-Photon , Xenon/adverse effectsABSTRACT
We propose and validate an end-to-end deep learning pipeline employing multi-label learning as a tool for creating differential diagnoses of lung pathology as well as quantifying the extent and distribution of emphysema in chest CT images. The proposed pipeline first employs deep learning based volumetric lung segmentation using a 3D CNN to extract the entire lung out of CT images. Then, a multi-label learning model is exploited for the classification creation differential diagnoses for emphysema and then used to correlate with the emphysema diagnosed by radiologists. The five lung tissue patterns which are involved in most lung disease differential diagnoses were classified as: ground glass, fibrosis, micronodules (random, perilymphatic and centrilobular lung nodules), normal appearing lung, and emphysematous lung tissue. To the best of our knowledge, this is the first end-to-end deep learning pipeline for the creation of differential diagnoses for lung disease and the quantification of emphysema. A comparative analysis shows the performance of the proposed pipeline on two publicly available datasets.
Subject(s)
Pulmonary Emphysema , Deep Learning , Humans , Lung , Tomography, X-Ray ComputedABSTRACT
BACKGROUND AND PURPOSE: A major challenge in CT screening for lung cancer is limited specificity when distinguishing between malignant and non-malignant pulmonary nodules (PN). Malignant nodules have different mechanical properties and tissue characteristics ('stiffness') from non-malignant nodules. This study seeks to improve CT specificity by demonstrating in rats that measurements of volumetric ratios in PNs with varying composition can be determined by respiratory-gated dynamic CT imaging and that these ratios correlate with direct physical measurements of PN stiffness. METHODS AND MATERIALS: Respiratory-gated MicroCT images acquired at extreme tidal volumes of 9 rats with PNs from talc, matrigel and A549 human lung carcinoma were analyzed and their volumetric ratios (δ) derived. PN stiffness was determined by measuring the Young's modulus using atomic force microscopy (AFM) for each nodule excised immediately after MicroCT imaging. RESULTS: There was significant correlation (p=0.0002) between PN volumetric ratios determined by respiratory-gated CT imaging and the physical stiffness of the PNs determined from AFM measurements. CONCLUSION: We demonstrated proof of concept that PN volume changes measured non-invasively correlate with direct physical measurements of stiffness. These results may translate clinically into a means of improving the specificity of CT screening for lung cancer and/or improving individual prognostic assessments based on lung tumor stiffness.
Subject(s)
Lung Neoplasms/pathology , Solitary Pulmonary Nodule/pathology , Tomography, X-Ray Computed/methods , Animals , Cell Line, Tumor , Female , Humans , Lung Neoplasms/diagnostic imaging , Microscopy, Atomic Force , Rats , Solitary Pulmonary Nodule/diagnostic imaging , Tumor BurdenABSTRACT
The steadily growing amounts of digital neuroscientific data demands for a reliable, systematic, and computationally effective retrieval algorithm. In this paper, we present Neuron-Miner, which is a tool for fast and accurate reference-based retrieval within neuron image databases. The proposed algorithm is established upon hashing (search and retrieval) technique by employing multiple unsupervised random trees, collectively called as Hashing Forests (HF). The HF are trained to parse the neuromorphological space hierarchically and preserve the inherent neuron neighborhoods while encoding with compact binary codewords. We further introduce the inverse-coding formulation within HF to effectively mitigate pairwise neuron similarity comparisons, thus allowing scalability to massive databases with little additional time overhead. The proposed hashing tool has superior approximation of the true neuromorphological neighborhood with better retrieval and ranking performance in comparison to existing generalized hashing methods. This is exhaustively validated by quantifying the results over 31266 neuron reconstructions from Neuromorpho.org dataset curated from 147 different archives. We envisage that finding and ranking similar neurons through reference-based querying via Neuron Miner would assist neuroscientists in objectively understanding the relationship between neuronal structure and function for applications in comparative anatomy or diagnosis.
Subject(s)
Brain/cytology , Data Mining , Image Processing, Computer-Assisted/methods , Neurons/cytology , Software , Algorithms , Animals , Databases, Factual , Humans , Machine LearningABSTRACT
PURPOSE: To investigate the hypothesis that CT ventilation functional image-based IMRT plans designed to avoid irradiating highly-functional lung regions are comparable to single-photon emission CT (SPECT) ventilation functional image-based plans. METHODS AND MATERIALS: Three IMRT plans were created for eight thoracic cancer patients using: (1) CT ventilation functional images, (2) SPECT ventilation functional images, and (3) anatomic images (no functional images). CT ventilation images were created by deformable image registration of 4D-CT image data sets and quantitative analysis. The resulting plans were analyzed for the relationship between the deviations of CT-functional plan metrics from anatomic plan metrics (ΔCT-anatomic) and those of SPECT-functional plans (ΔSPECT-anatomic), and moreover for agreements of various metrics between the CT-functional and SPECT-functional plans. RESULTS: The relationship between ΔCT-anatomic and ΔSPECT-anatomic was strong (e.g., R=0.94; linear regression slope 0.71). The average differences and 95% limits of agreement between the CT-functional and SPECT-functional plan metrics (except for monitor units) for various structures were mostly less than 1% and 2%, respectively. CONCLUSIONS: This study demonstrated a reasonable agreement between the CT ventilation functional image-based IMRT plans and SPECT-functional plans, suggesting the potential for CT ventilation imaging to serve as a surrogate for SPECT ventilation in functional image-guided radiotherapy.
Subject(s)
Carcinoma, Non-Small-Cell Lung/radiotherapy , Lung Neoplasms/radiotherapy , Lung/diagnostic imaging , Radiotherapy Planning, Computer-Assisted/methods , Adult , Aged , Aged, 80 and over , Carcinoma, Non-Small-Cell Lung/diagnostic imaging , Female , Four-Dimensional Computed Tomography/methods , Humans , Lung Neoplasms/diagnostic imaging , Male , Middle Aged , Radiotherapy, Image-Guided , Respiration , Tomography, Emission-Computed, Single-Photon/methodsABSTRACT
BACKGROUND AND PURPOSE: To develop a noninvasive method for determining malignant pulmonary nodule (MPN) elasticity, and compare it against expert dual-observer manual contouring. METHODS AND MATERIALS: We analyzed breath-hold images at extreme tidal volumes of 23 patients with 30 MPN treated with stereotactic ablative radiotherapy. Deformable image registration (DIR) was applied to the breath-hold images to determine the volumes of the MPNs and a ring of surrounding lung tissue (ring) in each state. MPNs were also manually delineated on deep inhale and exhale images by two observers. Volumes were compared between observers and DIR by Dice similarity. Elasticity was defined as the absolute value of the volume ratio of the MPN minus one normalized to that of the ring. RESULTS: For all 30 tumors the Dice coefficient was 0.79±0.07 and 0.79±0.06 between DIR with observers 1 and 2, respectively, close to the inter-observer Dice value, 0.81±0.1. The elasticity of MPNs was 1.24±0.26, demonstrating that volume change of the MPN was less than that of the surrounding lung. CONCLUSION: We developed a noninvasive CT elastometry method based on DIR that measures the elasticity of biopsy-proven MPN. Our future direction would be to develop this method to distinguish malignant from benign nodules.
Subject(s)
Lung Neoplasms/diagnostic imaging , Adult , Aged , Aged, 80 and over , Elasticity , Female , Humans , Male , Middle Aged , Tomography, X-Ray Computed/methodsABSTRACT
A novel energy function for computing planar optical flow from X-ray CT images was presented and reported in detail in [1]. The technique combines four terms: brightness constancy, gradient constancy, continuity equation based on mass conservation, and discontinuity-preserving spatio-temporal smoothness. Both qualitative and quantitative evaluation of the proposed method demonstrated that the method results in significantly better angular errors than previous well-known techniques for optical flow estimation. A multi-scale approach to motion field computation based on this framework is presented in this paper. The proposed approach significantly speeds up the calculations, realizing computational savings. Additionally, an approach to determination of optimum values of scalar weights in the energy function is herein proposed. Normalized mutual information measured between the first image warped with the estimated motion and the second image is used to determine the optimum value for weighting parameters.
Subject(s)
Lung/diagnostic imaging , Tomography, X-Ray Computed/methods , Humans , Models, TheoreticalABSTRACT
An approximate model for the effect of respiration is that the cross section of the thoracic area under interrogation experience time-varying magnification and displacement along two perpendicular axes - we propose to model this motion as parametric affine motion. A theoretical framework for determination of parameters of affine motion modeling the global respiratory motion based on the sinogram data in the projection domain is described. It is assumed that the spatial image considered is a density image where conservation of mass holds.