Harmonization of chest CT scans for different doses and reconstruction methods

Quantification of Emphysema Progression at CT Using Simultaneous Volume, Noise, and Bias Lung Density Correction

Background CT attenuation is affected by lung volume, dosage, and scanner bias, leading to inaccurate emphysema progression measurements in multicenter studies. Purpose To develop and validate a method that simultaneously corrects volume, noise, and interscanner bias for lung density change estimation in emphysema progression at CT in a longitudinal multicenter study. Materials and Methods In this secondary analysis of the prospective Genetic Epidemiology of Chronic Obstructive Pulmonary Disease (COPDGene) study, lung function data were obtained from participants who completed baseline and 5-year follow-up visits from January 2008 to August 2017. CT emphysema progression was measured with volume-adjusted lung density (VALD) and compared with the joint volume-noise-bias-adjusted lung density (VNB-ALD). Reproducibility was studied under change of dosage protocol and scanner model with repeated acquisitions. Emphysema progression was visually scored in 102 randomly selected participants. A stratified analysis of clinical characteristics was performed that considered groups based on their combined lung density change measured by VALD and VNB-ALD. Results A total of 4954 COPDGene participants (mean age, 60 years ± 9 [SD]; 2511 male, 2443 female) were analyzed (1329 with repeated reduced-dose acquisition in the follow-up visit). Mean repeatability coefficients were 30 g/L ± 0.46 for VALD and 14 g/L ± 0.34 for VNB-ALD. VALD measurements showed no evidence of differences between nonprogressors and progressors (mean, -5.5 g/L ± 9.5 vs -8.6 g/L ± 9.6; P = .11), while VNB-ALD agreed with visual readings and showed a difference (mean, -0.67 g/L ± 4.8 vs -4.2 g/L ± 5.5; P < .001). Analysis of progression showed that VNB-ALD progressors had a greater decline in forced expiratory volume in 1 second (-42 mL per year vs -32 mL per year; Tukey-adjusted P = .002). Conclusion Simultaneously correcting volume, noise, and interscanner bias for lung density change estimation in emphysema progression at CT improved repeatability analyses and agreed with visual readings. It distinguished between progressors and nonprogressors and was associated with a greater decline in lung function metrics. Clinical trial registration no. NCT00608764 © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Goo in this issue.

Harmonizing CT Images via Physics-based Deep Neural Networks

The rendition of medical images influences the accuracy and precision of quantifications. Image variations or biases make measuring imaging biomarkers challenging. The objective of this paper is to reduce the variability of computed tomography (CT) quantifications for radiomics and biomarkers using physics-based deep neural networks (DNNs). With the proposed framework, it is possible to harmonize the different renditions of a single CT scan (with variations in reconstruction kernel and dose) into an image that is in close agreement with the ground truth. To this end, a generative adversarial network (GAN) model was developed where the generator is informed by the scanner's modulation transfer function (MTF). To train the network, a virtual imaging trial (VIT) platform was used to acquire CT images, from a set of forty computational models (XCAT) serving as the patient model. Phantoms with varying levels of pulmonary disease, such as lung nodules and emphysema, were used. We scanned the patient models with a validated CT simulator (DukeSim) modeling a commercial CT scanner at 20 and 100 mAs dose levels and then reconstructed the images by twelve kernels representing smooth to sharp kernels. An evaluation of the harmonized virtual images was conducted in four different ways: 1) visual quality of the images, 2) bias and variation in density-based biomarkers, 3) bias and variation in morphological-based biomarkers, and 4) Noise Power Spectrum (NPS) and lung histogram. The trained model harmonized the test set images with a structural similarity index of 0.95±0.1, a normalized mean squared error of 10.2±1.5%, and a peak signal-to-noise ratio of 31.8±1.5 dB. Moreover, emphysema-based imaging biomarkers of LAA-950 (-1.5±1.8), Perc15 (13.65±9.3), and Lung mass (0.1±0.3) had more precise quantifications.

Harmonization of technical image quality in computed tomography: comparison between different reconstruction algorithms and kernels from six scanners

Purpose. The radiology department faces a large number of reconstruction algorithms and kernels during their computed tomography (CT) optimization process. These reconstruction methods are proprietary and ensuring consistent image quality between scanners is becoming increasingly difficult. This study contributes to solving this challenge in CT image quality harmonization by modifying and evaluating a reconstruction algorithm and kernel matching scheme.Methods. The Catphan 600 phantom was scanned with six different CT scanners from four vendors. The phantom was scanned with volumetric CT dose indices (CTDIvols) of 10 mGy and 40 mGy, and the data were reconstructed using 1 mm and 5 mm slices with each combination of reconstruction algorithm, body region kernel, and iterative and deep learning reconstruction strength. A matching scheme developed in previous research, which utilizes the noise power spectrum (NPS) and modulation transfer function (MTF), was modified based on our organization's needs and used to identify the matching reconstruction algorithms and kernels between different scanners.Results. The matching paradigm produced good matching results, and the mean ± standard deviation (median) matching function values for the different acquisition settings were (a value of 1 indicates a perfect match): CTDIvol 10 mGy, 1 mm slice: 0.78 ± 0.31 (0.94); CTDIvol 10 mGy, 5 mm slice: 0.75 ± 0.33 (0.93); CTDIvol 40 mGy, 1 mm slice: 0.81 ± 0.28 (0.95); CTDIvol 40 mGy, 5 mm slice: 0.75 ± 0.33 (0.93). In general, soft reconstruction kernels, i.e., noise-reducing kernels that reduce sharpness, of one vendor were matched with the soft kernels of another vendor, and vice versa for sharper kernels. Conclusions. Combined quantitative assessment of NPS and MTF allows effective strategy for harmonization of technical image quality between different CT scanners. A software was also shared to support CT image quality harmonization in other institutions.

Harmonization of in-plane resolution in CT using multiple reconstructions from single acquisitions

PURPOSE

To providea methodology that removes the spatial variability of in-plane resolution using different CT reconstructions. The methodology does not require any training, sinogram, or specific reconstruction method.

METHODS

The methodology is formulated as a reconstruction problem. The desired sharp image is modeled as an unobservable variable to be estimated from an arbitrary number of observations with spatially variant resolution. The methodology comprises three steps: (1) density harmonization, which removes the density variability across reconstructions; (2) point spread function (PSF) estimation, which estimates a spatially variant PSF with arbitrary shape; (3) deconvolution, which is formulated as a regularized least squares problem. The assessment was performed with CT scans of phantoms acquired with three different Siemens scanners (Definition AS, Definition AS+, Drive). Four low-dose acquisitions reconstructed with backprojection and iterative methods were used for the resolution harmonization. A sharp, high-dose (HD) reconstruction was used as a validation reference. The different factors affecting the in-plane resolution (radial, angular, and longitudinal) were studied with regression analysis of the edge decay (between 10% and 90% of the edge spread function (ESF) amplitude).

RESULTS

Results showed that the in-plane resolution improves remarkably and the spatial variability is substantially reduced without compromising the noise characteristics. The modulated transfer function (MTF) also confirmed a pronounced increase in resolution. The resolution improvement was also tested by measuring the wall thickness of tubes simulating airways. In all scanners, the resolution harmonization obtained better performance than the HD, sharp reconstruction used as a reference (up to 50 percentage points). The methodology was also evaluated in clinical scans achieving a noise reduction and a clear improvement in thin-layered structures. The estimated ESF and MTF confirmed the resolution improvement.

CONCLUSION

We propose a versatile methodology to reduce the spatial variability of in-plane resolution in CT scans by leveraging different reconstructions available in clinical studies. The methodology does not require any sinogram, training, or specific reconstruction, and it is not limited to a fixed number of input images. Therefore, it can be easily adopted in multicenter studies and clinical practice. The results obtained with our resolution harmonization methodology evidence its suitability to reduce the spatially variant in-plane resolution in clinical CT scans without compromising the reconstruction's noise characteristics. We believe that the resolution increase achieved by our methodology may contribute in more accurate and reliable measurements of small structures such as vasculature, airways, and wall thickness.

Privacy preserving distributed learning classifiers - Sequential learning with small sets of data

BACKGROUND

Artificial intelligence (AI) typically requires a significant amount of high-quality data to build reliable models, where gathering enough data within a single institution can be particularly challenging. In this study we investigated the impact of using sequential learning to exploit very small, siloed sets of clinical and imaging data to train AI models. Furthermore, we evaluated the capacity of such models to achieve equivalent performance when compared to models trained with the same data over a single centralized database.

METHODS

We propose a privacy preserving distributed learning framework, learning sequentially from each dataset. The framework is applied to three machine learning algorithms: Logistic Regression, Support Vector Machines (SVM), and Perceptron. The models were evaluated using four open-source datasets (Breast cancer, Indian liver, NSCLC-Radiomics dataset, and Stage III NSCLC).

FINDINGS

The proposed framework ensured a comparable predictive performance against a centralized learning approach. Pairwise DeLong tests showed no significant difference between the compared pairs for each dataset.

INTERPRETATION

Distributed learning contributes to preserve medical data privacy. We foresee this technology will increase the number of collaborative opportunities to develop robust AI, becoming the default solution in scenarios where collecting enough data from a single reliable source is logistically impossible. Distributed sequential learning provides privacy persevering means for institutions with small but clinically valuable datasets to collaboratively train predictive AI while preserving the privacy of their patients. Such models perform similarly to models that are built on a larger central dataset.

Statistical characterization of the linear attenuation coefficient in polychromatic CT scans

PURPOSE

To provide a unifying statistical model that characterizes the integrated x-ray intensity at the detector after logarithmic transformation and can be extended to the characterization of computed tomography (CT) numbers in the reconstructed image.

METHODS

We study the statistical characteristics of polyenergetic x-ray beams in the detector. Firstly, we consider the characterization of the integrated x-ray intensity at the detector through a probabilistic model (compound Poisson) that describes its statistics. We analyze its properties and derive the probabilistic distribution after the logarithmic transformation analytically. Finally, we propose a more tractable probabilistic distribution with the same features observed in the characterization, the noncentral Gamma (nc-Gamma). This distribution exhibits desirable properties for the statistical characterization across the reconstruction process. We assess the assumptions adopted in the derivation of the statistical models throughout Monte Carlo simulations and validate them with a water phantom and a lung phantom acquired in a Siemens clinical CT scan. We evaluate the statistical similarities between the theoretical distribution and the nc-Gamma using a power analysis with a Kolmogorov-Smirnov test for a 95% confidence level.

RESULTS

The Kolmogorov-Smirnov goodness-of-fit test obtained for the Monte Carlo simulation shows an extremely high agreement between the empirical distribution of the post-logarithmic-integrated x-ray intensity and the nc-Gamma. The experimental validation performed with both phantoms confirmed the excellent match between the theoretical distribution, the proposed nc-Gamma, and sample distributions in all situations.

CONCLUSION

We derive an analytical model describing the post-log distribution of the linear attenuation coefficient in the sensor for polychromatic CT scans. We also demonstrate that the nc-Gamma distribution approximates well the theoretical distribution. This distribution also approximates well the CT numbers after reconstruction since it naturally extends across linear operations involved in filtered back projection reconstructions. This probabilistic model may provide the analytical foundation to define new likelihood-based reconstruction methodologies for polychromatic scans.