In-silico study of the impact of system design parameters on microcalcification detection in wide-angle digital breast tomosynthesis.

Purpose: Accurate detection of microcalcifications ( µ Calcs ) is crucial for the early detection of breast cancer. Some clinical studies have indicated that digital breast tomosynthesis (DBT) systems with a wide angular range have inferior µ Calc detectability compared with those with a narrow angular range. This study aims to (1) provide guidance for optimizing wide-angle (WA) DBT for improving µ Calcs detectability and (2) prioritize key optimization factors. Approach: An in-silico DBT pipeline was constructed to evaluate µ Calc detectability of a WA DBT system under various imaging conditions: focal spot motion (FSM), angular dose distribution (ADS), detector pixel pitch, and detector electronic noise (EN). Images were simulated using a digital anthropomorphic breast phantom inserted with 120 µ m µ Calc clusters. Evaluation metrics included the signal-to-noise ratio (SNR) of the filtered channel observer and the area under the receiver operator curve (AUC) of multiple-reader multiple-case analysis. Results: Results showed that FSM degraded µ Calcs sharpness and decreased the SNR and AUC by 5.2% and 1.8%, respectively. Non-uniform ADS increased the SNR by 62.8% and the AUC by 10.2% for filtered backprojection reconstruction with a typical clinical filter setting. When EN decreased from 2000 to 200 electrons, the SNR and AUC increased by 21.6% and 5.0%, respectively. Decreasing the detector pixel pitch from 85 to 50 µ m improved the SNR and AUC by 55.6% and 7.5%, respectively. The combined improvement of a 50 µ m pixel pitch and EN200 was 89.2% in the SNR and 12.8% in the AUC. Conclusions: Based on the magnitude of impact, the priority for enhancing µ Calc detectability in WA DBT is as follows: (1) utilizing detectors with a small pixel pitch and low EN level, (2) allocating a higher dose to central projections, and (3) reducing FSM. The results from this study can potentially provide guidance for DBT system optimization in the future.

A deep learning anthropomorphic model observer for a detection task in PET.

BACKGROUND: Lesion detection is one of the most important clinical tasks in positron emission tomography (PET) for oncology. An anthropomorphic model observer (MO) designed to replicate human observers (HOs) in a detection task is an important tool for assessing task-based image quality. The channelized Hotelling observer (CHO) has been the most popular anthropomorphic MO. Recently, deep learning MOs (DLMOs), mostly based on convolutional neural networks (CNNs), have been investigated for various imaging modalities. However, there have been few studies on DLMOs for PET. PURPOSE: The goal of the study is to investigate whether DLMOs can predict HOs better than conventional MOs such as CHO in a two-alternative forced-choice (2AFC) detection task using PET images with real anatomical variability. METHODS: Two types of DLMOs were implemented: (1) CNN DLMO, and (2) CNN-SwinT DLMO that combines CNN and Swin Transformer (SwinT) encoders. Lesion-absent PET images were reconstructed from clinical data, and lesion-present images were reconstructed with adding simulated lesion sinogram data. Lesion-present and lesion-absent PET image pairs were labeled by eight HOs consisting of four radiologists and four image scientists in a 2AFC detection task. In total, 2268 pairs of lesion-present and lesion-absent images were used for training, 324 pairs for validation, and 324 pairs for test. CNN DLMO, CNN-SwinT DLMO, CHO with internal noise, and non-prewhitening matched filter (NPWMF) were compared in the same train-test paradigm. For comparison, six quantitative metrics including prediction accuracy, mean squared errors (MSEs) and correlation coefficients, which measure how well a MO predicts HOs, were calculated in a 9-fold cross-validation experiment. RESULTS: In terms of the accuracy and MSE metrics, CNN DLMO and CNN-SwinT DLMO showed better performance than CHO and NPWMF, and CNN-SwinT DLMO showed the best performance among the MOs evaluated. CONCLUSIONS: DLMO can predict HOs more accurately than conventional MOs such as CHO in PET lesion detection. Combining SwinT and CNN encoders can improve the DLMO prediction performance compared to using CNN only.

Low-contrast detectability of photon-counting-detector CT at different scan modes and image types in comparison with energy-integrating-detector CT.

Purpose: We aim to compare the low-contrast detectability of a clinical whole-body photon-counting-detector (PCD)-CT at different scan modes and image types with an energy-integrating-detector (EID)-CT. Approach: We used a channelized Hotelling observer (CHO) previously optimized for quality control purposes. An American College of Radiology CT accreditation phantom was scanned on both PCD-CT and EID-CT with 10 phantom positionings. For PCD-CT, images were generated using two scan modes, standard resolution (SR) and ultra-high-resolution (UHR); two image types, virtual monochromatic images at 70 keV and low-energy threshold (T3D); both filtered-back-projection (FBP) and iterative reconstruction (IR) reconstruction methods; and three reconstruction kernels. For each positioning, three repeated scans were acquired for each scan mode, image type, and CTDIvol of 6, 12, and 24 mGy. For EID-CT, images acquired from scans (10 positionings × 3 repeats × 3 doses) were reconstructed using the closest counterpart FBP and IR kernels. CHO was applied to calculate the index of detectability (d') on both scanners. Results: With the smooth Br44 kernel, the d' of UHR was mostly comparable with that of the SR mode (difference: -11.4% to 8.3%, p=0.020 to 0.956), and the T3D images had a higher d' (difference: 0.7% to 25.6%) than 70 keV images on PCD-CT. Compared with the EID-CT, UHR-T3D of PCD-CT had non-inferior d' (difference: -2.7% to 12.9%) with IR and non-superior d' (difference: 0.8% to 11.2%) with FBP using the Br44 kernel. PCD-CT produced higher d' than EID-CT by 61.8% to 247.1% with the sharper reconstruction kernels. Conclusions: The comparison between PCD-CT and EID-CT was significantly influenced by the reconstruction method and kernel. With a smooth kernel that is typically used in low-contrast detection tasks, the PCD-CT demonstrated low-contrast detectability that was comparable to EID-CT with IR and showed no superiority when using FBP. With the use of sharper kernels, the PCD-CT significantly outperformed EID-CT in low-contrast detectability.

Applying Self-Supervised Learning to Image Quality Assessment in Chest CT Imaging.

Many new reconstruction techniques have been deployed to allow low-dose CT examinations. Such reconstruction techniques exhibit nonlinear properties, which strengthen the need for a task-based measure of image quality. The Hotelling observer (HO) is the optimal linear observer and provides a lower bound of the Bayesian ideal observer detection performance. However, its computational complexity impedes its widespread practical usage. To address this issue, we proposed a self-supervised learning (SSL)-based model observer to provide accurate estimates of HO performance in very low-dose chest CT images. Our approach involved a two-stage model combining a convolutional denoising auto-encoder (CDAE) for feature extraction and dimensionality reduction and a support vector machine for classification. To evaluate this approach, we conducted signal detection tasks employing chest CT images with different noise structures generated by computer-based simulations. We compared this approach with two supervised learning-based methods: a single-layer neural network (SLNN) and a convolutional neural network (CNN). The results showed that the CDAE-based model was able to achieve similar detection performance to the HO. In addition, it outperformed both SLNN and CNN when a reduced number of training images was considered. The proposed approach holds promise for optimizing low-dose CT protocols across scanner platforms.

3D printed phantom with 12 000 submillimeter lesions to improve efficiency in CT detectability assessment.

BACKGROUND: The detectability performance of a CT scanner is difficult to precisely quantify when nonlinearities are present in reconstruction. An efficient detectability assessment method that is sensitive to small effects of dose and scanner settings is desirable. We previously proposed a method using a search challenge instrument: a phantom is embedded with hundreds of lesions at random locations, and a model observer is used to detect lesions. Preliminary tests in simulation and a prototype showed promising results. PURPOSE: In this work, we fabricated a full-size search challenge phantom with design updates, including changes to lesion size, contrast, and number, and studied our implementation by comparing the lesion detectability from a nonprewhitening (NPW) model observer between different reconstructions at different exposure levels, and by estimating the instrument sensitivity to detect changes in dose. METHODS: Designed to fit into QRM anthropomorphic phantoms, our search challenge phantom is a cylindrical insert 10 cm wide and 4 cm thick, embedded with 12 000 lesions (nominal width of 0.6 mm, height of 0.8 mm, and contrast of -350 HU), and was fabricated using PixelPrint, a 3D printing technique. The insert was scanned alone at a high dose to assess printing accuracy. To evaluate lesion detectability, the insert was placed in a QRM thorax phantom and scanned from 50 to 625 mAs with increments of 25 mAs, once per exposure level, and the average of all exposure levels was used as high-dose reference. Scans were reconstructed with three different settings: filtered-backprojection (FBP) with Br40 and Br59, and Sinogram Affirmed Iterative Reconstruction (SAFIRE) with strength level 5 and Br59 kernel. An NPW model observer was used to search for lesions, and detection performance of different settings were compared using area under the exponential transform of free response ROC curve (AUC). Using propagation of uncertainty, the sensitivity to changes in dose was estimated by the percent change in exposure due to one standard deviation of AUC, measured from 5 repeat scans at 100, 200, 300, and 400 mAs. RESULTS: The printed insert lesions had an average position error of 0.20 mm compared to printing reference. As the exposure level increases from 50 mAs to 625 mAs, the lesion detectability AUCs increase from 0.38 to 0.92, 0.42 to 0.98, and 0.41 to 0.97 for FBP Br40, FBP Br59, and SAFIRE Br59, respectively, with a lower rate of increase at higher exposure level. FBP Br59 performed best with AUC 0.01 higher than SAFIRE Br59 on average and 0.07 higher than FBP Br40 (all P < 0.001). The standard deviation of AUC was less than 0.006, and the sensitivity to detect changes in mAs was within 2% for FBP Br59. CONCLUSIONS: Our 3D-printed search challenge phantom with 12 000 submillimeter lesions, together with an NPW model observer, provide an efficient CT detectability assessment method that is sensitive to subtle effects in reconstruction and is sensitive to small changes in dose.

Probabilistic U-Net model observer for the DDC method in CT scan protocol optimization.

Optimizing complex imaging procedures within Computed Tomography, considering both dose and image quality, presents significant challenges amidst rapid technological advancements and the adoption of machine learning (ML) methods. A crucial metric in this context is the Difference-Detailed Curve, which relies on human observer studies. However, these studies are labor-intensive and prone to both inter- and intra-observer variability. To tackle these issues, a ML-based model observer utilizing the U-Net architecture and a Bayesian methodology is proposed. In order to train a model observer unaffected by the spatial arrangement of low-contrast objects, the image preprocessing incorporates a Gaussian Process-based noise model. Additionally, gradient-weighted class activation mapping is utilized to gain insights into the model observer's decision-making process. By training on data from a diverse group of observers, well-calibrated probabilistic predictions that quantify observer variability are achieved. Leveraging the principles of Beta regression, the Bayesian methodology is used to derive a model observer performance metric, effectively gauging the model observer's strength in terms of an 'effective number of observers'. Ultimately, this framework enables to predict the DDC distribution by applying thresholds to the inferred probabilities (Part of this work has been presented at: Stocker D, Sommer C, Gueng S, Stäuble J, Özden I, Griessinger J, Weyland M S, Lutters G, Scheidegger S (2023). Probabilistic U-Net Model Observer for the DDC Method in CT Scan Protocol Optimization. The 56th SSRMP Annual Meeting 2023, November 30. - December 1., 2023, Luzern, Switzerland).

Channelized hotelling observer-based low-contrast detectability on the ACR CT accreditation phantom: Part II. Repeatability study.

BACKGROUND: Objective and quantitative evaluation for low-contrast detectability that correlates with human observer performance is lacking for routine CT quality control testing. Channelized Hotelling observer (CHO) is considered a strong candidate to fill the need but has long been deemed impractical to implement due to its requirement of a large number of repeated scans in order to provide accurate and precise estimates of index of detectability (d'). In our previous work, we optimized a CHO model observer on the American College of Radiology (ACR) CT accreditation phantom and achieved accurate measurement of d' with only 1-3 repeat scans. PURPOSE: In this work, we aim to validate the repeatability of the proposed CHO-based low-contrast evaluation on four scanner models using the ACR CT accreditation phantom. METHODS: The repeatability test was performed on four different scanners from two major CT manufacturers: Siemens Force and Alpha; Canon Prism and Prime SP. An ACR CT phantom was scanned 10 times, each time after repositioning of the phantom. For each repositioning, 3 repeated scans were acquired at 24, 12, and 6 mGy on all four scanner models. CHO was applied at the measured dose levels for different low-contrast object sizes (4-6 mm). The CHO was also applied to images created using deep learning-based reconstructions on Canon Prism and to four different scan/reconstruction modes on the Siemens Alpha, a photon-counting-detector (PCD)-CT. The repeatability was evaluated by the probability that a measurement would fall within the ±15% tolerance (P<15% ). RESULTS: With the CHO setting optimized for the ACR phantom and the use of 3 repeated scans and 9 non-overlapping slices per scan, the CHO measurement could provide high repeatability with P<15% of 98.8%-99.9% at 12 mGy with IR reconstruction on all four scanners. On scanner A, P<15% were 91.5%-99.9% at the three dose levels and for all three object sizes while the numbers were 93.6%-99.998% on scanner B. P<15% were 96.5%-97.2% for the two deep learning reconstructions and 97.0%-99.97% for the four scan/reconstruction modes on the PCD-CT. CONCLUSION: The CHO provided highly repeatable measurements with over 95% probability that a CHO measurement would lie within the ±15% tolerance for most of the dose levels and object sizes on the ACR phantom. The repeatability was maintained when the CHO was applied to images created with a commercial deep learning-based reconstruction and various scan/reconstruction modes on a PCD-CT. This study demonstrates that practical implementation of CHO for routine quality control and performance evaluation is feasible.

Task-based detectability in anatomical background in digital mammography, digital breast tomosynthesis and synthetic mammography.

Objective.Determining the detectability of targets for the different imaging modalities in mammography in the presence of anatomical background noise is challenging. This work proposes a method to compare the image quality and detectability of targets in digital mammography (DM), digital breast tomosynthesis (DBT) and synthetic mammography.Approach. The low-frequency structured noise produced by a water phantom with acrylic spheres was used to simulate anatomical background noise for the different types of images. A method was developed to apply the non-prewhitening observer model with eye filter (NPWE) in these conditions. A homogeneous poly(methyl) methacrylate phantom with a 0.2 mm thick aluminium disc was used to calculate 2D in-plane modulation transfer function (MTF), noise power spectrum (NPS), noise equivalent quanta, and system detective quantum efficiency for 30, 50 and 70 mm thicknesses. The in-depth MTFs of DBT volumes were determined using a thin tungsten wire. The MTF, system NPS and anatomical NPS were used in the NPWE model to calculate the threshold gold thickness of the gold discs contained in the CDMAM phantom, which was taken as reference. Main results.The correspondence between the NPWE model and the CDMAM phantom (linear Pearson correlation 0.980) yielded a threshold detectability index that was used to determine the threshold diameter of spherical microcalcifications and masses. DBT imaging improved the detection of masses, which depended mostly on the reduction of anatomical background noise. Conversely, DM images yielded the best detection of microcalcifications.Significance.The method presented in this study was able to quantify image quality and object detectability for the different imaging modalities and levels of anatomical background noise.

Model-based deep CNN-regularized reconstruction for digital breast tomosynthesis with a task-based CNN image assessment approach.

Objective. Digital breast tomosynthesis (DBT) is a quasi-three-dimensional breast imaging modality that improves breast cancer screening and diagnosis because it reduces fibroglandular tissue overlap compared with 2D mammography. However, DBT suffers from noise and blur problems that can lower the detectability of subtle signs of cancers such as microcalcifications (MCs). Our goal is to improve the image quality of DBT in terms of image noise and MC conspicuity.Approach. We proposed a model-based deep convolutional neural network (deep CNN or DCNN) regularized reconstruction (MDR) for DBT. It combined a model-based iterative reconstruction (MBIR) method that models the detector blur and correlated noise of the DBT system and the learning-based DCNN denoiser using the regularization-by-denoising framework. To facilitate the task-based image quality assessment, we also proposed two DCNN tools for image evaluation: a noise estimator (CNN-NE) trained to estimate the root-mean-square (RMS) noise of the images, and an MC classifier (CNN-MC) as a DCNN model observer to evaluate the detectability of clustered MCs in human subject DBTs.Main results. We demonstrated the efficacies of CNN-NE and CNN-MC on a set of physical phantom DBTs. The MDR method achieved low RMS noise and the highest detection area under the receiver operating characteristic curve (AUC) rankings evaluated by CNN-NE and CNN-MC among the reconstruction methods studied on an independent test set of human subject DBTs.Significance. The CNN-NE and CNN-MC may serve as a cost-effective surrogate for human observers to provide task-specific metrics for image quality comparisons. The proposed reconstruction method shows the promise of combining physics-based MBIR and learning-based DCNNs for DBT image reconstruction, which may potentially lead to lower dose and higher sensitivity and specificity for MC detection in breast cancer screening and diagnosis.

Pre-whitened matched filter and convolutional neural network based model observer performance for mass lesion detection in non-contrast breast CT.

BACKGROUND: Mathematical model observers have been shown to reasonably predict human observer performance and are useful when human observer studies are infeasible. Recently, convolutional neural networks (CNNs) have also been used as substitutes for human observers, and studies have shown their utility as an optimal observer. In this study, a CNN model observer is compared to the pre-whitened matched filter (PWMF) model observer in detecting simulated mass lesions inserted into 253 acquired breast computed tomography (bCT) images from patients imaged at our institution. PURPOSE: To compare CNN and PWMF model observers for detecting signal-known-exactly (SKE) location-known-exactly (LKE) simulated lesions in bCT images with real anatomical backgrounds, and to use these model observers collectively to optimize parameters and understand trends in performance with breast CT. METHODS: Spherical lesions with different diameters (1, 3, 5, 9 mm) were mathematically inserted into reconstructed patient bCT image data sets to mimic 3D mass lesions in the breast. 2D images were generated by extracting the center slice along the axial dimension or by slice averaging across adjacent slices to model thicker sections (0.4, 1.2, 2.0, 6.0, 12.4, 20.4 mm). The role of breast density was retrospectively studied using the range of breast densities intrinsic to the patient bCT data sets. In addition, mass lesions were mathematically inserted into Gaussian images matched to the mean and noise power spectrum of the bCT images to better understand the performance of the CNN in the context of a known ideal observer (the PWMF). The simulated Gaussian and bCT images were divided into training and testing data sets. Each training data set consisted of 91 600 images, and each testing data set consisted of 96 000 images. A CNN and PWMF was trained on the Gaussian training images, and a different CNN and PWMF was trained on the bCT training images. The trained model observers were tested, and receiver operating characteristic (ROC) curve analysis was used to evaluate detection performance. The area under the ROC curve (AUC) was the primary performance metric used to compare the model observers. RESULTS: In the Gaussian background, the CNN performed essentially identically to the PWMF across lesion sizes and section thicknesses. In the bCT background, the CNN outperformed the PWMF across lesion size, breast density, and most section thicknesses. These findings suggest that there are higher-order features in bCT images that are harnessed by the CNN observer but are inaccessible to the PWMF. CONCLUSIONS: The CNN performed equivalently to the ideal observer in Gaussian textures. In bCT background, the CNN captures more diagnostic information than the PWMF and may be a more pertinent observer when conducting optimal performance studies in breast CT images.

Strategy to implement a convolutional neural network based ideal model observer via transfer learning for multi-slice simulated breast CT images.

Objective.In this work, we propose a convolutional neural network (CNN)-based multi-slice ideal model observer using transfer learning (TL-CNN) to reduce the required number of training samples.Approach.To train model observers, we generate simulated breast CT image volumes that are reconstructed using the FeldkampDavisKress algorithm with a ramp and Hanning-weighted ramp filter. The observer performance is evaluated on the background-known-statistically (BKS)/signal-known-exactly task with a spherical signal, and the BKS/signal-known-statistically task with random signal generated by the stochastic grown method. We compare the detectability of the CNN-based model observer with that of conventional linear model observers for multi-slice images (i.e. a multi-slice channelized Hotelling observer (CHO) and volumetric CHO). We also analyze the detectability of the TL-CNN for different numbers of training samples to examine its performance robustness to a limited number of training samples. To further analyze the effectiveness of transfer learning, we calculate the correlation coefficients of filter weights in the CNN-based multi-slice model observer.Main results.When using transfer learning for the CNN-based multi-slice ideal model observer, the TL-CNN provides the same performance with a 91.7% reduction in the number of training samples compared to that when transfer learning is not used. Moreover, compared to the conventional linear model observer, the proposed CNN-based multi-slice model observers achieve 45% higher detectability in the signal-known-statistically detection tasks and 13% higher detectability in the SKE detection tasks. In correlation coefficient analysis, it is observed that the filters in most of the layers are highly correlated, demonstrating the effectiveness of the transfer learning for multi-slice model observer training.Significance.Deep learning-based model observers require large numbers of training samples, and the required number of training samples increases as the dimensions of the image (i.e. the number of slices) increase. With applying transfer learning, the required number of training samples is significantly reduced without performance drop.

Need for objective task-based evaluation of deep learning-based denoising methods: A study in the context of myocardial perfusion SPECT.

BACKGROUND: Artificial intelligence-based methods have generated substantial interest in nuclear medicine. An area of significant interest has been the use of deep-learning (DL)-based approaches for denoising images acquired with lower doses, shorter acquisition times, or both. Objective evaluation of these approaches is essential for clinical application. PURPOSE: DL-based approaches for denoising nuclear-medicine images have typically been evaluated using fidelity-based figures of merit (FoMs) such as root mean squared error (RMSE) and structural similarity index measure (SSIM). However, these images are acquired for clinical tasks and thus should be evaluated based on their performance in these tasks. Our objectives were to: (1) investigate whether evaluation with these FoMs is consistent with objective clinical-task-based evaluation; (2) provide a theoretical analysis for determining the impact of denoising on signal-detection tasks; and (3) demonstrate the utility of virtual imaging trials (VITs) to evaluate DL-based methods. METHODS: A VIT to evaluate a DL-based method for denoising myocardial perfusion SPECT (MPS) images was conducted. To conduct this evaluation study, we followed the recently published best practices for the evaluation of AI algorithms for nuclear medicine (the RELAINCE guidelines). An anthropomorphic patient population modeling clinically relevant variability was simulated. Projection data for this patient population at normal and low-dose count levels (20%, 15%, 10%, 5%) were generated using well-validated Monte Carlo-based simulations. The images were reconstructed using a 3-D ordered-subsets expectation maximization-based approach. Next, the low-dose images were denoised using a commonly used convolutional neural network-based approach. The impact of DL-based denoising was evaluated using both fidelity-based FoMs and area under the receiver operating characteristic curve (AUC), which quantified performance on the clinical task of detecting perfusion defects in MPS images as obtained using a model observer with anthropomorphic channels. We then provide a mathematical treatment to probe the impact of post-processing operations on signal-detection tasks and use this treatment to analyze the findings of this study. RESULTS: Based on fidelity-based FoMs, denoising using the considered DL-based method led to significantly superior performance. However, based on ROC analysis, denoising did not improve, and in fact, often degraded detection-task performance. This discordance between fidelity-based FoMs and task-based evaluation was observed at all the low-dose levels and for different cardiac-defect types. Our theoretical analysis revealed that the major reason for this degraded performance was that the denoising method reduced the difference in the means of the reconstructed images and of the channel operator-extracted feature vectors between the defect-absent and defect-present cases. CONCLUSIONS: The results show the discrepancy between the evaluation of DL-based methods with fidelity-based metrics versus the evaluation on clinical tasks. This motivates the need for objective task-based evaluation of DL-based denoising approaches. Further, this study shows how VITs provide a mechanism to conduct such evaluations computationally, in a time and resource-efficient setting, and avoid risks such as radiation dose to the patient. Finally, our theoretical treatment reveals insights into the reasons for the limited performance of the denoising approach and may be used to probe the effect of other post-processing operations on signal-detection tasks.

Spatio-temporal generalized Model Observers methods for Low Contrast Detectability assessment in digital angiography: Application to moving targets.

The purpose of this work is to investigate the feasibility of spatio-temporal generalized Model Observer methods for protocol optimization programs in the field of interventional radiography. Two Model Observers were taken under examination: a Channelized Hotelling Observer with 24 spatio-temporal Gabor channels and a Non Pre-Whitening Model Observer with two different implementations of the spatio-temporal contrast sensitivity function. The images of targets, both stationary and in motion, were acquired in fluoroscopic mode using a CDRAD phantom for signal-present images and an homogenous slab of PMMA for signal-absent ones. After the processing, these images were used to build three series of two alternative forced choice experiments, designed to simulate tasks of clinical interest, and submitted to three human observers in order to set a goal on detectability. A first set of images was used for model tuning and subsequently the verified models were validated throughout a second set of images. Results from the validation phase, for both models, show good agreement with the human observer performances (Root Mean Square Error RMSE ≤ 12%). The tuning phase emerges as a crucial step in building models for angiographic dynamic images; the final agreement underlines the good capability of these spatio-temporal models in simulating human performances, allowing to consider them as a useful and worthwhile tool in protocol optimization when dynamic images are involved.

UNet and MobileNet CNN-based model observers for CT protocol optimization: comparative performance evaluation by means of phantom CT images.

Purpose: The aim of this work is the development and characterization of a model observer (MO) based on convolutional neural networks (CNNs), trained to mimic human observers in image evaluation in terms of detection and localization of low-contrast objects in CT scans acquired on a reference phantom. The final goal is automatic image quality evaluation and CT protocol optimization to fulfill the ALARA principle. Approach: Preliminary work was carried out to collect localization confidence ratings of human observers for signal presence/absence from a dataset of 30,000 CT images acquired on a PolyMethyl MethAcrylate phantom containing inserts filled with iodinated contrast media at different concentrations. The collected data were used to generate the labels for the training of the artificial neural networks. We developed and compared two CNN architectures based respectively on Unet and MobileNetV2, specifically adapted to achieve the double tasks of classification and localization. The CNN evaluation was performed by computing the area under localization-ROC curve (LAUC) and accuracy metrics on the test dataset. Results: The mean of absolute percentage error between the LAUC of the human observer and MO was found to be below 5% for the most significative test data subsets. An elevated inter-rater agreement was achieved in terms of S-statistics and other common statistical indices. Conclusions: Very good agreement was measured between the human observer and MO, as well as between the performance of the two algorithms. Therefore, this work is highly supportive of the feasibility of employing CNN-MO combined with a specifically designed phantom for CT protocol optimization programs.

Accurate and efficient measurement of channelized Hotelling observer-based low-contrast detectability on the ACR CT accreditation phantom.

BACKGROUND: Current CT quality control (QC) for low-contrast detectability relies on visual inspection and measurement of contrast-to-noise ratio (CNR). However, CNR numbers become unreliable when it comes to nonlinear methods, such as iterative reconstruction (IR) and deep-learning-based techniques. Image quality metrics using channelized Hotelling observer (CHO) have been validated to be well correlated with human observer performance on phantom-based and patient-based tasks, but it has not been widely used in routine CT QC mainly because the CHO calculation typically requires a large number of repeated scans in order to provide accurate and precise estimate of index of detectability (d'). PURPOSE: The main goal of this work is to optimize channel filters and other CHO parameters and accurately estimate the low-contrast detectability with minimum number of repeated scans for the widely used American College of Radiology (ACR) CT accreditation phantom so that it can become practically feasible for routine CT QC tests. METHODS: To provide a converged d' value, an ACR phantom was repeatedly scanned 100 times at three dose levels (24, 12, and 6 mGy). Images were reconstructed with two kernels (FBP Br44 and IR Br44-3). d' as a function of number of repeated scans was determined for different number of background regions of interest (ROIs), different number of low-contrast objects, different number of slices per each object, and different channel filter options. A reference d' was established using the optimized CHO setting, and the bias of d' was quantified using the d' calculated from all 100 repeated scans. The variation of d' at each condition was estimated using a resampling method combining random subsampling among 100 repeated scans and bootstrapping of the ensembles of signal and background ROIs. RESULTS: Optimized parameters in CHO calculation were determined: two background ROIs per object, four objects per low-contrast object size, nine non-overlapping slices per object, and a 4-channel Gabor filter. The bias and uncertainty were estimated at different numbers of repeated scans using these parameters. When only one single scan was used in the CHO calculation, the bias of d' was below 6.2% and the uncertainty 15.6-19.6% for the 6, 5, and 4 mm objects, while with three repeated scans the bias was below 2.0% and uncertainty 8.7-10.9% for the three object sizes. CONCLUSION: With optimized parameter settings in CHO, efficient and accurate measurement of low-contrast detectability on the commonly used ACR phantom becomes feasible, which could potentially lead to adoption of CHO-based low-contrast evaluation in routine QC tests.

Experimental feasibility of xenon-enhanced dual-energy radiography for imaging of lung function.

Chronic obstructive pulmonary disease (COPD) is a leading cause of death worldwide. We experimentally investigated the feasibility of two-dimensional xenon-enhanced dual-energy (XeDE) radiography for imaging of lung function. We optimized image quality under quantum-noise-limited conditions using a chest phantom consisting of a rectangular chamber representing the thoracic volume and PMMA slabs simulating x-ray attenuation by soft tissue. A sealed, air-filled cavity with thin PMMA walls was positioned inside the chamber to simulate a 2 cm thick ventilation defect. The chamber was ventilated with xenon and dual-energy imaging was performed using a diagnostic x-ray tube and a flat-panel detector. The contrast-to-noise ratio of ventilation defects normalized by patient x-ray exposure maximized at a kV-pair of approximately 60/140-kV and when approximately one third of the total exposure was allocated to the HE image. We used the optimized technique to image a second phantom that contained lung-parenchyma-mimicking PMMA clutter, rib-mimicking aluminum slats and an insert that simulated ventilation defects with thicknesses ranging from 0.5 cm to 2 cm and diameters ranging from 1 cm to 2 cm. From the resulting images we computed the area under the receiver operating characteristic curve (AUC) of the non-prewhitening model observer with an eye filter and internal noise. For a xenon concentration of 75%, good AUCs (i.e. 0.8-0.9) to excellent AUCs (i.e. >0.9) were obtained when the defect diameter is greater than 1.3 cm and defect thickness is 1 cm. When the xenon concentration was reduced to 50%, the AUC was ∼0.9 for defects 1.2 cm in diameter and ∼1.5 cm in thickness. Two-dimensional XeDE radiography may therefore enable detection of functional abnormalities associated with early-stage COPD, for which xenon ventilation defects can occupy up to 20% of the lung volume, and should be further developed as a low-cost alternative to MRI-based approaches and a low-dose alternative to CT-based approaches.

Investigating the limited performance of a deep-learning-based SPECT denoising approach: An observer-study-based characterization.

Multiple objective assessment of image-quality (OAIQ)-based studies have reported that several deep-learning (DL)-based denoising methods show limited performance on signal-detection tasks. Our goal was to investigate the reasons for this limited performance. To achieve this goal, we conducted a task-based characterization of a DL-based denoising approach for individual signal properties. We conducted this study in the context of evaluating a DL-based approach for denoising single photon-emission computed tomography (SPECT) images. The training data consisted of signals of different sizes and shapes within a clustered-lumpy background, imaged with a 2D parallel-hole-collimator SPECT system. The projections were generated at normal and 20% low-count level, both of which were reconstructed using an ordered-subset-expectation-maximization (OSEM) algorithm. A convolutional neural network (CNN)-based denoiser was trained to process the low-count images. The performance of this CNN was characterized for five different signal sizes and four different signal-to-background ratio (SBRs) by designing each evaluation as a signal-known-exactly/background-known-statistically (SKE/BKS) signal-detection task. Performance on this task was evaluated using an anthropomorphic channelized Hotelling observer (CHO). As in previous studies, we observed that the DL-based denoising method did not improve performance on signal-detection tasks. Evaluation using the idea of observer-study-based characterization demonstrated that the DL-based denoising approach did not improve performance on the signal-detection task for any of the signal types. Overall, these results provide new insights on the performance of the DL-based denoising approach as a function of signal size and contrast. More generally, the observer study-based characterization provides a mechanism to evaluate the sensitivity of the method to specific object properties, and may be explored as analogous to characterizations such as modulation transfer function for linear systems. Finally, this work underscores the need for objective task-based evaluation of DL-based denoising approaches.

A statistical alternative to current measures of image quality in digital mammography.

Objective. Mammogram image quality in European breast screening systems is defined by threshold gold thickness (T) assessment of the CDMAM contrast-detail phantom. Previous studies have outlined several limitations of the phantom including expense, number of images required and inter-phantom manufacturing variability. Two alternative approaches to image quality assessment for routine quality control are examined and compared to the CDMAM technique: (i) A detectability index (d') based on a non-prewhitened model observer with an eye filter (NPWE) and(ii) A statistical estimate of contrast based on image noise levels (CSTAT).Approach. Thed' calculation follows a previously published methodology based on the NNPS and contrast, both measured from an image of 5 cm of PMMA containing a 0.2 mm Al target, as well as the MTF measured under standard conditions. For the proposed statistical method, pixels in the centre of the same NNPS image were re-binned into a range of equivalent CDMAM target areas. For any area, the minimum contrast necessary to distinguish a signal from the background,CSTAT, is 3.29σat a 95% level of confidence, whereσis the standard deviation of the background pixels. Theoretical analysis predicts a simple relationships betweenCSTAT,Tandd'. Measured values ofCSTATwere compared toTandd' as a function of air kerma at the detector for ten digital mammography systems from three different manufacturers.Main Results. Theoretical relationships betweenCSTAT,d' andTwere demonstrated. Minimum acceptable image quality performance for 0.10 and 0.25 mm diameter discs, defined by the European Guidelines in terms ofT, are equivalent tod' values of 0.85 and 5.36 and thresholdCSTATvalues of 0.055 and 0.022.Significance. Strong correlations between log(T), log(d') and log(CSTAT) suggest that either alternative approach produces information corresponding to that obtained using the CDMAM.CSTATshould be considered as a simple, objective and cost-effective alternative to routine image quality assessment in mammography.

Deep-learning model observer for a low-contrast hepatic metastases localization task in computed tomography.

PURPOSE: Conventional model observers (MO) in CT are often limited to a uniform background or varying background that is random and can be modeled in an analytical form. It is unclear if these conventional MOs can be readily generalized to predict human observer performance in clinical CT tasks that involve realistic anatomical background. Deep-learning-based model observers (DL-MO) have recently been developed, but have not been validated for challenging low contrast diagnostic tasks in abdominal CT. We consequently sought to validate a DL-MO for a low-contrast hepatic metastases localization task. METHODS: We adapted our recently developed DL-MO framework for the liver metastases localization task. Our previously-validated projection-domain lesion-/noise-insertion techniques were used to synthesize realistic positive and low-dose abdominal CT exams, using the archived patient projection data. Ten experimental conditions were generated, which involved different lesion sizes/contrasts, radiation dose levels, and image reconstruction types. Each condition included 100 trials generated from a patient cohort of 7 cases. Each trial was presented as liver image patches (160×160×5 voxels). The DL-MO performance was calculated for each condition and was compared with human observer performance, which was obtained by three sub-specialized radiologists in an observer study. The performance of DL-MO and radiologists was gauged by the area under localization receiver-operating-characteristic curves. The generalization performance of the DL-MO was estimated with the repeated twofold cross-validation method over the same set of trials used in the human observer study. A multi-slice Channelized Hoteling Observers (CHO) was compared with the DL-MO across the same experimental conditions. RESULTS: The performance of DL-MO was highly correlated to that of radiologists (Pearson's correlation coefficient: 0.987; 95% CI: [0.942, 0.997]). The performance level of DL-MO was comparable to that of the grouped radiologists, that is, the mean performance difference was -3.3%. The CHO performance was poorer than the grouped radiologist performance, before internal noise could be added. The correlation between CHO and radiologists was weaker (Pearson's correlation coefficient: 0.812, and 95% CI: [0.378, 0.955]), and the corresponding performance bias (-29.5%) was statistically significant. CONCLUSION: The presented study demonstrated the potential of using the DL-MO for image quality assessment in patient abdominal CT tasks.

The regression detectability index RDI for mammography images of breast phantoms with calcification-like objects and anatomical background.

Currently, quality assurance measurements in mammography are performed on unprocessed images. For diagnosis, however, radiologists are provided with processed images. This image processing is optimised for images of human anatomy and therefore does not always perform satisfactorily with technical phantoms. To overcome this problem, it may be possible to use anthropomorphic phantoms reflecting the anatomic structure of the human breast in place of technical phantoms when carrying out task-specific quality assessment using model observers. However, the use of model observers is hampered by the fact that a large number of images needs to be acquired. A recently published novel observer called the regression detectability index (RDI) needs significantly fewer images, but requires the background of the images to be flat. Therefore, to be able to apply the RDI to images of anthropomorphic phantoms, the anatomic background needs to be removed. For this, a procedure in which the anatomical structures are fitted by thin plate spline (TPS) interpolation has been developed. When the object to be detected is small, such as a calcification-like lesion, it is shown that the anatomic background can be removed successfully by subtracting the TPS interpolation, which makes the background-free image accessible to the RDI. We have compared the detectability obtained by the RDI with TPS background subtraction to results of the channelized Hotelling observer (CHO) and human observers. With the RDI, results for the detectabilityd'can be obtained using 75% fewer images compared to the CHO, while the same uncertainty ofd'is achieved. Furthermore, the correlation ofd'(RDI) with the results of human observers is at least as good as that ofd'(CHO) with human observers.