Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function.

In CT imaging of the head, multiple image series are routinely reconstructed with different kernels and slice thicknesses. Reviewing the redundant information is an inefficient process for radiologists. We address this issue with a convolutional neural network (CNN)-based technique, synthesiZed Improved Resolution and Concurrent nOise reductioN (ZIRCON), that creates a single, thin, low-noise series that combines the favorable features from smooth and sharp head kernels. ZIRCON uses a CNN model with an autoencoder U-Net architecture that accepts two input channels (smooth- and sharp-kernel CT images) and combines their salient features to produce a single CT image. Image quality requirements are built into a task-based loss function with a smooth and sharp loss terms specific to anatomical regions. The model is trained using supervised learning with paired routine-dose clinical non-contrast head CT images as training targets and simulated low-dose (25%) images as training inputs. One hundred unique de-identified clinical exams were used for training, ten for validation, and ten for testing. Visual comparisons and contrast measurements of ZIRCON revealed that thinner slices and the smooth-kernel loss function improved gray-white matter contrast. Combined with lower noise, this increased visibility of small soft-tissue features that would be otherwise impaired by partial volume averaging or noise. Line profile analysis showed that ZIRCON images largely retained sharpness compared to the sharp-kernel input images. ZIRCON combined desirable image quality properties of both smooth and sharp input kernels into a single, thin, low-noise series suitable for both brain and skull imaging.

Pediatric evaluations for deep learning CT denoising.

BACKGROUND: Deep learning (DL) CT denoising models have the potential to improve image quality for lower radiation dose exams. These models are generally trained with large quantities of adult patient image data. However, CT, and increasingly DL denoising methods, are used in both adult and pediatric populations. Pediatric body habitus and size can differ significantly from adults and vary dramatically from newborns to adolescents. Ensuring that pediatric subgroups of different body sizes are not disadvantaged by DL methods requires evaluations capable of assessing performance in each subgroup. PURPOSE: To assess DL CT denoising in pediatric and adult-sized patients, we built a framework of computer simulated image quality (IQ) control phantoms and evaluation methodology. METHODS: The computer simulated IQ phantoms in the framework featured pediatric-sized versions of standard CatPhan 600 and MITA-LCD phantoms with a range of diameters matching the mean effective diameters of pediatric patients ranging from newborns to 18 years old. These phantoms were used in simulating CT images that were then inputs for a DL denoiser to evaluate performance in different sized patients. Adult CT test images were simulated using standard-sized phantoms scanned with adult scan protocols. Pediatric CT test images were simulated with pediatric-sized phantoms and adjusted pediatric protocols. The framework's evaluation methodology consisted of denoising both adult and pediatric test images then assessing changes in image quality, including noise, image sharpness, CT number accuracy, and low contrast detectability. To demonstrate the use of the framework, a REDCNN denoising model trained on adult patient images was evaluated. To validate that the DL model performance measured with the proposed pediatric IQ phantoms was representative of performance in more realistic patient anatomy, anthropomorphic pediatric XCAT phantoms of the same age range were also used to compare noise reduction performance. RESULTS: Using the proposed pediatric-sized IQ phantom framework, size differences between adult and pediatric-sized phantoms were observed to substantially influence the adult trained DL denoising model's performance. When applied to adult images, the DL model achieved a 60% reduction in noise standard deviation without substantial loss in sharpness in mid or high spatial frequencies. However, in smaller phantoms the denoising performance dropped due to different image noise textures resulting from the smaller field of view (FOV) between adult and pediatric protocols. In the validation study, noise reduction trends in the pediatric-sized IQ phantoms were found to be consistent with those found in anthropomorphic phantoms. CONCLUSION: We developed a framework of using pediatric-sized IQ phantoms for pediatric subgroup evaluation of DL denoising models. Using the framework, we found the performance of an adult trained DL denoiser did not generalize well in the smaller diameter phantoms corresponding to younger pediatric patient sizes. Our work suggests noise texture differences from FOV changes between adult and pediatric protocols can contribute to poor generalizability in DL denoising and that the proposed framework is an effective means to identify these performance disparities for a given model.

An FDA Guide on Indications for Use and Device Reporting of Artificial Intelligence-Enabled Devices: Significance for Pediatric Use.

Radiology has been a pioneer in adopting artificial intelligence (AI)-enabled devices into the clinic. However, initial clinical experience has identified concerns of inconsistent device performance across different patient populations. Medical devices, including those using AI, are cleared by the FDA for their specific indications for use (IFUs). IFU describes the disease or condition the device will diagnose or treat, including a description of the intended patient population. Performance data evaluated during the premarket submission support the IFU and include the intended patient population. Understanding the IFUs of a given device is thus critical to ensuring that the device is used properly and performs as expected. When devices do not perform as expected or malfunction, medical device reporting is an important way to provide feedback about the device to the manufacturer, the FDA, and other users. This article describes the ways to retrieve the IFU and performance data information as well as the FDA medical device reporting systems for unexpected performance discrepancy. It is crucial that imaging professionals, including radiologists, know how to access and use these tools to improve the informed use of medical devices for patients of all ages.

Empirical beam hardening and ring artifact correction for x-ray grating interferometry (EBHC-GI).

PURPOSE: Talbot-Lau grating interferometry enables the use of polychromatic x-ray sources, extending the range of potential applications amenable to phase contrast imaging. However, these sources introduce beam hardening effects not only from the samples but also from the gratings. As a result, grating inhomogeneities due to manufacturing imperfections can cause spectral nonuniformity artifacts when used with polychromatic sources. Consequently, the different energy dependencies of absorption, phase, and visibility contrasts impose challenges that so far have limited the achievable image quality. The purpose of this work was to develop and validate a correction strategy for grating-based x-ray imaging that accounts for beam hardening generated from both the imaged object and the gratings. METHODS: The proposed two-variable polynomial expansion strategy was inspired by work performed to address beam hardening from a primary modulator. To account for the multicontrast nature of grating interferometry, this approach was extended to each contrast to obtain three sets of correction coefficients, which were determined empirically from a calibration scan. The method's feasibility was demonstrated using a tabletop Talbot-Lau grating interferometer micro-computed tomography (CT) system using CT acquisitions of a water sample and a silicon sample, representing low and high atomic number materials. Spectral artifacts such as cupping and ring artifacts were quantified using mean squared error (MSE) from the beam-hardening-free target image and standard deviation within a reconstructed image of the sample. Finally, the model developed using the water sample was applied to a fixated murine lung sample to demonstrate robustness for similar materials. RESULTS: The water sample's absorption CT image was most impacted by spectral artifacts, but following correction to decrease ring artifacts, an 80% reduction in MSE and 57% reduction in standard deviation was observed. The silicon sample created severe artifacts in all contrasts, but following correction, MSE was reduced by 94% in absorption, 96% in phase, and 90% in visibility images. These improvements were due to the removal of ring artifacts for all contrasts and reduced cupping in absorption and phase images and reduced capping in visibility images. When the water calibration coefficients were applied to the lung sample, ring artifacts most prominent in the absorption contrast were eliminated. CONCLUSIONS: The described method, which was developed to remove artifacts in absorption, phase, and normalized visibility micro-CT images due to beam hardening in the system gratings and imaged object, reduced the MSE by up to 96%. The method depends on calibrations that can be performed on any system and does not require detailed knowledge of the x-ray spectrum, detector energy response, grating attenuation properties and imperfections, or the geometry and composition of the imaged object.

Forward model for propagation-based x-ray phase contrast imaging in parallel- and cone-beam geometry.

We demonstrate a fast, flexible, and accurate paraxial wave propagation model to serve as a forward model for propagation-based X-ray phase contrast imaging (XPCI) in parallel-beam or cone-beam geometry. This model incorporates geometric cone-beam effects into the multi-slice beam propagation method. It enables rapid prototyping and is well suited to serve as a forward model for propagation-based X-ray phase contrast tomographic reconstructions. Furthermore, it is capable of modeling arbitrary objects, including those that are strongly or multi-scattering. Simulation studies were conducted to compare our model to other forward models in the X-ray regime, such as the Mie and full-wave Rytov solutions.