Medical X-ray Research Articles

Risk communication and public response to potential radiation emergencies in New York City.

Public knowledge and awareness about radiation (both natural and manmade) tend to be low, while perceived risk of radiation tends to be high. High perceptions of risk associated with radiation have been linked to the affect heuristic and general feelings of dread, which are often not proportionate to actual risk. For example, studies routinely show that members of the public rate the risk of radiation from nuclear power plants as significantly higher (and riskier) than radiation from medical X-rays. This disconnect can have implications for adoption of protective actions during a potential radiation emergency and the perceived efficacy of these actions. This study explores how risk communication efforts influence public risk perceptions, intended protective action, and perceived efficacy of those actions. Using unique data from a survey of New York City adults, we analyze how information provision using different formats-no information, an infographic, an informational video-impact perceptions and response to a hypothetical radiation emergency. We hypothesize that respondents who receive some information, either through the infographic or the video, will have higher perceived efficacy and are more likely to take protective action. Findings suggest that providing information about what to do during a radiation emergency has a statistically significant impact on both perceived efficacy and adoption of protective action. Respondents who saw the informational video were most likely to say that they would take the correct protective actions and had the highest perceived efficacy, followed by those who saw the infographic.

Development and justification of a method for presenting radiation risks associated with medical exposure of patients

The practical implementation of radiation protection principles in the medical exposure of patients often tends to fall into one of two extremes: either excessive simplification of the methodology for assessing radiation health detriment, or its excessive complexity. An example of an excessively simplified approach is the assessment of radiation risk to patients using effective dose and radiation damage coefficients (nominal risk coefficients) as presented in Norms of the Radiation Safety NRB-99/2009. An example of an unjustifiably complex approach can be considered Tables 1-2 in MR 2.6.1.0215-20 "Assessment of radiation risk to patients during X-ray diagnostic radiological examinations," which indicate "lifetime risk values of death taking into account harm from reduced quality of life due to cancer of various organs and tissues and genetic effects from medical examinations" for a wide range of medical diagnostic X-ray radiological examinations in five-year age groups of patients. The main shortcomings of the simplified approach can be considered the lack of differences in risk assessment between individuals of different sexes and ages, although the fact of higher radiosensitivity in children compared to adults and in women compared to men can be considered universally recognized. The risk assessment approach proposed in MR 2.6.1.0215-20 addresses these shortcomings. However, in the view of the authors of this study, it offers an unnecessarily detailed picture considering uncertainties inherent in risk assessments at low doses, as well as uncertainties in the method of interpopulation transfer of radiation risk proposed by the International Commission on Radiological Protection. The aim of this study was to develop and justify a simpler and more straightforward method of presenting information on radiation risks associated with medical X-ray radiological examinations, free from the main drawbacks of the two aforementioned methods. To achieve this goal, radiation risks were calculated using two methods (using effective dose and using the risk model of the International Commission on Radiological Protection). A comparative analysis of the calculation results was conducted with estimates presented in Tables 1-2 of MR 2.6.1.0215-20. As a result of the analysis, an original applied method for presenting qualitative characteristics of radiation risks was developed for use in prescribing X-ray radiological examinations and informing patients about potential health risks. The practical outcome of the study is the formation of a table of radiation risks associated with conducting studies on patients from the Russian population, using the developed method of presenting information on radiation risks.

Single activator doped fluoride nanoparticles for X-ray excited high-resolution and delayed flexible nonplanar imaging

Lanthanide doped fluoride nanoparticles (NPs) exhibit adjustable X-ray excited optical/persistent luminescence (XEOL/XEPL) and possessing extensive potential in XEPL-based delay imaging. However, the high-resolution and delayed XEPL-based delay imaging always requires strict introduction of sensitized or heterovalent ions and the construction of complex structures for excellent XEPL performance of lanthanide-doped fluoride NPs, which is attributed to a lack of effective and convenient strategies. Here we show that the intense XEPL strength was obtained by single Tb3+ activated ion doping and simple CsY2F7@NaYF4 core/shell structure for X-ray excited high-resolution and delayed flexible nonplanar imaging. The strict introduction and control of sensitized or heterovalent ions, and the construction of intricate multilayer nanostructure are avoided. In addition, we enable the NPs to achieve high and long-term XEPL performance by simply thermal processing strategy. We successfully mixed minimized amount of the NPs into flexible film to create a scintillation screen which consequently exhibit high surface smoothness, environmental stability, and transparency. Our results demonstrate high resolution delayed imaging of the NPs integrated flexible film for the interior of nonplanar objects. The spatial resolution of 14.3 lp mm−1 is substantially higher than the typical spatial resolutions required for medical CT scans (2 lp mm−1) and X-ray imaging (3.5 lp mm−1). These findings will promote the development of advanced delayed X-ray nonplanar imaging applications.

Output Voltage Ripple Suppression of PPSS-LCC Converter for Medical X-Ray Application Based on Repetitive Control

Output Voltage Ripple Suppression of <i>PPSS</i>-<i>LCC</i> Converter for Medical X-Ray Application Based on Repetitive Control

Insights into enhanced learning through virtual reality

IntroductionThis study investigates the impact of Immersive Virtual Reality (VR) on learning outcomes in radiography training using Virtual Medical Coaching's X-Ray Pro VR software. VR offers immersive experiential learning that may enhance academic performance and professional preparedness. MethodsThe study involved 130 students from two consecutive cohorts: one using traditional training methods and the other using VR software. We employed a mixed-methods approach, integrating quantitative measures of academic performance and clinical competency with qualitative insights into student engagement, enjoyment, and stress levels. The VR group used Virtual Medical Coaching simulation software, while the traditional group used physical simulation machines (Siemens Healthineers) to practice positioning anthropomorphic manikins, adjusting exposure settings, and handling X-ray equipment. Data were collected at four points: immediately post-training and at 1 month, 3 months, and 8 months follow-up. ResultsThe VR cohort showed significantly higher knowledge retention, work readiness, and lower stress levels compared to the traditional cohort. Among the participants trained in VR, a vast majority (59 out of 65) reported a high level of satisfaction with the immersive nature of the training, emphasizing its realism and direct applicability to clinical scenarios. Quantitative analysis revealed superior performance metrics for the VR group across all measured time points (p < 0.001). Qualitative feedback highlighted enhanced engagement and enjoyment among VR-trained students. ConclusionThe results of this study clearly indicate that VR training significantly enhances learning outcomes in radiography education. The VR cohort exhibited higher knowledge retention, work readiness, student engagement, and enjoyment, along with lower stress levels and higher confidence compared to the traditional cohort. These findings support the integration of VR into professional training curricula to improve student performance and engagement.

Visualization of spatial dose distribution for effective radiation protection education in interventional radiology: obtaining high-accuracy spatial doses.

In recent years, eye lens exposure among radiation workers has become a serious concern in medical X-ray fluoroscopy and interventional radiology (IVR), highlighting the need for radiation protection education and training. This study presents a method that can maintain high accuracy when calculating spatial dose distributions obtained via Monte Carlo simulation and establishes another method to three-dimensionally visualize radiation using the obtained calculation results for contributing to effective radiation-protection education in X-ray fluoroscopy and IVR. The Monte Carlo particle and heavy ion transport code system (PHITS, Ver. 3.24) was used for calculating the spatial dose distribution generated by an angiography device. We determined the peak X-ray tube voltage and half value layer using Raysafe X2 to define the X-ray spectrum from the source and calculated the X-ray spectrum from the measured results using an approximation formula developed by Tucker et al. Further, we performed measurements using the "jungle-gym" method under the same conditions as the Monte Carlo calculations for verifying the accuracy of the latter. An optically stimulated luminescence dosimeter (nanoDot dosimeter) was used as the measuring instrument. In addition, we attempted to visualize radiation using ParaView (version 5.12.0-RC2) using the spatial dose distribution confirmed by the above calculations. A comparison of the measured and Monte Carlo calculated spatial dose distributions revealed that some areas showed large errors (12.3 and 24.2%) between the two values. These errors could be attributed to the scattering and absorption of X-rays caused by the jungle gym method, which led to uncertain measurements, and (2) the angular and energy dependencies of the nanoDot dosimetry. These two causes explain the errors in the actual values, and thus, the Monte Carlo calculations proposed in this study can be considered to have high-quality X-ray spectra and high accuracy. We successfully visualized the three-dimensional spatial dose distribution for direct and scattered X-rays separately using the obtained spatial dose distribution. We established a method to verify the accuracy of Monte Carlo calculations performed through the procedures considered in this study. Various three-dimensional spatial dose distributions were obtained with assured accuracy by applying the Monte Carlo calculation (e.g., changing the irradiation angle and adding a protective plate). Effective radiation-protection education can be realized by combining the present method with highly reliable software to visualize dose distributions.

Development of fetal magnetic resonance imaging (MRI) phantom.

Anthropomorphic phantoms play an important role in routine clinical practice. They can be used to calibrate magnetic resonance imaging (MRI) scanners, control the diagnostic equipment quality, and reduce the acquisition time. The latter is especially critical for diagnosing fetal anomalies, which requires optimal image quality within the shortest possible time. This paper aims to develop an MRI fetal phantom and determine the materials that best mimic the magnetic resonance (MR) characteristics of its internal organs. Future phantom features will include simulations of fetal limb movements. A single MRI study of a pregnant woman at 20 weeks 3 days of gestation was used as a reference and for image segmentation. Anonymized Digital Imaging and Communication in Medicine (DICOM) files were imported into 3D Slicer v. 5.2.1 for segmentation of the uterus, fetus, and internal organs. Based on the performed segmentation, a three-dimensional model was obtained for printing on a 3D printer. The mold was 3D printed on an Anycubic Photon M3 Max printer. The paper showcases the selection and manufacturing of compositions to simulate the relaxation times of the fetal organs. Formulations for emulsions and carrageenan- and agar-based hydrogels are presented. The selected compositions were used to fill the 3D printed model. Statistical analysis showed no significant differences in absolute and relative signal values obtained from scans of a pregnant woman at 20 weeks and 3 days and a fetal phantom. During the study, an anthropomorphic fetal phantom was constructed, filled with compositions with relaxation times T1 and T2 similar to the control values of the corresponding tissues. The phantom can be used to set up and optimize fetal MRI protocols, train and educate medical students, residents, graduate students, and X-ray technicians, as well as to timely control image quality and equipment serviceability.

Assessment of foetal dose and occupational exposure for pregnant workers in nuclear medicine using the Taiwanese pregnancy phantom.

For pregnant workers in nuclear medicine, radiation doses can pose a risk to their foetus. However, foetal radiation doses cannot be measured directly. In this study, a method of estimating foetal radiation doses was developed through simulations and measurements of phantoms of pregnant women in the three trimesters. The uterus and abdominal surface doses for monoenergetic photons (137Cs) and medical diagnostic X-rays were measured, and uterine dose conversion coefficients (UDCCs) were calculated. The accuracy of the UDCC estimates were validated for measurements from thermoluminescent dosemeter (TLD) chips and TLD badges on the abdomen or chest. The foetal effective dose could be estimated using TLD chips and TLD badges on the abdomen or chest, or through literature estimation method. The proposed method can be used to easily and accurately estimate foetal effective doses from chest-worn TLD badges, ensuring accurate estimation in the early stage of pregnancy when a worker may not yet be wearing an abdominal badge. A flowchart for applying the UDCC method to approximate a foetal dose is also provided to ensure that total doses remain below the maximum of 1mSv recommended in the International Commission on Radiological Protection 103 guidelines.

Radiation shielding calculation for interventional radiology: An updated workload survey using a dose monitoring software

Shielding design is an essential aspect of radiation protection. It is necessary to ensure that barriers safeguard workers, patients, the general public, and the environment from the harmful radiation emitted by X-ray machines. The National Council on Radiation Protection and Measurements (NCRP) 147 method is widely accepted within the radiation protection experts’ (RPEs) community for structural shielding design for medical X-ray imaging facilities. However, these indications are based on data collected in 1996. In recent years, interventional radiology procedures have seen significant developments. Therefore, it is important to evaluate whether updating the data on workload in the different specialities is necessary.We extracted all interventional radiology exposure data parameters from three angiographs from two vendors using dose monitoring software for 3066 procedures and 214,697 individual exposures. The workload distribution as a function of the kVp for five interventional rooms was calculated by summing all exposures and then normalising them by the number of patients.Analysing this data, we obtained new transmission curves through lead, concrete and gypsum wallboard, finding the parameters (α, β, and γ) in the Archer equation for the secondary radiation. Finally, our aim was to share an example of shielding calculations for haemodynamics and neuroangiography rooms to illustrate the impact of updated transmission data.

Medical X-ray images enhancement based on super resolution convolution neural network

&lt;p&gt;Pneumonia is a severe lung infection, chest X-ray (CXR) image preferred to find infection. Real images lost its quality, resolution and other feature due to transmission. So good qualitative datasets are very limited. Quality enhancement in medical images is challenging task for researchers. And quality in clinical diagnosis of any disease in deep learning play a very important role. So, this paper presents an aspect with importance of quality in medical images CXR of a particular dataset and how to enhance and create new images with high quality resolution, that is re-used for classification in deep learning. Super resolution convolutional neural netwok (SRCNN) is deep learning based method, which is used for improving resolution in image. Super resolution means low resolution (LR) images from dataset is to be reconstructed or magnified into high resolution (HR). The objective behind this study is to measure the effect of super resolution with quality index, peak signal-to-noise ratio (PSNR), mean squared error (MSE), and structural similarity index measure (SSIM). This experinment performed on 200 images with 10 batches, each batch has 20 images from Kermany dataset, select LR images and converted into HR with SRCNN. Then we find PSNR value of image is increase upto 2 to 5 DB, and MSE of qood quality images is near to zero and MSE decrease up to 20-25, SSIM value have little variation due to same pattern is found in input and output images. Enhancement means highlight or improve the region of interest of pneumonic images. Main goal of this study is to preapare a modified dataset which is further used for classification.&lt;/p&gt;

Special MCP-N thermoluminescent detectors (TLD) with better angular dependence

One of the basic parameters in the use of thermoluminescent detectors is the angular characteristics, i.e. receiving different signals depending on the angle of radiation exposure. The TLD detector is typically installed in a slide/badge and here the angular characteristics may have an influence, adding non-uniform coverage of the detectors with filters for the correct determination of the dose as a function of energy, because the energy characteristics are usually not flat.As part of the work, special MCP-N (LiF:Mg, Cu, P) detectors of various shapes were prepared to improve the angular characteristics. The detectors were round pellets with the same circular shape with a diameter of 4.5 mm, and a modified form inside (large and small drilled holes) or a modified surface (cavity or hemisphere). For comparison, standard MCP-N type detectors were used. Detectors in special boxes were exposed to X-rays with an energy of 80 keV. The radiation beam was formed by a medical X-ray apparatus adapted for experiments at the Warsaw University of Technology in Warsaw. Exposures were made for angles of 0, 30, 45, 50, and 90° for a dose of 1.5 mGy.Results show that it is possible to improve the angular characteristics for detectors with modified shapes. Better angular characteristics will allow for more accurate measurements, in particular in comparison tests, for accreditation and other calibrations. Especially improved angular dependence detectors could be useful for H*(10) environmental measurements.

Developing a novel radiosensitive metal ion-loaded PVA/NBT film as medical low-dose X-ray dosimeter

The hazards associated with ionizing radiation exposure has spurred the advancement of reliable and precise dosimetry techniques. Currently, there still lacks a viable and efficient dosimetry methods for measuring radiation expo-sure within low dosage ranges. This raises a concern in certain domains, such as radiodiagnostics and practitioners that employ low-dose X-ray radiation. Hence, this study aims to produce PVA/NBT film dosimeters that are capable in effectively detecting low dosages of X-ray radiation. Moreover, PVA/NBT films loaded by various metal ions will also be studied. The films were synthesized by the solvent casting method. The films were subjected to X-ray irradiation with a dosage ranging from 0 to 2.626 mGy. The films were further examined through EDX analysis, UV–Vis spectroscopy, optical density measurement, FTIR spectroscopy, and SEM analysis. It was observed that X-ray radiation exposure caused a broadening in the UV–Vis absorbance spectrum, which enhances its light absorption capability and making it more opaque. Furthermore, the careful selection of metal ion enhanced the film's ability in detecting X-ray radiation. Silver-loaded films were proven to significantly enhance the linearity and sensitivity of the PVA/NBT dosimeter film. These findings showed that silver-loaded PVA/NBT films exhibit a promising prospect for further applications in the domain of medical radiation.

Breaking the barriers of electron-driven x-ray radiation in crystals

Parametric x-ray radiation (PXR) is a prospective mechanism for producing directional, tunable, and quasi-coherent x-rays in laboratory-scale dimensions, yet it is limited by heat dissipation and self-absorption. Resolving these limits, we show the PXR source flux is suitable for medical imaging and x-ray spectroscopy. We discuss the experimental feasibility of these findings for a compact commercial PXR source.

Effectiveness of an X-ray shielding sheet with lower shielding ability to enable both bone mineral density determination and morphological diagnosis

Bone Mineral Density (BMD) can be determined by applying the Digital Image Processing (DIP) method using medical X-ray diagnosis. Although only the second metacarpal bone is analyzed in this approach, other parts of the body are exposed to X-ray radiation. We here propose a novel procedure in which parts of the hand surrounding the area of interest are shielded with an X-ray shielding sheet having low shielding performance. In our procedure, the main diagnostic area is not shielded, and other areas are covered with an X-ray shielding sheet with a low shielding performance. The sheet was fabricated by embedding Bi2O3 particles in resin sheet. We assessed the clinical performance of this method using three types of hand phantoms and conventional diagnostic X-ray equipment. The dose reduction for the entire hand region was evaluated by the Dose Area Product (DAPHand), which was measured with a small dosimeter, and the hand area was determined from the X-ray image. The X-ray image of the second metacarpal bone is affected by the contribution of X-rays that penetrate the object of interest and are scattered in other areas of the hand. Because our X-ray shielding sheet suppressed the generation of scattered X-rays, the pixel value of the second metacarpal bone and corresponding BMD value are varied. To address this issue, we developed a correction algorithm. We found that the X-ray shields with Dose Reduction Factor (DRF) values of 40–60% were appropriate for our methodology. Our method was estimated to have a percentage uncertainty of approximately 5% for the derivation of BMD values. Morphological information of the hand and bones could thus be clearly observed. We verified that both morphological diagnosis and quantitative determination of BMD are possible when DIP procedure is conducted using our shield.

Dendrochronology with a medical X-ray photon counting computed tomography scanner

Dendrochronology traditionally involves invasive techniques that might harm cultural heritage artifacts. Non-invasive approaches using X-ray technology and computed tomography (CT) have emerged, but dedicated CT systems are limited in availability. This study explored the viability of using a recently developed photon counting medical CT scanner for growth ring analysis. Six wood samples from pine, oak and beech with varying growth rates were studied. Comparisons were made between traditional methods and CT scanning with measurements performed on the untreated surface, a prepared surface and on CT images. For samples with annual rings wider than 0.3–0.4 mm, CT scanning performed well, while samples with thinner rings or indistinct ring boundaries presented challenges. The combination of traditional tree-ring measurements and analyses of CT scanned images might be a superior study approach generating extra data and information.

SPEI-FL: Serverless Privacy Edge Intelligence-Enabled Federated Learning in Smart Healthcare Systems

Smart healthcare systems promise significant benefits for fast and accurate medical decisions. However, working with personal health data presents new privacy issues and constraints that must be solved from a cybersecurity perspective. Edge intelligence-enabled federated learning is a new scheme that utilises decentralised computing that allows data analytics to be carried out at the edge of a network, enhancing data privacy. However, this scheme suffers from privacy attacks, including inference, free-riding, and man-in-the-middle attacks, especially with serverless computing for allocating resources to user needs. Edge intelligence-enabled federated learning requires client data insertion and deletion to authenticate genuine clients and a serverless computing capability to ensure the security of collaborative machine learning models. This work introduces a serverless privacy edge intelligence-based federated learning (SPEI-FL) framework to address these issues. SPEI-FL includes a federated edge aggregator and authentication method to improve the data privacy of federated learning and allow client adaptation and removal without impacting the overall learning processes. It also can classify intruders through serverless computing processes. The proposed framework was evaluated with the unstructured COVID-19 medical chest x-rays and MNIST digit datasets, and the structured BoT-IoT dataset. The performance of the framework is comparable with existing authentication methods and reported a higher accuracy than comparable methods (approximately 90% as compared with the 81% reported by peer methods). The proposed authentication method prevents the exposure of sensitive patient information during medical device authentication and would become the cornerstone of the next generation of medical security with serverless computing.

Influence of water saturation and water memory on CO2 hydrate formation/dissociation in porous media under flowing condition

Injection of high-pressure CO2 into depleted gas reservoirs can lead to low temperatures promoting formation of hydrate in the near wellbore area resulting in reduced injection rates. The design of effective mitigation methods requires an understanding of the impact of crucial parameters on the formation and dissociation of CO2 hydrate within the porous medium under flowing conditions. This study investigates the influence of water saturation (ranging from 20 % to 40 %) on the saturation and kinetics of CO2 hydrate during continuous CO2 injection. The experiments were conducted under a medical X-ray computed tomography (CT) to monitor the dynamics of hydrate growth inside the core and to calculate the hydrate saturation profile. The experimental data reveal increase in CO2 hydrate saturation with increasing water saturation levels. The extent of permeability reduction is strongly dependent on the initial water saturation: beyond a certain water saturation the core is fully blocked. For water saturations representative of the depleted gas fields, although the amount of generated hydrate is not sufficient to fully block the CO2 flow path, a significant reduction in permeability (approximately 80 %) is measured. It is also observed that the volume of water + hydrate phases increases during hydrate formation, indicating a lower-than-water density for CO2 hydrate. Having a history of hydrate at the same water saturation leads to an increase in CO2 consumption compared to the primary formation of hydrate, confirming the existence of the water memory effect in porous media.

Young’s double-slit interference with single hard X-ray photons

Double-slit interference experiments using monochromatic hard X-rays with the energy of 25 keV are presented. The experiments were performed at a synchrotron source with a distance of 110 m between the interferometer and the detector to produce an interference pattern with a sufficiently broad period that could be adequately sampled by a photon-counting detector with 75 micrometre pixels. In the single-particle version of the experiment, over one million image frames with a single registered photon in each one were collected. The sum of these frames showed a clear presence of the interference pattern with the expected period. Subsequent analysis provided an objective estimation of the minimal number of detected photons required to determine, in accordance with the Rose criterion, the presence of the photon interference. Apart from a general theoretical interest, these investigations were aimed at exploring the possibility of medical X-ray phase-contrast imaging in photon-counting regime at minimal radiation doses.

Enhancing medical image classification with generative AI using latent denoising diffusion probabilistic model and wiener filtering approach

In the evolving landscape of medical diagnostics, a paradigm shift is being catalysed by the advent of generative artificial intelligence. Medical X-ray, CT and MRI images are essential diagnostic tools used by healthcare professionals to assess various musculoskeletal conditions. However, obtaining a sufficient number of medical images for training deep learning models can be challenging due to limited access to labelled data. Hence, the authors propose a novel Latent diffusion process for synthesizing medical images that closely resemble real patient images, aiming to address the challenge of limited access to labelled data in medical diagnostics. Leveraging deep learning and generative modelling techniques, this method synthesizes high-fidelity images that closely mimic real patient scans. By introducing noise and subsequently training the model to denoise, the approach captures intricate patterns inherent in authentic medical images. Among the four datasets utilized, the Diabetes Retinopathy dataset demonstrates superior performance, achieving the highest Mean Structural Similarity Index (MSSIM) score of 0.57 (compared to the dataset baseline of 0.62) and an accuracy of 93.75% when passed through the proposed pipeline. The Cataract dataset, registered a MSSIM score of 0.51 (versus the dataset baseline of 0.53) and an accuracy score of 97.52%, while the Knee OA dataset follows closely with MSSIM and accuracy scores of 0.65 (in contrast to the dataset baseline of 0.63) and 68.66% respectively. The results obtained are then compared with the results generated by the other state of the art models.

From explanation to intervention: Interactive knowledge extraction from Convolutional Neural Networks used in radiology.

Deep Learning models such as Convolutional Neural Networks (CNNs) are very effective at extracting complex image features from medical X-rays. However, the limited interpretability of CNNs has hampered their deployment in medical settings as they failed to gain trust among clinicians. In this work, we propose an interactive framework to allow clinicians to ask what-if questions and intervene in the decisions of a CNN, with the aim of increasing trust in the system. The framework translates a layer of a trained CNN into a measurable and compact set of symbolic rules. Expert interactions with visualizations of the rules promote the use of clinically-relevant CNN kernels and attach meaning to the rules. The definition and relevance of the kernels are supported by radiomics analyses and permutation evaluations, respectively. CNN kernels that do not have a clinically-meaningful interpretation are removed without affecting model performance. By allowing clinicians to evaluate the impact of adding or removing kernels from the rule set, our approach produces an interpretable refinement of the data-driven CNN in alignment with medical best practice.