Journal Of Digital Imaging[JOURNAL]
Sun
200 papers
ChimeraNet: U-Net for Hair Detection in Dermoscopic Skin Lesion Images.
Lama N, Kasmi R, Hagerty JR … +6 more , Stanley RJ, Young R, Miinch J, Nepal J, Nambisan A, Stoecker WV
J Digit Imaging · 2023 Apr · PMID 36385676 · Full text
Hair and ruler mark structures in dermoscopic images are an obstacle preventing accurate image segmentation and detection of critical network features. Recognition and removal of hairs from images can be challenging, esp...
Hair and ruler mark structures in dermoscopic images are an obstacle preventing accurate image segmentation and detection of critical network features. Recognition and removal of hairs from images can be challenging, especially for hairs that are thin, overlapping, faded, or of similar color as skin or overlaid on a textured lesion. This paper proposes a novel deep learning (DL) technique to detect hair and ruler marks in skin lesion images. Our proposed ChimeraNet is an encoder-decoder architecture that employs pretrained EfficientNet in the encoder and squeeze-and-excitation residual (SERes) structures in the decoder. We applied this approach at multiple image sizes and evaluated it using the publicly available HAM10000 (ISIC2018 Task 3) skin lesion dataset. Our test results show that the largest image size (448 × 448) gave the highest accuracy of 98.23 and Jaccard index of 0.65 on the HAM10000 (ISIC 2018 Task 3) skin lesion dataset, exhibiting better performance than for two well-known deep learning approaches, U-Net and ResUNet-a. We found the Dice loss function to give the best results for all measures. Further evaluated on 25 additional test images, the technique yields state-of-the-art accuracy compared to 8 previously reported classical techniques. We conclude that the proposed ChimeraNet architecture may enable improved detection of fine image structures. Further application of DL techniques to detect dermoscopy structures is warranted.
An Intelligent System to Enhance the Performance of Brain Tumor Diagnosis from MR Images.
Shiney TSS, Jerome SA
J Digit Imaging · 2023 Apr · PMID 36385675 · Full text
In the human body, cancer is caused by aberrant cell proliferation. Brain tumors are created when cells in the human brain proliferate out of control. Brain tumors consist of two types: benign and malignant. The aberrant...
In the human body, cancer is caused by aberrant cell proliferation. Brain tumors are created when cells in the human brain proliferate out of control. Brain tumors consist of two types: benign and malignant. The aberrant parts of benign tumors, which contain dormant tumor cells, can be cured with the appropriate medication. On the other hand, malignant tumors are tumors that contain abnormal cells and an unorganized area of these abnormal cells that cannot be treated with medication. Therefore, surgery is required to remove these brain tumors. Brain cancers are manually identified and diagnosed by a skilled radiologist using traditional procedures. It's a lengthy and error-prone procedure. As a result, it is unsuitable for emerging countries with large populations. So computer-assisted automatic identification and diagnosis of brain tumors are recommended. This work proposes and implements a CAD system for the diagnosis of brain cancers using magnetic resonance imaging (MRI). Preprocessing, segmentation, feature extraction, and classification are the stages of automatic brain MRI processing that necessitate software based on a sophisticated algorithm. Image normalization with contourlet transform (INCT) is used in the preprocessing step to remove undesirable or noisy data. The performance metrics such as PSNR, MSE, and RMSE are computed. Then, the modified hierarchical k-means with firefly clustering (MHKFC) technique is used in the segmentation step to precisely recover the afflicted (tumor) area from the preprocessed image. The enhanced monarch butterfly optimization (EMBO) is used to select and then extract the most important gray-level co-occurrence matrix feature from the segmented image. The classification task was finally completed using the adaptive neuro-fuzzy inference system (ANFIS). The overall classification accuracy is 95.4% ( BRATS 2015), 96.6% ( BRATS 2021), and 93.7% (clinical data) is obtained.
Myocardial Perfusion SPECT Imaging Radiomic Features and Machine Learning Algorithms for Cardiac Contractile Pattern Recognition.
Sabouri M, Hajianfar G, Hosseini Z … +9 more , Amini M, Mohebi M, Ghaedian T, Madadi S, Rastgou F, Oveisi M, Bitarafan Rajabi A, Shiri I, Zaidi H
J Digit Imaging · 2023 Apr · PMID 36376780 · Full text
A U-shaped contraction pattern was shown to be associated with a better Cardiac resynchronization therapy (CRT) response. The main goal of this study is to automatically recognize left ventricular contractile patterns us...
A U-shaped contraction pattern was shown to be associated with a better Cardiac resynchronization therapy (CRT) response. The main goal of this study is to automatically recognize left ventricular contractile patterns using machine learning algorithms trained on conventional quantitative features (ConQuaFea) and radiomic features extracted from Gated single-photon emission computed tomography myocardial perfusion imaging (GSPECT MPI). Among 98 patients with standard resting GSPECT MPI included in this study, 29 received CRT therapy and 69 did not (also had CRT inclusion criteria but did not receive treatment yet at the time of data collection, or refused treatment). A total of 69 non-CRT patients were employed for training, and the 29 were employed for testing. The models were built utilizing features from three distinct feature sets (ConQuaFea, radiomics, and ConQuaFea + radiomics (combined)), which were chosen using Recursive feature elimination (RFE) feature selection (FS), and then trained using seven different machine learning (ML) classifiers. In addition, CRT outcome prediction was assessed by different treatment inclusion criteria as the study's final phase. The MLP classifier had the highest performance among ConQuaFea models (AUC, SEN, SPE = 0.80, 0.85, 0.76). RF achieved the best performance in terms of AUC, SEN, and SPE with values of 0.65, 0.62, and 0.68, respectively, among radiomic models. GB and RF approaches achieved the best AUC, SEN, and SPE values of 0.78, 0.92, and 0.63 and 0.74, 0.93, and 0.56, respectively, among the combined models. A promising outcome was obtained when using radiomic and ConQuaFea from GSPECT MPI to detect left ventricular contractile patterns by machine learning.
Radiologists Help Each Other to Improve Efficiency by Decreasing the Time to Look Up a Patient's Relevant Clinical History in the Electronic Medical Record.
Shafique U, Hillis SL, Park JM … +2 more , Kuehn DM, Rajput M
J Digit Imaging · 2023 Apr · PMID 36357753 · Full text
The study aims to prove that it takes less time to look up relevant clinical history from an electronic medical record (EMR) if the information is already provided in a specific space in the EMR by a fellow radiologist....
The study aims to prove that it takes less time to look up relevant clinical history from an electronic medical record (EMR) if the information is already provided in a specific space in the EMR by a fellow radiologist. Patients with complex oncological and surgical histories need frequent imaging, and every time a radiologist may spend a significant amount of time looking up the same clinical information as their peers. In collaboration with ACMIO and Radiant Epic team, a space labeled "Specialty Comments" was added to the SNAPSHOT of patient's chart in EMR. For our research purpose, the specialty comment was labeled as boxed history as a variable for data analysis. If the history was not provided in that particular space, it was labeled as without boxed history. Inclusion criteria included outpatients with complex oncological histories undergoing CT chest, abdomen, and pelvis with IV contrast. The time to look up history (LUT) was documented in minutes and seconds. Two assistant professors from Abdominal Imaging provided LUT. A total of 85 cases were included in the study, 39 with boxed history and 46 without boxed history. Comparing averages of the individual reader means for history, mean LUT differed by 2.03 min (without boxed history) versus 0.57 min (with boxed history), p < 0.0001. The t-test and the nonparametric Wilcoxon tests for a difference in the population means were highly significant (p < 0.0001). A history directed to radiologist's needs resulted in a statistically significant decrease in time spent by interpreting radiologists to look through the electronic medical records for patients with complex oncological histories. Availability of history pertinent to radiology has wide-ranging advantages, including quality reporting, decrease in turnaround time, reduction in interpretation errors, and radiologists' continued learning. The space for documenting clinical history may be reproduced, or some similar area may be developed by optimizing the electronic medical records.
Automated Registration and Color Labeling of Serial 3D Double Inversion Recovery MR Imaging for Detection of Lesion Progression in Multiple Sclerosis.
Park CC, Brummer ME, Sadigh G … +4 more , Saindane AM, Mullins ME, Allen JW, Hu R
J Digit Imaging · 2023 Apr · PMID 36352165 · Full text
Automated co-registration and subtraction techniques have been shown to be useful in the assessment of longitudinal changes in multiple sclerosis (MS) lesion burden, but the majority depend on T2-fluid-attenuated inversi...
Automated co-registration and subtraction techniques have been shown to be useful in the assessment of longitudinal changes in multiple sclerosis (MS) lesion burden, but the majority depend on T2-fluid-attenuated inversion recovery sequences. We aimed to investigate the use of a novel automated temporal color complement imaging (CCI) map overlapped on 3D double inversion recovery (DIR), and to assess its diagnostic performance for detecting disease progression in patients with multiple sclerosis (MS) as compared to standard review of serial 3D DIR images. We developed a fully automated system that co-registers and compares baseline to follow-up 3D DIR images and outputs a pseudo-color RGB map in which red pixels indicate increased intensity values in the follow-up image (i.e., progression; new/enlarging lesion), blue-green pixels represent decreased intensity values (i.e., disappearing/shrinking lesion), and gray-scale pixels reflect unchanged intensity values. Three neuroradiologists blinded to clinical information independently reviewed each patient using standard DIR images alone and using CCI maps based on DIR images at two separate exams. Seventy-six follow-up examinations from 60 consecutive MS patients who underwent standard 3 T MR brain MS protocol that included 3D DIR were included. Median cohort age was 38.5 years, with 46 women, 59 relapsing-remitting type MS, and median follow-up interval of 250 days (interquartile range: 196-394 days). Lesion progression was detected in 67.1% of cases using CCI review versus 22.4% using standard review, with a total of 182 new or enlarged lesions using CCI review versus 28 using standard review. There was a statistically significant difference between the two methods in the rate of all progressive lesions (P < 0.001, McNemar's test) as well as cortical progressive lesions (P < 0.001). Automated CCI maps using co-registered serial 3D DIR, compared to standard review of 3D DIR alone, increased detection rate of MS lesion progression in patients undergoing clinical brain MRI exam.
Mapping the Landscape of Care Providers' Quality Assurance Approaches for AI in Diagnostic Imaging.
Lundström C, Lindvall M
J Digit Imaging · 2023 Apr · PMID 36352164 · Full text
The discussion on artificial intelligence (AI) solutions in diagnostic imaging has matured in recent years. The potential value of AI adoption is well established, as are the potential risks associated. Much focus has, r...
The discussion on artificial intelligence (AI) solutions in diagnostic imaging has matured in recent years. The potential value of AI adoption is well established, as are the potential risks associated. Much focus has, rightfully, been on regulatory certification of AI products, with the strong incentive of being an enabling step for the commercial actors. It is, however, becoming evident that regulatory approval is not enough to ensure safe and effective AI usage in the local setting. In other words, care providers need to develop and implement quality assurance (QA) approaches for AI solutions in diagnostic imaging. The domain of AI-specific QA is still in an early development phase. We contribute to this development by describing the current landscape of QA-for-AI approaches in medical imaging, with focus on radiology and pathology. We map the potential quality threats and review the existing QA approaches in relation to those threats. We propose a practical categorization of QA approaches, based on key characteristics corresponding to means, situation, and purpose. The review highlights the heterogeneity of methods and practices relevant for this domain and points to targets for future research efforts.
Non-Expert Markings of Active Chronic Graft-Versus-Host Disease Photographs: Optimal Metrics of Training Effects.
Parks K, Liu X, Reasat T … +6 more , Khera Z, Baker LX, Chen H, Dawant BM, Saknite I, Tkaczyk ER
J Digit Imaging · 2023 Feb · PMID 36344635 · Full text
Lack of reliable measures of cutaneous chronic graft-versus-host disease (cGVHD) remains a significant challenge. Non-expert assistance in marking photographs of active disease could aid the development of automated segm...
Lack of reliable measures of cutaneous chronic graft-versus-host disease (cGVHD) remains a significant challenge. Non-expert assistance in marking photographs of active disease could aid the development of automated segmentation algorithms, but validated metrics to evaluate training effects are lacking. We studied absolute and relative error of marked body surface area (BSA), redness, and the Dice index as potential metrics of non-expert improvement. Three non-experts underwent an extensive training program led by a board-certified dermatologist to mark cGVHD in photographs. At the end of the 4-month training, the dermatologist confirmed that each trainee had learned to accurately mark cGVHD. The trainees' inter- and intra-rater intraclass correlation coefficient estimates were "substantial" to "almost perfect" for both BSA and total redness. For fifteen 3D photos of patients with cGVHD, the trainees' median absolute (relative) BSA error compared to expert marking dropped from 20 cm (29%) pre-training to 14 cm (24%) post-training. Total redness error decreased from 122 a*·cm (26%) to 95 a*·cm (21%). By contrast, median Dice index did not reflect improvement (0.76 to 0.75). Both absolute and relative BSA and redness errors similarly and stably reflected improvements from this training program, which the Dice index failed to capture.
Reducing Wait Times for Radiology Exams Around Holiday Periods: A Monte Carlo Simulation.
Pisharody VA, Yarmohammadi H, Ziv E … +4 more , Sotirchos VS, Alexander E, Sofocleous C, Erinjeri JP
J Digit Imaging · 2023 Feb · PMID 36344634 · Full text
Reducing patient wait times is a key operational goal and impacts patient outcomes. The purpose of this study is to explore the effects of different radiology scheduling strategies on exam wait times before and after hol...
Reducing patient wait times is a key operational goal and impacts patient outcomes. The purpose of this study is to explore the effects of different radiology scheduling strategies on exam wait times before and after holiday periods at an outpatient imaging facility using computer simulation. An idealized Monte Carlo simulation of exam scheduling at an outpatient imaging facility was developed based on the actual distribution of scheduled exams at outpatient radiology sites at a tertiary care medical center. Using this simulation, we examined three scheduling strategies: (1) no scheduling modifications, (2) increase imaging capacity before or after the holiday (i.e. increase facility hours), and (3) use a novel rolling release scheduling paradigm. In the third scenario, a fraction of exam slots are blocked to long-term follow-up exams and made available only closer to the exam date, thereby preventing long-term follow-up exams from filling the schedule and ensuring slots are available for non-follow-up exams. We examined the effect of these three scenarios on utilization and wait times, which we defined as the time from order placement to exam completion, during and after the holiday period. The baseline mean wait time for non-follow-up exams was 5.4 days in our simulation. When no scheduling modifications were made, there was a significant increase in wait times in the week preceding the holiday when compared to baseline (10.0 days vs 5.4 days, p < 0.01). Wait times remained elevated for 4 weeks following the holiday. Increasing imaging capacity during the holiday and post-holiday period by 20% reduced wait times by only 6.2% (9.38 days vs 10.0 days, p < 0.01). Increasing capacity by 50% resulted in a 7.1% reduction in wait times (9.28 days, p < 0.01), and increasing capacity by 100% resulted in a 13% reduction in wait times (8.75 days, p < 0.01). In comparison, using a rolling release model produced a reduction in peak wait times equivalent to doubling capacity (8.76 days, p < 0.01) when 45% of slots were reserved. Improvements in wait times persisted even when rolling release was limited to the 3 weeks preceding or 1 week following the holiday period. Releasing slots on a rolling basis did not significantly decrease utilization or increase wait times for long-term follow-up exams except in extreme scenarios where 80% or more of slots were reserved for non-follow-up exams. A rolling release scheduling paradigm can significantly reduce wait time fluctuations around holiday periods without requiring additional capacity or impacting utilization.
An Explainable Convolutional Neural Network for the Early Diagnosis of Alzheimer's Disease from 18F-FDG PET.
De Santi LA, Pasini E, Santarelli MF … +2 more , Genovesi D, Positano V
J Digit Imaging · 2023 Feb · PMID 36344633 · Full text
Convolutional Neural Networks (CNN) which support the diagnosis of Alzheimer's Disease using 18F-FDG PET images are obtaining promising results; however, one of the main challenges in this domain is the fact that these m...
Convolutional Neural Networks (CNN) which support the diagnosis of Alzheimer's Disease using 18F-FDG PET images are obtaining promising results; however, one of the main challenges in this domain is the fact that these models work as black-box systems. We developed a CNN that performs a multiclass classification task of volumetric 18F-FDG PET images, and we experimented two different post hoc explanation techniques developed in the field of Explainable Artificial Intelligence: Saliency Map (SM) and Layerwise Relevance Propagation (LRP). Finally, we quantitatively analyze the explanations returned and inspect their relationship with the PET signal. We collected 2552 scans from the Alzheimer's Disease Neuroimaging Initiative labeled as Cognitively Normal (CN), Mild Cognitive Impairment (MCI), and Alzheimer's Disease (AD) and we developed and tested a 3D CNN that classifies the 3D PET scans into its final clinical diagnosis. The model developed achieves, to the best of our knowledge, performances comparable with the relevant literature on the test set, with an average Area Under the Curve (AUC) for prediction of CN, MCI, and AD 0.81, 0.63, and 0.77 respectively. We registered the heatmaps with the Talairach Atlas to perform a regional quantitative analysis of the relationship between heatmaps and PET signals. With the quantitative analysis of the post hoc explanation techniques, we observed that LRP maps were more effective in mapping the importance metrics in the anatomic atlas. No clear relationship was found between the heatmap and the PET signal.
Natural Language Processing Model for Identifying Critical Findings-A Multi-Institutional Study.
Banerjee I, Davis MA, Vey BL … +6 more , Mazaheri S, Khan F, Zavaletta V, Gerard R, Gichoya JW, Patel B
J Digit Imaging · 2023 Feb · PMID 36344632 · Full text
Improving detection and follow-up of recommendations made in radiology reports is a critical unmet need. The long and unstructured nature of radiology reports limits the ability of clinicians to assimilate the full repor...
Improving detection and follow-up of recommendations made in radiology reports is a critical unmet need. The long and unstructured nature of radiology reports limits the ability of clinicians to assimilate the full report and identify all the pertinent information for prioritizing the critical cases. We developed an automated NLP pipeline using a transformer-based ClinicalBERT model which was fine-tuned on 3 M radiology reports and compared against the traditional BERT model. We validated the models on both internal hold-out ED cases from EUH as well as external cases from Mayo Clinic. We also evaluated the model by combining different sections of the radiology reports. On the internal test set of 3819 reports, the ClinicalBERT model achieved 0.96 f1-score while the BERT also achieved the same performance using the reason for exam and impression sections. However, ClinicalBERT outperformed BERT on the external test dataset of 2039 reports and achieved the highest performance for classifying critical finding reports (0.81 precision and 0.54 recall). The ClinicalBERT model has been successfully applied to large-scale radiology reports from 5 different sites. Automated NLP system that can analyze free-text radiology reports, along with the reason for the exam, to identify critical radiology findings and recommendations could enable automated alert notifications to clinicians about the need for clinical follow-up. The clinical significance of our proposed model is that it could be used as an additional layer of safeguard to clinical practice and reduce the chance of important findings reported in a radiology report is not overlooked by clinicians as well as provide a way to retrospectively track large hospital databases for evaluating the documentation of the critical findings.
Accelerated Diffusion-Weighted MR Image Reconstruction Using Deep Neural Networks.
Aamir F, Aslam I, Arshad M … +1 more , Omer H
J Digit Imaging · 2023 Feb · PMID 36333593 · Full text
Under-sampling in diffusion-weighted imaging (DWI) decreases the scan time that helps to reduce off-resonance effects, geometric distortions, and susceptibility artifacts; however, it leads to under-sampling artifacts. I...
Under-sampling in diffusion-weighted imaging (DWI) decreases the scan time that helps to reduce off-resonance effects, geometric distortions, and susceptibility artifacts; however, it leads to under-sampling artifacts. In this paper, diffusion-weighted MR image (DWI-MR) reconstruction using deep learning (DWI U-Net) is proposed to recover artifact-free DW images from variable density highly under-sampled k-space data. Additionally, different optimizers, i.e., RMSProp, Adam, Adagrad, and Adadelta, have been investigated to choose the best optimizers for DWI U-Net. The reconstruction results are compared with the conventional Compressed Sensing (CS) reconstruction. The quality of the recovered images is assessed using mean artifact power (AP), mean root mean square error (RMSE), mean structural similarity index measure (SSIM), and mean apparent diffusion coefficient (ADC). The proposed method provides up to 61.1%, 60.0%, 30.4%, and 28.7% improvements in the mean AP value of the reconstructed images in our experiments with different optimizers, i.e., RMSProp, Adam, Adagrad, and Adadelta, respectively, as compared to the conventional CS at an acceleration factor of 6 (i.e., AF = 6). The results of DWI U-Net with the RMSProp, Adam, Adagrad, and Adadelta optimizers show 13.6%, 10.0%, 8.7%, and 8.74% improvements, respectively, in terms of mean SSIM with respect to the conventional CS at AF = 6. Also, the proposed technique shows 51.4%, 29.5%, 24.04%, and 18.0% improvements in terms of mean RMSE using the RMSProp, Adam, Adagrad, and Adadelta optimizers, respectively, with reference to the conventional CS at AF = 6. The results confirm that DWI U-Net performs better than the conventional CS reconstruction. Also, when comparing the different optimizers in DWI U-Net, RMSProp provides better results than the other optimizers.
Improved Fine-Tuning of In-Domain Transformer Model for Inferring COVID-19 Presence in Multi-Institutional Radiology Reports.
Chambon P, Cook TS, Langlotz CP
J Digit Imaging · 2023 Feb · PMID 36323915 · Full text
Building a document-level classifier for COVID-19 on radiology reports could help assist providers in their daily clinical routine, as well as create large numbers of labels for computer vision models. We have developed...
Building a document-level classifier for COVID-19 on radiology reports could help assist providers in their daily clinical routine, as well as create large numbers of labels for computer vision models. We have developed such a classifier by fine-tuning a BERT-like model initialized from RadBERT, its continuous pre-training on radiology reports that can be used on all radiology-related tasks. RadBERT outperforms all biomedical pre-trainings on this COVID-19 task (P<0.01) and helps our fine-tuned model achieve an 88.9 macro-averaged F1-score, when evaluated on both X-ray and CT reports. To build this model, we rely on a multi-institutional dataset re-sampled and enriched with concurrent lung diseases, helping the model to resist to distribution shifts. In addition, we explore a variety of fine-tuning and hyperparameter optimization techniques that accelerate fine-tuning convergence, stabilize performance, and improve accuracy, especially when data or computational resources are limited. Finally, we provide a set of visualization tools and explainability methods to better understand the performance of the model, and support its practical use in the clinical setting. Our approach offers a ready-to-use COVID-19 classifier and can be applied similarly to other radiology report classification tasks.
Deep Learning for Image Enhancement and Correction in Magnetic Resonance Imaging-State-of-the-Art and Challenges.
Chen Z, Pawar K, Ekanayake M … +3 more , Pain C, Zhong S, Egan GF
J Digit Imaging · 2023 Feb · PMID 36323914 · Full text
Magnetic resonance imaging (MRI) provides excellent soft-tissue contrast for clinical diagnoses and research which underpin many recent breakthroughs in medicine and biology. The post-processing of reconstructed MR image...
Magnetic resonance imaging (MRI) provides excellent soft-tissue contrast for clinical diagnoses and research which underpin many recent breakthroughs in medicine and biology. The post-processing of reconstructed MR images is often automated for incorporation into MRI scanners by the manufacturers and increasingly plays a critical role in the final image quality for clinical reporting and interpretation. For image enhancement and correction, the post-processing steps include noise reduction, image artefact correction, and image resolution improvements. With the recent success of deep learning in many research fields, there is great potential to apply deep learning for MR image enhancement, and recent publications have demonstrated promising results. Motivated by the rapidly growing literature in this area, in this review paper, we provide a comprehensive overview of deep learning-based methods for post-processing MR images to enhance image quality and correct image artefacts. We aim to provide researchers in MRI or other research fields, including computer vision and image processing, a literature survey of deep learning approaches for MR image enhancement. We discuss the current limitations of the application of artificial intelligence in MRI and highlight possible directions for future developments. In the era of deep learning, we highlight the importance of a critical appraisal of the explanatory information provided and the generalizability of deep learning algorithms in medical imaging.
Imaging Informatics Fellowship Curriculum: Building Consensus on the Most Critical Topics and the Future of the Informatics Fellowship.
Gerard R, Makeeva V, Vey B … +7 more , Cook TS, Nagy P, Filice RW, Wang KC, Balthazar P, Harri P, Safdar NM
J Digit Imaging · 2023 Feb · PMID 36316619 · Full text
The existing fellowship imaging informatics curriculum, established in 2004, has not undergone formal revision since its inception and inaccurately reflects present-day radiology infrastructure. It insufficiently equips...
The existing fellowship imaging informatics curriculum, established in 2004, has not undergone formal revision since its inception and inaccurately reflects present-day radiology infrastructure. It insufficiently equips trainees for today's informatics challenges as current practices require an understanding of advanced informatics processes and more complex system integration. We sought to address this issue by surveying imaging informatics fellowship program directors across the country to determine the components and cutline for essential topics in a standardized imaging informatics curriculum, the consensus on essential versus supplementary knowledge, and the factors individual programs may use to determine if a newly developed topic is an essential topic. We further identified typical program structural elements and sought fellowship director consensus on offering official graduate trainee certification to imaging informatics fellows. Here, we aim to provide an imaging informatics fellowship director consensus on topics considered essential while still providing a framework for informatics fellowship programs to customize their individual curricula.
Fidelity of 3D Printed Brains from MRI Scan in Children with Pathology (Prior Hypoxic Ischemic Injury).
Chacko A, Rungsiprakarn P, Erlic I … +2 more , Thai NJ, Andronikou S
J Digit Imaging · 2023 Feb · PMID 36280655 · Full text
Cortical injury on the surface of the brain in children with hypoxic ischemic injury (HII) can be difficult to demonstrate to non-radiologists and lay people using brain images alone. Three-dimensional (3D) printing is h...
Cortical injury on the surface of the brain in children with hypoxic ischemic injury (HII) can be difficult to demonstrate to non-radiologists and lay people using brain images alone. Three-dimensional (3D) printing is helpful to communicate the volume loss and pathology due to HII in children's brains. 3D printed models represent the brain to scale and can be held up against models of normal brains for appreciation of volume loss. If 3D printed brains are to be used for formal communication, e.g., with medical colleagues or in court, they should have high fidelity of reproduction of the actual size of patients' brains. Here, we evaluate the size fidelity of 3D printed models from MRI scans of the brain, in children with prior HII. Twelve 3D prints of the brain were created from MRI scans of children with HII and selected to represent a variety of cortical pathologies. Specific predetermined measures of the 3D prints were made and compared to measures in matched planes on MRI. Fronto-occipital length (FOL) and bi-temporal/bi-parietal diameters (BTD/BPD) demonstrated high interclass correlations (ICC). Correlations were moderate to weak for hemispheric height, temporal height, and pons-cerebellar thickness. The average standard error of measurement (SEM) was 0.48 cm. Our results demonstrate high correlations in overall measurements of each 3D printed model derived from brain MRI scans versus the original MRI, evidenced by high ICC values for FOL and BTD/BPD. Measures with low correlation values can be explained by variability in matching the plane of measurement to the MRI slice orientation.
Automated Multimodal Machine Learning for Esophageal Variceal Bleeding Prediction Based on Endoscopy and Structured Data.
Wang Y, Hong Y, Wang Y … +10 more , Zhou X, Gao X, Yu C, Lin J, Liu L, Gao J, Yin M, Xu G, Liu X, Zhu J
J Digit Imaging · 2023 Feb · PMID 36279027 · Full text
Esophageal variceal (EV) bleeding is a severe medical emergency related to cirrhosis. Early identification of cirrhotic patients with at a high risk of EV bleeding is key to improving outcomes and optimizing medical reso...
Esophageal variceal (EV) bleeding is a severe medical emergency related to cirrhosis. Early identification of cirrhotic patients with at a high risk of EV bleeding is key to improving outcomes and optimizing medical resources. This study aimed to evaluate the feasibility of automated multimodal machine learning (MMML) for predicting EV bleeding by integrating endoscopic images and clinical structured data. This study mainly includes three steps: step 1, developing deep learning (DL) models using EV images by 12-month bleeding on TensorFlow (backbones include ResNet, Xception, EfficientNet, ViT and ConvMixer); step 2, training and internally validating MMML models integrating clinical structured data and DL model outputs to predict 12-month EV bleeding on an H2O-automated machine learning platform (algorithms include DL, XGBoost, GLM, GBM, RF, and stacking); and step 3, externally testing MMML models. Furthermore, existing clinical indices, e.g., the MELD score, Child‒Pugh score, APRI, and FIB-4, were also examined. Five DL models were transfer learning to the binary classification of EV endoscopic images at admission based on the occurrence or absence of bleeding events during the 12-month follow-up. An EfficientNet model achieved the highest accuracy of 0.868 in the validation set. Then, a series of MMML models, integrating clinical structured data and the output of the EfficientNet model, were automatedly trained to predict 12-month EV bleeding. A stacking model showed the highest accuracy (0.932), sensitivity (0.952), and F1-score (0.879) in the test dataset, which was also better than the existing indices. This study is the first to evaluate the feasibility of automated MMML in predicting 12-month EV bleeding based on endoscopic images and clinical variables.
Artificial Intelligence-Powered Clinical Decision Support and Simulation Platform for Radiology Trainee Education.
Shah C, Davtyan K, Nasrallah I … +2 more , Bryan RN, Mohan S
J Digit Imaging · 2023 Feb · PMID 36279026 · Full text
Technological tools can redesign traditional approaches to radiology education, for example, with simulation cases and via computer-generated feedback. In this study, we investigated the use of an AI-powered, Bayesian in...
Technological tools can redesign traditional approaches to radiology education, for example, with simulation cases and via computer-generated feedback. In this study, we investigated the use of an AI-powered, Bayesian inference-based clinical decision support (CDS) software to provide automated "real-time" feedback to trainees during interpretation of clinical and simulation brain MRI examinations. Radiology trainees participated in sessions in which they interpreted 3 brain MRIs: two cases from a routine clinical worklist (one without and one with CDS) and a teaching file-based simulation case with CDS. The CDS software required trainees to input imaging features and differential diagnoses, after which inferred diagnoses were displayed, and the case was reviewed with an attending neuroradiologist. An observer timed each case, including time spent on education, and trainees completed a survey rating their confidence in their findings and the educational value of the case. Ten trainees reviewed 75 brain MRI examinations during 25 reading sessions. Trainees had slightly lower confidence in their findings and diagnosis and rated the educational value slightly higher for simulation cases with CDS compared to clinical cases without CDS (p < 0.05). There were no significant differences in ratings of clinical cases with or without CDS. No differences in overall timing were found among the reading scenarios. Simulation cases with "CDS-provided feedback" may improve the educational value of interpreting imaging studies at a workstation without adding additional time. Further investigation will help drive innovation in trainee education, which may be particularly relevant in this era of increasing remote work and asynchronous attending review.
Toward Robust Partial-Image Based Template Matching Techniques for MRI-Guided Interventions.
Lee EJ, Farzinfard S, Yarmolenko P … +2 more , Cleary K, Monfaredi R
J Digit Imaging · 2023 Feb · PMID 36271210 · Full text
We have developed an MRI-safe needle guidance toolkit for MRI-guided interventions intended to enable accurate positioning for needle-based procedures. The toolkit allows intuitive and accurate needle angulation and entr...
We have developed an MRI-safe needle guidance toolkit for MRI-guided interventions intended to enable accurate positioning for needle-based procedures. The toolkit allows intuitive and accurate needle angulation and entry point positioning according to an MRI-based plan, using a flexible, patterned silicone 2D grid. The toolkit automatically matches the grid on MRI planning images with a physical silicon grid placed conformally on the patient's skin and provides the Interventional Radiologist an easy-to-use guide showing the needle entry point on the silicon grid as well as needle angle information. The radiologist can use this guide along with a 2-degree-of-freedom (rotation and angulation relative to the entry point) hand-held needle guide to place the needle into the anatomy of interest. The initial application that we are considering for this toolkit is arthrography, a diagnostic procedure to evaluate the joint space condition. However, this toolkit could be used for any needle-based and percutaneous procedures such as MRI-guided biopsy and facet joint injection. For matching the images, we adopt a transformation parameter estimation technique using the phase-only correlation method in the frequency domain. We investigated the robustness of this method against rotation, displacement, and Rician noise. The algorithm was able to successfully match all the dataset images. We also investigated the accuracy of identifying the entry point from registered template images as a prerequisite for a future targeting study. Application of the template matching algorithm to locate the needle entry points within the MRI dataset resulted in an average entry point location estimation accuracy of 0.12 ±0.2 mm. This promising result motivates a more detailed assessment of this algorithm in the future including a targeting study on a silicon phantom with embedded plastic targets to investigate the end-to-end accuracy of this automatic template matching algorithm in the interventional MRI room.
Event-Based Clinical Finding Extraction from Radiology Reports with Pre-trained Language Model.
Lau W, Lybarger K, Gunn ML … +1 more , Yetisgen M
J Digit Imaging · 2023 Feb · PMID 36253581 · Full text
Radiology reports contain a diverse and rich set of clinical abnormalities documented by radiologists during their interpretation of the images. Comprehensive semantic representations of radiological findings would enabl...
Radiology reports contain a diverse and rich set of clinical abnormalities documented by radiologists during their interpretation of the images. Comprehensive semantic representations of radiological findings would enable a wide range of secondary use applications to support diagnosis, triage, outcomes prediction, and clinical research. In this paper, we present a new corpus of radiology reports annotated with clinical findings. Our annotation schema captures detailed representations of pathologic findings that are observable on imaging ("lesions") and other types of clinical problems ("medical problems"). The schema used an event-based representation to capture fine-grained details, including assertion, anatomy, characteristics, size, and count. Our gold standard corpus contained a total of 500 annotated computed tomography (CT) reports. We extracted triggers and argument entities using two state-of-the-art deep learning architectures, including BERT. We then predicted the linkages between trigger and argument entities (referred to as argument roles) using a BERT-based relation extraction model. We achieved the best extraction performance using a BERT model pre-trained on 3 million radiology reports from our institution: 90.9-93.4% F1 for finding triggers and 72.0-85.6% F1 for argument roles. To assess model generalizability, we used an external validation set randomly sampled from the MIMIC Chest X-ray (MIMIC-CXR) database. The extraction performance on this validation set was 95.6% for finding triggers and 79.1-89.7% for argument roles, demonstrating that the model generalized well to the cross-institutional data with a different imaging modality. We extracted the finding events from all the radiology reports in the MIMIC-CXR database and provided the extractions to the research community.
Content-Based Medical Image Retrieval System for Skin Melanoma Diagnosis Based on Optimized Pair-Wise Comparison Approach.
Rout NK, Ahirwal MK, Atulkar M
J Digit Imaging · 2023 Feb · PMID 36253580 · Full text
Medical image analysis for perfect diagnosis of disease has become a very challenging task. Due to improper diagnosis, required medical treatment may be skipped. Proper diagnosis is needed as suspected lesions could be m...
Medical image analysis for perfect diagnosis of disease has become a very challenging task. Due to improper diagnosis, required medical treatment may be skipped. Proper diagnosis is needed as suspected lesions could be missed by the physician's eye. Hence, this problem can be settled up by better means with the investigation of similar case studies present in the healthcare database. In this context, this paper substantiates an assistive system that would help dermatologists for accurate identification of 23 different kinds of melanoma. For this, 2300 dermoscopic images were used to train the skin-melanoma similar image search system. The proposed system uses feature extraction by assigning dynamic weights to the low-level features based on the individual characteristics of the searched images. Optimal weights are obtained by the newly proposed optimized pair-wise comparison (OPWC) approach. The uniqueness of the proposed approach is that it provides the dynamic weights to the features of the searched image instead of applying static weights. The proposed approach is supported by analytic hierarchy process (AHP) and meta-heuristic optimization algorithms such as particle swarm optimization (PSO), JAYA, genetic algorithm (GA), and gray wolf optimization (GWO). The proposed approach has been tested with images of 23 classes of melanoma and achieved significant precision and recall. Thus, this approach of skin melanoma image search can be used as an expert assistive system to help dermatologists/physicians for accurate identification of different types of melanomas.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe