Quantifying effects of blood pressure control on neuroimaging utilization in a large multi-institutional healthcare population.

OBJECTIVES: Essential hypertension is a common chronic condition that can exacerbate or complicate various neurological diseases that may necessitate neuroimaging. Given growing medical imaging costs and the need to understand relationships between population blood pressure control and neuroimaging utilization, we seek to quantify the relationship between maximum blood pressure recorded in a given year and same-year utilization of neuroimaging CT or MR in a large healthcare population. METHODS: A retrospective population-based cohort study was performed by extracting aggregate data from a multi-institutional dataset of patient encounters from 2016, 2018, and 2020 using an informatics platform (Cosmos) consisting of de-duplicated data from over 140 academic and non-academic health systems, comprising over 137 million unique patients. A population-based sample of all patients with recorded blood pressures of at least 50 mmHg DBP or 90 mmHg SBP were included. Cohorts were identified based on maximum annual SBP and DBP meeting or exceeding pre-defined thresholds. For each cohort, we assessed neuroimaging CT and MR utilization, defined as the percentage of patients undergoing ≥1 neuroimaging exam of interest in the same calendar year. RESULTS: The multi-institutional population consisted of >38 million patients for the most recent calendar year analyzed, with overall utilization of 3.8-5.1% for CT and 1.5-2.0% for MR across the study period. Neuroimaging utilization increased substantially with increasing annual maximum BP. Even a modest BP increase to 140 mmHg systolic or 90 mmHg diastolic is associated with 3-4-fold increases in MR and 5-7-fold increases in CT same-year imaging compared to BP values below 120 mmHg / 80 mmHg. CONCLUSION: Higher annual maximum recorded blood pressure is associated with higher same-year neuroimaging CT and MR utilization rates. These observations are relevant to public health efforts on hypertension management to mitigate costs associated with growing imaging utilization.

Lessons Learned in Building Expertly Annotated Multi-Institution Datasets and Hosting the RSNA AI Challenges.

The Radiological Society of North America (RSNA) has held artificial intelligence competitions to tackle real-world medical imaging problems at least annually since 2017. This article examines the challenges and processes involved in organizing these competitions, with a specific emphasis on the creation and curation of high-quality datasets. The collection of diverse and representative medical imaging data involves dealing with issues of patient privacy and data security. Furthermore, ensuring quality and consistency in data, which includes expert labeling and accounting for various patient and imaging characteristics, necessitates substantial planning and resources. Overcoming these obstacles requires meticulous project management and adherence to strict timelines. The article also highlights the potential of crowdsourced annotation to progress medical imaging research. Through the RSNA competitions, an effective global engagement has been realized, resulting in innovative solutions to complex medical imaging problems, thus potentially transforming health care by enhancing diagnostic accuracy and patient outcomes. Keywords: Use of AI in Education, Artificial Intelligence © RSNA, 2024.

Performance of the Winning Algorithms of the RSNA 2022 Cervical Spine Fracture Detection Challenge.

Purpose To evaluate and report the performance of the winning algorithms of the Radiological Society of North America Cervical Spine Fracture AI Challenge. Materials and Methods The competition was open to the public on Kaggle from July 28 to October 27, 2022. A sample of 3112 CT scans with and without cervical spine fractures (CSFx) were assembled from multiple sites (12 institutions across six continents) and prepared for the competition. The test set had 1093 scans (private test set: n = 789; mean age, 53.40 years ± 22.86 [SD]; 509 males; public test set: n = 304; mean age, 52.51 years ± 20.73; 189 males) and 847 fractures. The eight top-performing artificial intelligence (AI) algorithms were retrospectively evaluated, and the area under the receiver operating characteristic curve (AUC) value, F1 score, sensitivity, and specificity were calculated. Results A total of 1108 contestants composing 883 teams worldwide participated in the competition. The top eight AI models showed high performance, with a mean AUC value of 0.96 (95% CI: 0.95, 0.96), mean F1 score of 90% (95% CI: 90%, 91%), mean sensitivity of 88% (95% Cl: 86%, 90%), and mean specificity of 94% (95% CI: 93%, 96%). The highest values reported for previous models were an AUC of 0.85, F1 score of 81%, sensitivity of 76%, and specificity of 97%. Conclusion The competition successfully facilitated the development of AI models that could detect and localize CSFx on CT scans with high performance outcomes, which appear to exceed known values of previously reported models. Further study is needed to evaluate the generalizability of these models in a clinical environment. Keywords: Cervical Spine, Fracture Detection, Machine Learning, Artificial Intelligence Algorithms, CT, Head/Neck Supplemental material is available for this article. © RSNA, 2024.

The infundibulochiasmatic angle and the favorability of an endoscopic endonasal approach in type IV craniopharyngioma: illustrative case.

BACKGROUND: Lesions located in the floor of the third ventricle are among the most difficult to access in neurosurgery. The neurovascular structures can limit transcranial exposure, whereas tumor extension into the third ventricle can limit visualization and access. The midline transsphenoidal route is an alternative approach to tumor invading the third ventricle if the tumor is localized at its anterior half and a working space between the optic apparatus and the pituitary infundibulum exists. The authors introduce the "infundibulochiasmatic angle," a valuable measurement supporting the feasibility of the translamina terminalis endoscopic endonasal approach (EEA) for resection of type IV craniopharyngiomas. OBSERVATIONS: Due to a favorable infundibulochiasmatic angle measurement on preoperative magnetic resonance imaging (MRI), an endoscopic endonasal transsellar transtubercular approach was performed to resect a type IV craniopharyngioma. At 2-month follow-up, the patient's neurological exam was unremarkable, with improvement in bitemporal hemianopsia. Postoperative MRI confirmed gross-total tumor resection. LESSONS: The infundibulochiasmatic angle is a radiological tool for evaluating the feasibility of EEA when resecting tumors in the anterior half of the third ventricle. Advantages include reduced brain retraction and excellent rates of resection, with minimal postoperative risks of cerebrospinal fluid leakage and permanent pituitary dysfunction.

Data Liberation and Crowdsourcing in Medical Research: The Intersection of Collective and Artificial Intelligence.

In spite of an exponential increase in the volume of medical data produced globally, much of these data are inaccessible to those who might best use them to develop improved health care solutions through the application of advanced analytics such as artificial intelligence. Data liberation and crowdsourcing represent two distinct but interrelated approaches to bridging existing data silos and accelerating the pace of innovation internationally. In this article, we examine these concepts in the context of medical artificial intelligence research, summarizing their potential benefits, identifying potential pitfalls, and ultimately making a case for their expanded use going forward. A practical example of a crowdsourced competition using an international medical imaging dataset is provided. Keywords: Artificial Intelligence, Data Liberation, Crowdsourcing © RSNA, 2023.

A New Finding on Magnetic Resonance Imaging for Diagnosis of Hemifacial Spasm with High Accuracy and Interobserver Correlation.

Among patients with clinical hemifacial spasm (HFS), imaging exams aim to identify the neurovascular conflict (NVC) location. It has been proven that the identification in the preoperative exam increases the rate of surgical success. Despite the description of specific magnetic resonance image (MRI) acquisitions, the site of neurovascular compression is not always visualized. The authors describe a new MRI finding that helps in the diagnosis of HFS, and evaluate the sensitivity, specificity, and interobserver correlation of the described sign. A cross-sectional study including cases of hemifacial spasm treated surgically from 1 August 2011 to 31 July 2021 was performed. The MRIs of the cases were independently evaluated by two experienced neuroradiologists, who were blinded regarding the side of the symptom. The neuroradiologists were assigned to evaluate the MRIs in two separate moments. Primarily, they evaluated whether there was a neurovascular conflict based on the standard technique. Following this initial analysis, the neuroradiologists received a file with the description of the novel sign, named Prevedello Sign (PS). In a second moment, the same neuroradiologists were asked to identify the presence of the PS and, if it was present, to report on which side. A total of 35 patients were included, mostly females (65.7%) with a mean age of 59.02 (+0.48). Since the 35 cases were independently evaluated by two neuroradiologists, a total of 70 reports were included in the analysis. The PS was present in 66 patients (sensitivity of 94.2%, specificity of 91.4% and positive predictive value of 90.9%). When both analyses were performed in parallel (standard plus PS), the sensitivity increased to 99.2%. Based on the findings of this study, the authors conclude that PS is helpful in determining the neurovascular conflict location in patients with HFS. Its presence, combined with the standard evaluation, increases the sensitivity of the MRI to over 99%, without increasing risks of harm to patients or resulting in additional costs.

Endoscopic endonasal marsupialization of rathke cleft cysts: clinical outcomes and risk factors analysis of visual impairment, pituitary dysfunction, and CSF leak.

OBJECTIVE: This paper assesses the clinical and imaging characteristics, histopathological findings, and treatment outcomes of patients with Rathke's cleft cyst (RCC), as well as identifies potential risk factors for preoperative visual and pituitary dysfunction, intraoperative cerebrospinal fluid (CSF) leak, and recurrence. Through analyzing these factors, the study aims to contribute to the current understanding of the management of RCCs and identify opportunities for improving patient outcomes. METHODS: We performed a retrospective analysis of 45 RCC patients between ages 18-80 treated by Endoscopic Endonasal Approach (EEA) and cyst marsupialization between 2010 and 2022 at a single institution. RESULTS: The median patient age was 34, and 73% were female. The mean follow-up was 70 ± 43 months. Preoperative visual impairment correlated with cyst diameter (OR = 1.41, 95% CI = 1.07 to 1.85, p-value = 0.01) and older age (OR = 1.06, 95% CI = 1.01 to 1.11, p-value = 0.02). Intraoperative CSF leaks were 11 times more likely for cysts ≥ 2 cm (OR = 11.3, 95% CI = 1.25 to 97.37, p-value = 0.03), with the odds of leakage doubling for every 0.1 cm increase in cyst size (OR = 1.41, 95% CI = 1.08 to 1.84, p-value = 0.01). Preoperative RCC appearing hypointense on T1 images demonstrated significantly higher CSF leak rates than hyperintense lesions (OR = 122.88, 95% CI = 1.5 to 10077.54, p-value = 0.03). Preoperative pituitary hypofunction was significantly more likely in patients with the presence of inflammation on histopathology (OR = 20.53, 95% CI = 2.20 to 191.45, p-value = 0.008 ) and T2 hyperintensity on magnetic resonance imaging (MRI) sequences (OR = 23.2, 95% CI = 2.56 to 211.02, p-value = 0.005). Notably, except for the hyperprolactinemia, no postoperative improvement was observed in pituitary function. CONCLUSION: Carefully considering risk factors, surgeons can appropriately counsel patients and deliver expectations for complications and long-term results. In contrast to preoperative visual impairment, preoperative pituitary dysfunction was found to have the least improvement post-surgery. It was the most significant permanent complication, with our data indicating the link to the cyst signal intensity on T2 MR and inflammation on histopathology. Earlier surgical intervention might improve the preservation of pituitary function.

The RSNA Cervical Spine Fracture CT Dataset.

This dataset is composed of cervical spine CT images with annotations related to fractures; it is available at https://www.kaggle.com/competitions/rsna-2022-cervical-spine-fracture-detection/.

Detecting and Characterizing Inferior Vena Cava Filters on Abdominal Computed Tomography with Data-Driven Computational Frameworks.

Two data-driven algorithms were developed for detecting and characterizing Inferior Vena Cava (IVC) filters on abdominal computed tomography to assist healthcare providers with the appropriate management of these devices to decrease complications: one based on 2-dimensional data and transfer learning (2D + TL) and an augmented version of the same algorithm which accounts for the 3-dimensional information leveraging recurrent convolutional neural networks (3D + RCNN). The study contains 2048 abdominal computed tomography studies obtained from 439 patients who underwent IVC filter placement during the 10-year period from January 1st, 2009, to January 1st, 2019. Among these, 399 patients had retrievable filters, and 40 had non-retrievable filter types. The reference annotations for the filter location were obtained through a custom-developed interface. The ground truth annotations for the filter types were determined based on the electronic medical record and physician review of imaging. The initial stage of the framework returns a list of locations containing metallic objects based on the density of the structure. The second stage processes the candidate locations and determines which one contains an IVC filter. The final stage of the pipeline classifies the filter types as retrievable vs. non-retrievable. The computational models are trained using Tensorflow Keras API on an Nvidia Quadro GV100 system. We utilized a fine-tuning supervised training strategy to conduct our experiments. We find that the system achieves high sensitivity on detecting the filter locations with a high confidence value. The 2D + TL model achieved a sensitivity of 0.911 and a precision of 0.804, and the 3D + RCNN model achieved a sensitivity of 0.923 and a precision of 0.853 for filter detection. The system confidence for the IVC location predictions is high: 0.993 for 2D + TL and 0.996 for 3D + RCNN. The filter type prediction component of the system achieved 0.945 sensitivity, 0.882 specificity, and 0.97 AUC score with 2D + TL and 0. 940 sensitivity, 0.927 specificity, and 0.975 AUC score with 3D + RCNN. With the intent to create tools to improve patient outcomes, this study describes the initial phase of a computational framework to support healthcare providers in detecting patients with retained IVC filters, so an individualized decision can be made to remove these devices when appropriate, to decrease complications. To our knowledge, this is the first study that curates abdominal computed tomography (CT) scans and presents an algorithm for automated detection and characterization of IVC filters.

Artificial Intelligence in Neuroradiology: A Review of Current Topics and Competition Challenges.

There is an expanding body of literature that describes the application of deep learning and other machine learning and artificial intelligence methods with potential relevance to neuroradiology practice. In this article, we performed a literature review to identify recent developments on the topics of artificial intelligence in neuroradiology, with particular emphasis on large datasets and large-scale algorithm assessments, such as those used in imaging AI competition challenges. Numerous applications relevant to ischemic stroke, intracranial hemorrhage, brain tumors, demyelinating disease, and neurodegenerative/neurocognitive disorders were discussed. The potential applications of these methods to spinal fractures, scoliosis grading, head and neck oncology, and vascular imaging were also reviewed. The AI applications examined perform a variety of tasks, including localization, segmentation, longitudinal monitoring, diagnostic classification, and prognostication. While research on this topic is ongoing, several applications have been cleared for clinical use and have the potential to augment the accuracy or efficiency of neuroradiologists.

Prediction of model generalizability for unseen data: Methodology and case study in brain metastases detection in T1-Weighted contrast-enhanced 3D MRI.

BACKGROUND AND PURPOSE: A medical AI system's generalizability describes the continuity of its performance acquired from varying geographic, historical, and methodologic settings. Previous literature on this topic has mostly focused on "how" to achieve high generalizability (e.g., via larger datasets, transfer learning, data augmentation, model regularization schemes), with limited success. Instead, we aim to understand "when" the generalizability is achieved: Our study presents a medical AI system that could estimate its generalizability status for unseen data on-the-fly. MATERIALS AND METHODS: We introduce a latent space mapping (LSM) approach utilizing Fréchet distance loss to force the underlying training data distribution into a multivariate normal distribution. During the deployment, a given test data's LSM distribution is processed to detect its deviation from the forced distribution; hence, the AI system could predict its generalizability status for any previously unseen data set. If low model generalizability is detected, then the user is informed by a warning message integrated into a sample deployment workflow. While the approach is applicable for most classification deep neural networks (DNNs), we demonstrate its application to a brain metastases (BM) detector for T1-weighted contrast-enhanced (T1c) 3D MRI. The BM detection model was trained using 175 T1c studies acquired internally (from the authors' institution) and tested using (1) 42 internally acquired exams and (2) 72 externally acquired exams from the publicly distributed Brain Mets dataset provided by the Stanford University School of Medicine. Generalizability scores, false positive (FP) rates, and sensitivities of the BM detector were computed for the test datasets. RESULTS AND CONCLUSION: The model predicted its generalizability to be low for 31% of the testing data (i.e., two of the internally and 33 of the externally acquired exams), where it produced (1) ∼13.5 false positives (FPs) at 76.1% BM detection sensitivity for the low and (2) ∼10.5 FPs at 89.2% BM detection sensitivity for the high generalizability groups respectively. These results suggest that the proposed formulation enables a model to predict its generalizability for unseen data.

Pituitary Stalk Stretch Predicts Postoperative Diabetes Insipidus After Pituitary Macroadenoma Transsphenoidal Resection.

BACKGROUND: Manipulation of the pituitary stalk, posterior pituitary gland, and hypothalamus during transsphenoidal pituitary adenoma resection can cause disruption of water electrolyte regulation leading to diabetes insipidus (DI). OBJECTIVE: To determine whether pituitary stalk stretch is an independent risk factor for postoperative DI after pituitary adenoma resection. METHODS: A retrospective review was performed of patients undergoing endoscopic endonasal resection of pituitary macroadenoma between July 2010 and December 2016 by a single neurosurgeon. We analyzed preoperative and postoperative imaging metrics to assess predictors for postoperative DI. RESULTS: Of the 234 patients undergoing resection, 41 (17.5%) developed postoperative DI. DI was permanent in 10 (4.3%) and transient in 31 (13.2%). The pituitary stalk stretch, measured as the change in stalk length from preoperative to postoperative imaging, was greater in the DI compared with the non-DI group (10.1 mm vs 5.9 mm, P < .0001). The pituitary stalk stretch was associated with DI with significant difference in mean pituitary stalk stretch between non-DI group vs DI group (5.9 mm vs 10.1 mm, P < .0001). Multivariate analysis revealed that pituitary stalk stretch >10 mm was a significant independent predictor of postoperative DI [odds ratios = 2.56 (1.10-5.96), P = .029]. When stratified into transient and permanent DI, multivariable analysis showed that pituitary stalk stretch >10 mm was a significant independent predictor of transient DI [odds ratios = 2.71 (1.0-7.1), P = .046] but not permanent DI. CONCLUSION: Postoperative pituitary stalk stretch after transsphenoidal pituitary adenoma surgery is an important factor for postoperative DI. We propose a reconstruction strategy to mitigate stalk stretch.

Intraoperative Prediction of Long-Term Remission in Acromegaly.

BACKGROUND: The features of long-term remission in acromegaly adenomectomy are incompletely understood. An intraoperative predictor for long-term outcome would be valuable for assessing resection of growth hormone (GH)-secreting tumors in real-time. OBJECTIVE: To evaluate whether intraoperative GH measurement could predict long-term outcomes for acromegaly. METHODS: In 47 patients, peripheral blood GH levels were measured thrice intraoperatively: once before tumor dissection, once during tumor dissection, and once after tumor dissection. Long-term remission was defined by age-appropriate, normalized insulin-like growth factor-1 at most recent follow-up and a random GH less than 1.0 ng/mL. Patients were only considered to be in long-term remission without the use of postoperative medical therapy for acromegaly or radiation therapy. RESULTS: The median length of follow-up was 4.51 (range: 0.78-9.80) years. Long-term remission was achieved in 61.7% (29/47) of operations. Like previous studies, cavernous sinus invasion (odds ratio [OR]: 0.060; 95% CI: 0.014-0.260; P value < .01), suprasellar extension (OR: 0.191; 95% CI: 0.053-0.681; P value<.01), and tumor size greater than 1 cm (OR: 0.177; 95% CI: 0.003-0.917; P value = .03) were associated with not being in long-term remission. The minimum GH measured intraoperatively predicted long-term outcome (area under the curve: 0.7107; 95% CI: 0.537-0.884; P value < .01). The odds ratio of remission in patients with the lowest quartile minimum intraoperative GH compared with patients with the highest quartile minimum intraoperative GH was 27.0 (95% CI: 2.343-311.171; P value < .01). CONCLUSION: Minimum intraoperative GH may predict long-term outcome for acromegaly, which in principle could provide the pituitary neurosurgeon with real-time feedback and inform intraoperative decision making.

Advancing Brain Metastases Detection in T1-Weighted Contrast-Enhanced 3D MRI Using Noisy Student-Based Training.

The detection of brain metastases (BM) in their early stages could have a positive impact on the outcome of cancer patients. The authors previously developed a framework for detecting small BM (with diameters of <15 mm) in T1-weighted contrast-enhanced 3D magnetic resonance images (T1c). This study aimed to advance the framework with a noisy-student-based self-training strategy to use a large corpus of unlabeled T1c data. Accordingly, a sensitivity-based noisy-student learning approach was formulated to provide high BM detection sensitivity with a reduced count of false positives. This paper (1) proposes student/teacher convolutional neural network architectures, (2) presents data and model noising mechanisms, and (3) introduces a novel pseudo-labeling strategy factoring in the sensitivity constraint. The evaluation was performed using 217 labeled and 1247 unlabeled exams via two-fold cross-validation. The framework utilizing only the labeled exams produced 9.23 false positives for 90% BM detection sensitivity, whereas the one using the introduced learning strategy led to ~9% reduction in false detections (i.e., 8.44). Significant reductions in false positives (>10%) were also observed in reduced labeled data scenarios (using 50% and 75% of labeled data). The results suggest that the introduced strategy could be utilized in existing medical detection applications with access to unlabeled datasets to elevate their performances.

Mortality Prediction Analysis among COVID-19 Inpatients Using Clinical Variables and Deep Learning Chest Radiography Imaging Features.

The emergence of the COVID-19 pandemic over a relatively brief interval illustrates the need for rapid data-driven approaches to facilitate clinical decision making. We examined a machine learning process to predict inpatient mortality among COVID-19 patients using clinical and chest radiographic data. Modeling was performed with a de-identified dataset of encounters prior to widespread vaccine availability. Non-imaging predictors included demographics, pre-admission clinical history, and past medical history variables. Imaging features were extracted from chest radiographs by applying a deep convolutional neural network with transfer learning. A multi-layer perceptron combining 64 deep learning features from chest radiographs with 98 patient clinical features was trained to predict mortality. The Local Interpretable Model-Agnostic Explanations (LIME) method was used to explain model predictions. Non-imaging data alone predicted mortality with an ROC-AUC of 0.87 ± 0.03 (mean ± SD), while the addition of imaging data improved prediction slightly (ROC-AUC: 0.91 ± 0.02). The application of LIME to the combined imaging and clinical model found HbA1c values to contribute the most to model prediction (17.1 ± 1.7%), while imaging contributed 8.8 ± 2.8%. Age, gender, and BMI contributed 8.7%, 8.2%, and 7.1%, respectively. Our findings demonstrate a viable explainable AI approach to quantify the contributions of imaging and clinical data to COVID mortality predictions.

Trends and Predictors of Imaging Utilization by Modality within an Academic Health System's Active Patient Population.

RATIONALE AND OBJECTIVES: Evaluate trends and demographic predictors of imaging utilization at a university-affiliated health system. MATERIALS AND METHODS: In this single-institution retrospective study, per capita estimates of imaging utilization among patients active in the health system were computed by cross-referencing all clinical encounters (2004-2016) for 1,628,980 unique patients with a listing of 6,157,303 diagnostic radiology encounters. Time trends in imaging utilization and effects of gender, race/ethnicity, and age were assessed, with subgroup analyses performed by imaging modality. Utilization was analyzed as both a continuous and binary outcome variable. RESULTS: Over 13 years, total diagnostic exams rose 6.8% a year (285,947-622,196 exams per annum), while the active population size grew 7.0% a year (244,238-543,290 active patients per annum). Per capita utilization peaked in 2007 at 1.33 studies/patient/year before dropping to 1.06 from 2011 to 2015. Latest per capita utilization was 0.22 for computed tomography, 0.10 for MR, 0.20 for US, 0.03 for NM, 0.51 for radiography, and 0.07 for mammography. Over the study period, ultrasound utilization doubled, whereas NM and radiography utilization decreased. computed tomography, MR, and mammography showed no significant net change. Univariate analysis of utilization as a continuous variable showed statistically significant effects of gender, race/ethnicity, and age (P < 0.0001), with utilization higher in males and Blacks and lower in Asian/Pacific Islanders and Hispanics. Utilization increased with age, except for a decline after age 75. Many of the effects of age, gender, and race/ethnicity were also found when analyzing the binarized utilization variable. CONCLUSIONS: Although absolute counts of imaging studies more than doubled, the net change in per capita utilization over the study period was minimal. Variations in utilization across age, gender, and race/ethnicity may reflect differential health needs and/or access disparities, warranting future studies.

Augmented networks for faster brain metastases detection in T1-weighted contrast-enhanced 3D MRI.

Early detection of brain metastases (BM) is one of the determining factors for the successful treatment of patients with cancer; however, the accurate detection of small BM lesions (< 15 mm) remains a challenging task. We previously described a framework for the detection of small BM in single-sequence gadolinium-enhanced T1-weighted 3D MRI datasets. It combined classical image processing (IP) with a dedicated convolutional neural network, taking approximately 30 s to process each dataset due to computation-intensive IP stages. To overcome the speed limitation, this study aims to reformulate the framework via an augmented pair of CNNs (eliminating the IP) to reduce the processing times while preserving the BM detection performance. Our previous implementation of the BM detection algorithm utilized Laplacian of Gaussians (LoG) for the candidate selection portion of the solution. In this study, we introduce a novel BM candidate detection CNN (cdCNN) to replace this classical IP stage. The network is formulated to have (1) a similar receptive field as the LoG method, and (2) a bias for the detection of BM lesion loci. The proposed CNN is later augmented with a classification CNN to perform the BM detection task. The cdCNN achieved 97.4% BM detection sensitivity when producing 60 K candidates per 3D MRI dataset, while the LoG achieved 96.5% detection sensitivity with 73 K candidates. The augmented BM detection framework generated on average 9.20 false-positive BM detections per patient for 90% sensitivity, which is comparable with our previous results. However, it processes each 3D data in 1.9 s, presenting a 93.5% reduction in the computation time.

Training Strategies for Radiology Deep Learning Models in Data-limited Scenarios.

Data-driven approaches have great potential to shape future practices in radiology. The most straightforward strategy to obtain clinically accurate models is to use large, well-curated and annotated datasets. However, patient privacy constraints, tedious annotation processes, and the limited availability of radiologists pose challenges to building such datasets. This review details model training strategies in scenarios with limited data, insufficiently labeled data, and/or limited expert resources. This review discusses strategies to enlarge the data sample, decrease the time burden of manual supervised labeling, adjust the neural network architecture to improve model performance, apply semisupervised approaches, and leverage efficiencies from pretrained models. Keywords: Computer-aided Detection/Diagnosis, Transfer Learning, Limited Annotated Data, Augmentation, Synthetic Data, Semisupervised Learning, Federated Learning, Few-Shot Learning, Class Imbalance.

Natural Language Processing of Radiology Text Reports: Interactive Text Classification.

This report presents a hands-on introduction to natural language processing (NLP) of radiology reports with deep neural networks in Google Colaboratory (Colab) to introduce readers to the rapidly evolving field of NLP. The implementation of the Google Colab notebook was designed with code hidden to facilitate learning for noncoders (ie, individuals with little or no computer programming experience). The data used for this module are the corpus of radiology reports from the Indiana University chest x-ray collection available from the National Library of Medicine's Open-I service. The module guides learners through the process of exploring the data, splitting the data for model training and testing, preparing the data for NLP analysis, and training a deep NLP model to classify the reports as normal or abnormal. Concepts in NLP, such as tokenization, numericalization, language modeling, and word embeddings, are demonstrated in the module. The module is implemented in a guided fashion with the authors presenting the material and explaining concepts. Interactive features and extensive text commentary are provided directly in the notebook to facilitate self-guided learning and experimentation with the module. Keywords: Neural Networks, Negative Expression Recognition, Natural Language Processing, Computer Applications, Informatics © RSNA, 2021.