Integration and evaluation of chest X-ray artificial intelligence in clinical practice.

Purpose: To integrate and evaluate an artificial intelligence (AI) system that assists in checking endotracheal tube (ETT) placement on chest x-rays (CXRs) in clinical practice. Approach: In clinical use over 17 months, 214 CXR images were ordered to check ETT placement with AI assistance by intensive care unit (ICU) physicians. The system was built on the SimpleMind Cognitive AI platform and integrated into a clinical workflow. It automatically identified the ETT and checked its placement relative to the trachea and carina. The ETT overlay and misplacement alert messages generated by the AI system were compared with radiology reports as the reference. A survey study was also conducted to evaluate usefulness of the AI system in clinical practice. Results: The alert messages indicating that either the ETT was misplaced or not detected had a positive predictive value of 42% (21/50) and negative predictive value of 98% (161/164) based on the radiology reports. In the survey, radiologist and ICU physician users indicated that they agreed with the AI outputs and that they were useful. Conclusions: The AI system performance in real-world clinical use was comparable to that seen in previous experiments. Based on this and physician survey results, the system can be deployed more widely at our institution, using insights gained from this evaluation to make further algorithm improvements and quality assurance of the AI system.

Translating AI to Clinical Practice: Overcoming Data Shift with Explainability.

To translate artificial intelligence (AI) algorithms into clinical practice requires generalizability of models to real-world data. One of the main obstacles to generalizability is data shift, a data distribution mismatch between model training and real environments. Explainable AI techniques offer tools to detect and mitigate the data shift problem and develop reliable AI for clinical practice. Most medical AI is trained with datasets gathered from limited environments, such as restricted disease populations and center-dependent acquisition conditions. The data shift that commonly exists in the limited training set often causes a significant performance decrease in the deployment environment. To develop a medical application, it is important to detect potential data shift and its impact on clinical translation. During AI training stages, from premodel analysis to in-model and post hoc explanations, explainability can play a key role in detecting model susceptibility to data shift, which is otherwise hidden because the test data have the same biased distribution as the training data. Performance-based model assessments cannot effectively distinguish the model overfitting to training data bias without enriched test sets from external environments. In the absence of such external data, explainability techniques can aid in translating AI to clinical practice as a tool to detect and mitigate potential failures due to data shift. ©RSNA, 2023 Quiz questions for this article are available in the supplemental material.

Republished: Robotic assistant spinal angiography: a case report and technical considerations.

Robotic-assisted technology has shown to be promising in coronary and peripheral vascular interventions. Early case reports have also demonstrated its efficacy in neuro-interventions. However, there is no prior report demonstrating use of the robotic-assisted platform for spinal angiography. We report the feasibility of the robotic-assisted thoracic and lumbar spinal angiography.

Automated Endotracheal Tube Placement Check Using Semantically Embedded Deep Neural Networks.

RATIONALE AND OBJECTIVES: To develop artificial intelligence (AI) system that assists in checking endotracheal tube (ETT) placement on chest X-rays (CXRs) and evaluate whether it can move into clinical validation as a quality improvement tool. MATERIALS AND METHODS: A retrospective data set including 2000 de-identified images from intensive care unit patients was split into 1488 for training and 512 for testing. AI was developed to automatically identify the ETT, trachea, and carina using semantically embedded neural networks that combine a declarative knowledge base with deep neural networks. To check the ETT tip placement, a "safe zone" was computed as the region inside the trachea and 3-7 cm above the carina. Two AI outputs were evaluated: (1) ETT overlay, (2) ETT misplacement alert messages. Clinically relevant performance metrics were compared against prespecified thresholds of >85% overlay accuracy and positive predictive value (PPV) > 30% and negative predictive value NPV > 95% for alerts to move into clinical validation. RESULTS: An ETT was present in 285 of 512 test cases. The AI detected 95% (271/285) of ETTs, 233 (86%) of these with accurate tip localization. The system (correctly) did not generate an ETT overlay in 221/227 CXRs where the tube was absent for an overall overlay accuracy of 89% (454/512). The alert messages indicating that either the ETT was misplaced or not detected had a PPV of 83% (265/320) and NPV of 98% (188/192). CONCLUSION: The chest X-ray AI met prespecified performance thresholds to move into clinical validation.

Synthesizing MR Image Contrast Enhancement Using 3D High-Resolution ConvNets.

OBJECTIVE: Gadolinium-based contrast agents (GBCAs) have been widely used to better visualize disease in brain magnetic resonance imaging (MRI). However, gadolinium deposition within the brain and body has raised safety concerns about the use of GBCAs. Therefore, the development of novel approaches that can decrease or even eliminate GBCA exposure while providing similar contrast information would be of significant use clinically. METHODS: In this work, we present a deep learning based approach for contrast-enhanced T1 synthesis on brain tumor patients. A 3D high-resolution fully convolutional network (FCN), which maintains high resolution information through processing and aggregates multi-scale information in parallel, is designed to map pre-contrast MRI sequences to contrast-enhanced MRI sequences. Specifically, three pre-contrast MRI sequences, T1, T2 and apparent diffusion coefficient map (ADC), are utilized as inputs and the post-contrast T1 sequences are utilized as target output. To alleviate the data imbalance problem between normal tissues and the tumor regions, we introduce a local loss to improve the contribution of the tumor regions, which leads to better enhancement results on tumors. RESULTS: Extensive quantitative and visual assessments are performed, with our proposed model achieving a PSNR of 28.24 dB in the brain and 21.2 dB in tumor regions. CONCLUSION AND SIGNIFICANCE: Our results suggest the potential of substituting GBCAs with synthetic contrast images generated via deep learning.

Electronic Health Record-Integrated Tumor Board Application to Save Preparation Time and Reduce Errors.

PURPOSE: Multidisciplinary oncology meetings, or tumor boards (TBs), ensure and facilitate communication between specialties regarding the management of cancer cases to improve patient care. The organization of TB and the preparation and presentation of patient cases are typically inefficient processes that require the exchange of patient information via e-mail, the hunting for data and images in the electronic health record, and the copying and pasting of patient data into desktop presentation software. METHODS: We implemented a standards-based electronic health record-integrated application that automated several aspects of TB organization and preparation. We hypothesized that this application would increase the efficiency of TB preparation, reduce errors in patient entry, and enhance communication with the clinical team. Our experimental design used a prospective evaluation by pathologists who were timed in preparing for weekly TBs using both the new application and the conventional method. In addition, patient data entry errors associated with each method were tracked, and TB attendees completed a survey evaluating satisfaction with the new application. RESULTS: The total time savings for TB preparation using the digital TB application over the conventional method was 5 hours and 19 minutes, representing a 45% reduction in preparation time (P < .01). Survey results showed that 91% of respondents preferred the digital method and believed that it improved the flow of the TB meeting. In addition, most believed that the digital method had an impact on subsequent patient care. CONCLUSION: This study provides further evidence that new electronic systems have the potential to significantly improve the overall TB paradigm by optimizing and enhancing case organization, preparation, and presentation.

Angular needle tracker and stabilizer for image-guided interventions.

INTRODUCTION: Minimally invasive image-guided interventions have changed the face of procedural medicine. For these procedures, safety and efficacy depend on precise needle placement. Needle targeting devices help improve the accuracy of needle placement, but their use has not seen broad penetration. Some of these devices are costly and require major modifications to the clinical workflow. In this article, we developed a low-cost, disposable, and easy-to-use angulation tracking device, which was based on a redesigned commercial passive needle holder. MATERIAL AND METHODS: The new design provided real-time angulation information for needle tracking. In this design, two potentiometers were used as angulation sensors, and they were connected to two axes of the passive needle holder's arch structure through a 3 D-printed bridge structure. A control unit included an Arduino Pro Mini, a Bluetooth module, and two rechargeable batteries. The angulation was calculated and communicated in real time to a novel developed smartphone app, where real-time angulation information was displayed for guiding the operator to position the needle to the planned angles. RESULTS: The open-air test results showed that the average errors are 1.03° and 1.08° for left-right angulation and head-foot angulation, respectively. The animal cadaver tests revealed that the novel system had an average angular error of 3.2° and a radial distance error of 3.1 mm. CONCLUSIONS: The accuracy was comparable with some commercially available solutions. The novel and low-cost needle tracking device may find a role as part of a real-time precision approach to both planning and implementation of image-guided therapies.

Trends that Impact IR's Future.

The future of IR will evolve as a result of current trends in advances in medicine, disease biology, technology, and IR devices and accoutrements. Changes in the trends that lie at the center of the differentiation of IR from other treatment specialties will have the greatest impact. Differentiation revolves around image guidance knowledge and procedural treatment skills and involves three key components: access, mapping, and action, all of which have the common thread of imaging knowledge. The main trends that are discussed are: image-guided diagnostics (IgDx), image-guided treatment (IgRx), sub-specialization in IgRx, large device design for IgRx, multimodality IgRx, interdisciplinary IgRx, and decentralized IgRx growth. Success in attaining a patient-facing "front-line" patient position will determine the future not only of IR but of radiology as a field. IgRx is anti-commoditization immunization. KEY POINTS:: · It is useful to conceptually separate diagnostic (IgDx) and treatment (IgRx) IR procedures.. · Subspecialization in IgRx will innovate IR practices for all practitioners.. · Advances in IR-tailored imaging equipment and integration of multimodality imaging will create, expand, and facilitate new treatments.. · Other treatment disciplines will be integrated into IgRx in a complementary fashion.. · Expansion of IR services into outpatient imaging sites and outpatient clinics will help establish important direct patient care.. · The adoption of these trends will follow a "diffusion of innovations" sigmoid curve pattern spanning different time intervals.. CITATION FORMAT: · Enzmann D. Trends that Impact IR's Future. Fortschr Röntgenstr 2022; 194: 21 - 28.

Robotic assistant spinal angiography: a case report and technical considerations.

Robotic-assisted technology has shown to be promising in coronary and peripheral vascular interventions. Early case reports have also demonstrated its efficacy in neuro-interventions. However, there is no prior report demonstrating use of the robotic-assisted platform for spinal angiography. We report the feasibility of the robotic-assisted thoracic and lumbar spinal angiography.

Federated learning improves site performance in multicenter deep learning without data sharing.

OBJECTIVE: To demonstrate enabling multi-institutional training without centralizing or sharing the underlying physical data via federated learning (FL). MATERIALS AND METHODS: Deep learning models were trained at each participating institution using local clinical data, and an additional model was trained using FL across all of the institutions. RESULTS: We found that the FL model exhibited superior performance and generalizability to the models trained at single institutions, with an overall performance level that was significantly better than that of any of the institutional models alone when evaluated on held-out test sets from each institution and an outside challenge dataset. DISCUSSION: The power of FL was successfully demonstrated across 3 academic institutions while avoiding the privacy risk associated with the transfer and pooling of patient data. CONCLUSION: Federated learning is an effective methodology that merits further study to enable accelerated development of models across institutions, enabling greater generalizability in clinical use.

Continuous Learning AI in Radiology: Implementation Principles and Early Applications.

Artificial intelligence (AI) is becoming increasingly present in radiology and health care. This expansion is driven by the principal AI strengths: automation, accuracy, and objectivity. However, as radiology AI matures to become fully integrated into the daily radiology routine, it needs to go beyond replicating static models, toward discovering new knowledge from the data and environments around it. Continuous learning AI presents the next substantial step in this direction and brings a new set of opportunities and challenges. Herein, the authors discuss the main concepts and requirements for implementing continuous AI in radiology and illustrate them with examples from emerging applications.

Early-Stage Radiology Volume Effects and Considerations with the Coronavirus Disease 2019 (COVID-19) Pandemic: Adaptations, Risks, and Lessons Learned.

OBJECTIVE: The coronavirus disease 2019 (COVID-19) pandemic resulted in significant loss of radiologic volume as a result of shelter-at-home mandates and delay of non-time-sensitive imaging studies to preserve capacity for the pandemic. We analyze the volume-related impact of the COVID-19 pandemic on six academic medical systems (AMSs), three in high COVID-19 surge (high-surge) and three in low COVID-19 surge (low-surge) regions, and a large national private practice coalition. We sought to assess adaptations, risks of actions, and lessons learned. METHODS: Percent change of 2020 volume per week was compared with the corresponding 2019 volume calculated for each of the 14 imaging modalities and overall total, outpatient, emergency, and inpatient studies in high-surge AMSs and low-surge AMSs and the practice coalition. RESULTS: Steep examination volume drops occurred during week 11, with slow recovery starting week 17. The lowest total AMS volume drop was 40% compared with the same period the previous year, and the largest was 70%. The greatest decreases were seen with screening mammography and dual-energy x-ray absorptiometry scans, and the smallest decreases were seen with PET/CT, x-ray, and interventional radiology. Inpatient volume was least impacted compared with outpatient or emergency imaging. CONCLUSION: Large percentage drops in volume were seen from weeks 11 through 17, were seen with screening studies, and were larger for the high-surge AMSs than for the low-surge AMSs. The lowest drops in volume were seen with modalities in which delays in imaging had greater perceived adverse consequences.

Prostate Microstructure in Prostate Cancer Using 3-T MRI with Diffusion-Relaxation Correlation Spectrum Imaging: Validation with Whole-Mount Digital Histopathology.

Background Microstructural MRI has the potential to improve diagnosis and characterization of prostate cancer (PCa), but validation with histopathology is lacking. Purpose To validate ex vivo diffusion-relaxation correlation spectrum imaging (DR-CSI) in the characterization of microstructural tissue compartments in prostate specimens from men with PCa by using registered whole-mount digital histopathology (WMHP) as the reference standard. Materials and Methods Men with PCa who underwent 3-T MRI and robotic-assisted radical prostatectomy between June 2018 and January 2019 were prospectively studied. After prostatectomy, the fresh whole prostate specimens were imaged in patient-specific three-dimensionally printed molds by using 3-T MRI with DR-CSI and were then sliced to create coregistered WMHP slides. The DR-CSI spectral signal component fractions (fA, fB, fC) were compared with epithelial, stromal, and luminal area fractions (fepithelium, fstroma, flumen) quantified in PCa and benign tissue regions. A linear mixed-effects model assessed the correlations between (fA, fB, fC) and (fepithelium, fstroma, flumen), and the strength of correlations was evaluated by using Spearman correlation coefficients. Differences between PCa and benign tissues in terms of DR-CSI signal components and microscopic tissue compartments were assessed using two-sided t tests. Results Prostate specimens from nine men (mean age, 65 years ± 7 [standard deviation]) were evaluated; 20 regions from 17 PCas, along with 20 benign tissue regions of interest, were analyzed. Three DR-CSI spectral signal components (spectral peaks) were consistently identified. The fA, fB, and fC were correlated with fepithelium, fstroma, and flumen (all P < .001), with Spearman correlation coefficients of 0.74 (95% confidence interval [CI]: 0.62, 0.83), 0.80 (95% CI: 0.66, 0.89), and 0.67 (95% CI: 0.51, 0.81), respectively. PCa exhibited differences compared with benign tissues in terms of increased fA (PCa vs benign, 0.37 ± 0.05 vs 0.27 ± 0.06; P < .001), decreased fC (PCa vs benign, 0.18 ± 0.06 vs 0.31 ± 0.13; P = .01), increased fepithelium (PCa vs benign, 0.44 ± 0.13 vs 0.26 ± 0.16; P < .001), and decreased flumen (PCa vs benign, 0.14 ± 0.08 vs 0.27 ± 0.18; P = .004). Conclusion Diffusion-relaxation correlation spectrum imaging signal components correlate with microscopic tissue compartments in the prostate and differ between cancer and benign tissue. © RSNA, 2020 Online supplemental material is available for this article. See also the editorial by Lee and Hectors in this issue.

Radiology's Information Architecture Could Migrate to One Emulating That of Smartphones.

Diagnostic radiology (DxR), having had successful serial co-evolutions with imaging equipment and PACS, is faced with another. With a backdrop termed "globotics transition," it should create an IT and informatics infrastructure capable of integrating artificial intelligence (AI) into current critical communication functions of PACS and incorporating functions currently residing in balkanized products. DxR will face the challenge of adopting sustaining and disruptive AI innovations simultaneously. In this co-evolution, a major selection force for AI will be increasing the flow of information and patients; "increasing" means faster flow over larger areas defined by geography and content. Larger content includes a broader spectrum of imaging and nonimaging information streams that facilitate medical decision making. Evolution to faster flow will gravitate toward a hierarchical IT architecture consisting of many small channels feeding into fewer larger channels, something potentially difficult for current PACS. Smartphone-like architecture optimized for communication and integration could provide a large-channel backbone and many smaller feeding channels for basic functions, as well as those needing to innovate rapidly. New, more flexible architectures stimulate market competition in which DxR could act as an artificial selection force to influence development of faster increased flow in current PACS companies, in disruptors such as consolidated AI companies, or in entirely new entrants like Apple or Google. In this co-evolution, DxR should be able to stimulate design of a modern communication medium that increases the flow of information and decreases the time and energy necessary to absorb it, thereby creating even more indispensable clinical value for itself.

Joint Prostate Cancer Detection and Gleason Score Prediction in mp-MRI via FocalNet.

Multi-parametric MRI (mp-MRI) is considered the best non-invasive imaging modality for diagnosing prostate cancer (PCa). However, mp-MRI for PCa diagnosis is currently limited by the qualitative or semi-quantitative interpretation criteria, leading to inter-reader variability and a suboptimal ability to assess lesion aggressiveness. Convolutional neural networks (CNNs) are a powerful method to automatically learn the discriminative features for various tasks, including cancer detection. We propose a novel multi-class CNN, FocalNet, to jointly detect PCa lesions and predict their aggressiveness using Gleason score (GS). FocalNet characterizes lesion aggressiveness and fully utilizes distinctive knowledge from mp-MRI. We collected a prostate mp-MRI dataset from 417 patients who underwent 3T mp-MRI exams prior to robotic-assisted laparoscopic prostatectomy. FocalNet was trained and evaluated in this large study cohort with fivefold cross validation. In the free-response receiver operating characteristics (FROC) analysis for lesion detection, FocalNet achieved 89.7% and 87.9% sensitivity for index lesions and clinically significant lesions at one false positive per patient, respectively. For the GS classification, evaluated by the receiver operating characteristics (ROC) analysis, FocalNet received the area under the curve of 0.81 and 0.79 for the classifications of clinically significant PCa (GS ≥ 3 + 4) and PCa with GS ≥ 4 + 3, respectively. With the comparison to the prospective performance of radiologists using the current diagnostic guideline, FocalNet demonstrated comparable detection sensitivity for index lesions and clinically significant lesions, only 3.4% and 1.5% lower than highly experienced radiologists without statistical significance.

Deep transfer learning-based prostate cancer classification using 3 Tesla multi-parametric MRI.

PURPOSE: The purpose of the study was to propose a deep transfer learning (DTL)-based model to distinguish indolent from clinically significant prostate cancer (PCa) lesions and to compare the DTL-based model with a deep learning (DL) model without transfer learning and PIRADS v2 score on 3 Tesla multi-parametric MRI (3T mp-MRI) with whole-mount histopathology (WMHP) validation. METHODS: With IRB approval, 140 patients with 3T mp-MRI and WMHP comprised the study cohort. The DTL-based model was trained on 169 lesions in 110 arbitrarily selected patients and tested on the remaining 47 lesions in 30 patients. We compared the DTL-based model with the same DL model architecture trained from scratch and the classification based on PIRADS v2 score with a threshold of 4 using accuracy, sensitivity, specificity, and area under curve (AUC). Bootstrapping with 2000 resamples was performed to estimate the 95% confidence interval (CI) for AUC. RESULTS: After training on 169 lesions in 110 patients, the AUC of discriminating indolent from clinically significant PCa lesions of the DTL-based model, DL model without transfer learning and PIRADS v2 score ≥ 4 were 0.726 (CI [0.575, 0.876]), 0.687 (CI [0.532, 0.843]), and 0.711 (CI [0.575, 0.847]), respectively, in the testing set. The DTL-based model achieved higher AUC compared to the DL model without transfer learning and PIRADS v2 score ≥ 4 in discriminating clinically significant lesions in the testing set. CONCLUSION: The DeLong test indicated that the DTL-based model achieved comparable AUC compared to the classification based on PIRADS v2 score (p = 0.89).

A system using patient-specific 3D-printed molds to spatially align in vivo MRI with ex vivo MRI and whole-mount histopathology for prostate cancer research.

BACKGROUND: Patient-specific 3D-printed molds and ex vivo MRI of the resected prostate have been two important strategies to align MRI with whole-mount histopathology (WMHP) for prostate cancer (PCa) research, but the combination of these two strategies has not been systematically evaluated. PURPOSE: To develop and evaluate a system that combines patient-specific 3D-printed molds with ex vivo MRI (ExV) to spatially align in vivo MRI (InV), ExV, and WMHP in PCa patients. STUDY TYPE: Prospective cohort study. POPULATION: Seventeen PCa patients who underwent 3T MRI and robotic-assisted laparoscopic radical prostatectomy (RALP). FIELD STRENGTH/SEQUENCES: T2 -weighted turbo spin-echo sequences at 3T. ASSESSMENT: Immediately after RALP, the fresh whole prostate specimens were imaged in patient-specific 3D-printed molds by 3T MRI and then sectioned to create WMHP slides. The time required for ExV was measured to assess impact on workflow. InV, ExV, and WMHP images were registered. Spatial alignment was evaluated using: slide offset (mm) between ExV slice locations and WMHP slides; overlap of the 3D prostate contour on InV versus ExV using Dice's coefficient (0 to 1); and 2D target registration error (TRE, mm) between corresponding landmarks on InV, ExV, and WMHP. Data are reported as mean ± standard deviation (SD). STATISTICAL TESTING: Differences in 2D TRE before versus after registration were compared using the Wilcoxon signed-rank test (P < 0.05 considered significant). RESULTS: ExV (duration 115 ± 15 min) was successfully incorporated into the workflow for all cases. Absolute slide offset was 1.58 ± 1.57 mm. Dice's coefficient was 0.865 ± 0.035. 2D TRE was significantly reduced after registration (P < 0.01) with mean (±SD of per patient means) of 1.9 ± 0.6 mm for InV versus ExV, 1.4 ± 0.5 mm for WMHP versus ExV, and 2.0 ± 0.5 mm for WMHP versus InV. DATA CONCLUSION: The proposed system combines patient-specific 3D-printed molds with ExV to achieve spatial alignment between InV, ExV, and WMHP with mean 2D TRE of 1-2 mm. LEVEL OF EVIDENCE: 2 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2019;49:270-279.