Performance of a Deep Learning Algorithm for Automated Segmentation and Quantification of Traumatic Pelvic Hematomas on CT

Applications of Deep Learning in Trauma Radiology: A Narrative Review

Diagnostic imaging is essential in modern trauma care for initial evaluation and identifying injuries requiring intervention. Deep learning (DL) has become mainstream in medical image analysis and has shown promising efficacy for classification, segmentation, and lesion detection. This narrative review provides the fundamental concepts for developing DL algorithms in trauma imaging and presents an overview of current progress in each modality. DL has been applied to detect free fluid on Focused Assessment with Sonography for Trauma (FAST), traumatic findings on chest and pelvic X-rays, and computed tomography (CT) scans, identify intracranial hemorrhage on head CT, detect vertebral fractures, and identify injuries to organs like the spleen, liver, and lungs on abdominal and chest CT. Future directions involve expanding dataset size and diversity through federated learning, enhancing model explainability and transparency to build clinician trust, and integrating multimodal data to provide more meaningful insights into traumatic injuries. Though some commercial artificial intelligence products are Food and Drug Administration-approved for clinical use in the trauma field, adoption remains limited, highlighting the need for multi-disciplinary teams to engineer practical, real-world solutions. Overall, DL shows immense potential to improve the efficiency and accuracy of trauma imaging, but thoughtful development and validation are critical to ensure these technologies positively impact patient care.

Deep Learning for Automated Detection and Localization of Traumatic Abdominal Solid Organ Injuries on CT Scans

Computed tomography (CT) is the most commonly used diagnostic modality for blunt abdominal trauma (BAT), significantly influencing management approaches. Deep learning models (DLMs) have shown great promise in enhancing various aspects of clinical practice. There is limited literature available on the use of DLMs specifically for trauma image evaluation. In this study, we developed a DLM aimed at detecting solid organ injuries to assist medical professionals in rapidly identifying life-threatening injuries. The study enrolled patients from a single trauma center who received abdominal CT scans between 2008 and 2017. Patients with spleen, liver, or kidney injury were categorized as the solid organ injury group, while others were considered negative cases. Only images acquired from the trauma center were enrolled. A subset of images acquired in the last year was designated as the test set, and the remaining images were utilized to train and validate the detection models. The performance of each model was assessed using metrics such as the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value, and negative predictive value based on the best Youden index operating point. The study developed the models using 1302 (87%) scans for training and tested them on 194 (13%) scans. The spleen injury model demonstrated an accuracy of 0.938 and a specificity of 0.952. The accuracy and specificity of the liver injury model were reported as 0.820 and 0.847, respectively. The kidney injury model showed an accuracy of 0.959 and a specificity of 0.989. We developed a DLM that can automate the detection of solid organ injuries by abdominal CT scans with acceptable diagnostic accuracy. It cannot replace the role of clinicians, but we can expect it to be a potential tool to accelerate the process of therapeutic decisions for trauma care.

A vendor-agnostic, PACS integrated, and DICOM-compatible software-server pipeline for testing segmentation algorithms within the clinical radiology workflow

Background

Reproducible approaches are needed to bring AI/ML for medical image analysis closer to the bedside. Investigators wishing to shadow test cross-sectional medical imaging segmentation algorithms on new studies in real-time will benefit from simple tools that integrate PACS with on-premises image processing, allowing visualization of DICOM-compatible segmentation results and volumetric data at the radiology workstation.

Purpose

In this work, we develop and release a simple containerized and easily deployable pipeline for shadow testing of segmentation algorithms within the clinical workflow.

Methods

Our end-to-end automated pipeline has two major components- 1. A router/listener and anonymizer and an OHIF web viewer backstopped by a DCM4CHEE DICOM query/retrieve archive deployed in the virtual infrastructure of our secure hospital intranet, and 2. An on-premises single GPU workstation host for DICOM/NIfTI conversion steps, and image processing. DICOM images are visualized in OHIF along with their segmentation masks and associated volumetry measurements (in mL) using DICOM SEG and structured report (SR) elements. Since nnU-net has emerged as a widely-used out-of-the-box method for training segmentation models with state-of-the-art performance, feasibility of our pipleine is demonstrated by recording clock times for a traumatic pelvic hematoma nnU-net model.

Results

Mean total clock time from PACS send by user to completion of transfer to the DCM4CHEE query/retrieve archive was 5 min 32 s (± SD of 1 min 26 s). This compares favorably to the report turnaround times for whole-body CT exams, which often exceed 30 min, and illustrates feasibility in the clinical setting where quantitative results would be expected prior to report sign-off. Inference times accounted for most of the total clock time, ranging from 2 min 41 s to 8 min 27 s. All other virtual and on-premises host steps combined ranged from a minimum of 34 s to a maximum of 48 s.

Conclusion

The software worked seamlessly with an existing PACS and could be used for deployment of DL models within the radiology workflow for prospective testing on newly scanned patients. Once configured, the pipeline is executed through one command using a single shell script. The code is made publicly available through an open-source license at "https://github.com/vastc/," and includes a readme file providing pipeline config instructions for host names, series filter, other parameters, and citation instructions for this work.

Pulmonary contusion: automated deep learning-based quantitative visualization

PURPOSE

Rapid automated CT volumetry of pulmonary contusion may predict progression to Acute Respiratory Distress Syndrome (ARDS) and help guide early clinical management in at-risk trauma patients. This study aims to train and validate state-of-the-art deep learning models to quantify pulmonary contusion as a percentage of total lung volume (Lung Contusion Index, or auto-LCI) and assess the relationship between auto-LCI and relevant clinical outcomes.

METHODS

302 adult patients (age ≥ 18) with pulmonary contusion were retrospectively identified from reports between 2016 and 2021. nnU-Net was trained on manual contusion and whole-lung segmentations. Point-of-care candidate variables for multivariate regression included oxygen saturation, heart rate, and systolic blood pressure on admission. Logistic regression was used to assess ARDS risk, and Cox proportional hazards models were used to determine differences in ICU length of stay and mechanical ventilation time.

RESULTS

Mean Volume Similarity Index and mean Dice scores were 0.82 and 0.67. Interclass correlation coefficient and Pearson r between ground-truth and predicted volumes were 0.90 and 0.91. 38 (14%) patients developed ARDS. In bivariate analysis, auto-LCI was associated with ARDS (p < 0.001), ICU admission (p < 0.001), and need for mechanical ventilation (p < 0.001). In multivariate analyses, auto-LCI was associated with ARDS (p = 0.04), longer length of stay in the ICU (p = 0.02) and longer time on mechanical ventilation (p = 0.04). AUC of multivariate regression to predict ARDS using auto-LCI and clinical variables was 0.70 while AUC using auto-LCI alone was 0.68.

CONCLUSION

Increasing auto-LCI values corresponded with increased risk of ARDS, longer ICU admissions, and longer periods of mechanical ventilation.

Accelerating voxelwise annotation of cross-sectional imaging through AI collaborative labeling with quality assurance and bias mitigation

Background

precision-medicine quantitative tools for cross-sectional imaging require painstaking labeling of targets that vary considerably in volume, prohibiting scaling of data annotation efforts and supervised training to large datasets for robust and generalizable clinical performance. A straight-forward time-saving strategy involves manual editing of AI-generated labels, which we call AI-collaborative labeling (AICL). Factors affecting the efficacy and utility of such an approach are unknown. Reduction in time effort is not well documented. Further, edited AI labels may be prone to automation bias.

Purpose

In this pilot, using a cohort of CTs with intracavitary hemorrhage, we evaluate both time savings and AICL label quality and propose criteria that must be met for using AICL annotations as a high-throughput, high-quality ground truth.

Methods

57 CT scans of patients with traumatic intracavitary hemorrhage were included. No participant recruited for this study had previously interpreted the scans. nnU-net models trained on small existing datasets for each feature (hemothorax/hemoperitoneum/pelvic hematoma; n = 77-253) were used in inference. Two common scenarios served as baseline comparison- de novo expert manual labeling, and expert edits of trained staff labels. Parameters included time effort and image quality graded by a blinded independent expert using a 9-point scale. The observer also attempted to discriminate AICL and expert labels in a random subset (n = 18). Data were compared with ANOVA and post-hoc paired signed rank tests with Bonferroni correction.

Results

AICL reduced time effort 2.8-fold compared to staff label editing, and 8.7-fold compared to expert labeling (corrected p < 0.0006). Mean Likert grades for AICL (8.4, SD:0.6) were significantly higher than for expert labels (7.8, SD:0.9) and edited staff labels (7.7, SD:0.8) (corrected p < 0.0006). The independent observer failed to correctly discriminate AI and human labels.

Conclusion

For our use case and annotators, AICL facilitates rapid large-scale curation of high-quality ground truth. The proposed quality control regime can be employed by other investigators prior to embarking on AICL for segmentation tasks in large datasets.

A survey of ASER members on artificial intelligence in emergency radiology: trends, perceptions, and expectations

PURPOSE

There is a growing body of diagnostic performance studies for emergency radiology-related artificial intelligence/machine learning (AI/ML) tools; however, little is known about user preferences, concerns, experiences, expectations, and the degree of penetration of AI tools in emergency radiology. Our aim is to conduct a survey of the current trends, perceptions, and expectations regarding AI among American Society of Emergency Radiology (ASER) members.

METHODS

An anonymous and voluntary online survey questionnaire was e-mailed to all ASER members, followed by two reminder e-mails. A descriptive analysis of the data was conducted, and results summarized.

RESULTS

A total of 113 members responded (response rate 12%). The majority were attending radiologists (90%) with greater than 10 years' experience (80%) and from an academic practice (65%). Most (55%) reported use of commercial AI CAD tools in their practice. Workflow prioritization based on pathology detection, injury or disease severity grading and classification, quantitative visualization, and auto-population of structured reports were identified as high-value tasks. Respondents overwhelmingly indicated a need for explainable and verifiable tools (87%) and the need for transparency in the development process (80%). Most respondents did not feel that AI would reduce the need for emergency radiologists in the next two decades (72%) or diminish interest in fellowship programs (58%). Negative perceptions pertained to potential for automation bias (23%), over-diagnosis (16%), poor generalizability (15%), negative impact on training (11%), and impediments to workflow (10%).

CONCLUSION

ASER member respondents are in general optimistic about the impact of AI in the practice of emergency radiology and its impact on the popularity of emergency radiology as a subspecialty. The majority expect to see transparent and explainable AI models with the radiologist as the decision-maker.

Artificial intelligence CAD tools in trauma imaging: a scoping review from the American Society of Emergency Radiology (ASER) AI/ML Expert Panel

BACKGROUND

AI/ML CAD tools can potentially improve outcomes in the high-stakes, high-volume model of trauma radiology. No prior scoping review has been undertaken to comprehensively assess tools in this subspecialty.

PURPOSE

To map the evolution and current state of trauma radiology CAD tools along key dimensions of technology readiness.

METHODS

Following a search of databases, abstract screening, and full-text document review, CAD tool maturity was charted using elements of data curation, performance validation, outcomes research, explainability, user acceptance, and funding patterns. Descriptive statistics were used to illustrate key trends.

RESULTS

A total of 4052 records were screened, and 233 full-text articles were selected for content analysis. Twenty-one papers described FDA-approved commercial tools, and 212 reported algorithm prototypes. Works ranged from foundational research to multi-reader multi-case trials with heterogeneous external data. Scalable convolutional neural network-based implementations increased steeply after 2016 and were used in all commercial products; however, options for explainability were narrow. Of FDA-approved tools, 9/10 performed detection tasks. Dataset sizes ranged from < 100 to > 500,000 patients, and commercialization coincided with public dataset availability. Cross-sectional torso datasets were uniformly small. Data curation methods with ground truth labeling by independent readers were uncommon. No papers assessed user acceptance, and no method included human-computer interaction. The USA and China had the highest research output and frequency of research funding.

CONCLUSIONS

Trauma imaging CAD tools are likely to improve patient care but are currently in an early stage of maturity, with few FDA-approved products for a limited number of uses. The scarcity of high-quality annotated data remains a major barrier.

A vendor-agnostic, PACS integrated, and DICOMcompatible software-server pipeline for testing segmentation algorithms within the clinical radiology workflow

Background

Reproducible approaches are needed to bring AI/ML for medical image analysis closer to the bedside. Investigators wishing to shadow test cross-sectional medical imaging segmentation algorithms on new studies in real-time will benefit from simple tools that integrate PACS with on-premises image processing, allowing visualization of DICOM-compatible segmentation results and volumetric data at the radiology workstation.

Purpose

In this work, we develop and release a simple containerized and easily deployable pipeline for shadow testing of segmentation algorithms within the clinical workflow.

Methods

Our end-to-end automated pipeline has two major components-1. a router/listener and anonymizer and an OHIF web viewer backstopped by a DCM4CHEE DICOM query/retrieve archive deployed in the virtual infrastructure of our secure hospital intranet, and 2. An on-premises single GPU workstation host for DICOM/NIfTI conversion steps, and image processing. DICOM images are visualized in OHIF along with their segmentation masks and associated volumetry measurements (in mL) using DICOM SEG and structured report (SR) elements. Feasibility is demonstrated by recording clock times for a traumatic pelvic hematoma cascaded nnU-net model.

Results

Mean total clock time from PACS send by user to completion of transfer to the DCM4CHEE query/retrieve archive was 5 minutes 32 seconds (+/- SD of 1 min 26 sec). This compares favorably to the report turnaround times for whole-body CT exams, which often exceed 30 minutes. Inference times accounted for most of the total clock time, ranging from 2 minutes 41 seconds to 8 minutes 27 seconds. All other virtual and on-premises host steps combined ranged from a minimum of 34 seconds to a maximum of 48 seconds.

Conclusion

The software worked seamlessly with an existing PACS and could be used for deployment of DL models within the radiology workflow for prospective testing on newly scanned patients. Once configured, the pipeline is executed through one command using a single shell script. The code is made publicly available through an open-source license at "https://github.com/vastc/", and includes a readme file providing pipeline config instructions for host names, series filter, other parameters, and citation instructions for this work.

Science fiction or clinical reality: a review of the applications of artificial intelligence along the continuum of trauma care

Artificial intelligence (AI) and machine learning describe a broad range of algorithm types that can be trained based on datasets to make predictions. The increasing sophistication of AI has created new opportunities to apply these algorithms within within trauma care. Our paper overviews the current uses of AI along the continuum of trauma care, including injury prediction, triage, emergency department volume, assessment, and outcomes. Starting at the point of injury, algorithms are being used to predict severity of motor vehicle crashes, which can help inform emergency responses. Once on the scene, AI can be used to help emergency services triage patients remotely in order to inform transfer location and urgency. For the receiving hospital, these tools can be used to predict trauma volumes in the emergency department to help allocate appropriate staffing. After patient arrival to hospital, these algorithms not only can help to predict injury severity, which can inform decision-making, but also predict patient outcomes to help trauma teams anticipate patient trajectory. Overall, these tools have the capability to transform trauma care. AI is still nascent within the trauma surgery sphere, but this body of the literature shows that this technology has vast potential. AI-based predictive tools in trauma need to be explored further through prospective trials and clinical validation of algorithms.

Toward automated interpretable AAST grading for blunt splenic injury

BACKGROUND

The American Association for the Surgery of Trauma (AAST) splenic organ injury scale (OIS) is the most frequently used CT-based grading system for blunt splenic trauma. However, reported inter-rater agreement is modest, and an algorithm that objectively automates grading based on transparent and verifiable criteria could serve as a high-trust diagnostic aid.

PURPOSE

To pilot the development of an automated interpretable multi-stage deep learning-based system to predict AAST grade from admission trauma CT.

METHODS

Our pipeline includes 4 parts: (1) automated splenic localization, (2) Faster R-CNN-based detection of pseudoaneurysms (PSA) and active bleeds (AB), (3) nnU-Net segmentation and quantification of splenic parenchymal disruption (SPD), and (4) a directed graph that infers AAST grades from detection and segmentation results. Training and validation is performed on a dataset of adult patients (age ≥ 18) with voxelwise labeling, consensus AAST grading, and hemorrhage-related outcome data (n = 174).

RESULTS

AAST classification agreement (weighted κ) between automated and consensus AAST grades was substantial (0.79). High-grade (IV and V) injuries were predicted with accuracy, positive predictive value, and negative predictive value of 92%, 95%, and 89%. The area under the curve for predicting hemorrhage control intervention was comparable between expert consensus and automated AAST grading (0.83 vs 0.88). The mean combined inference time for the pipeline was 96.9 s.

CONCLUSIONS

The results of our method were rapid and verifiable, with high agreement between automated and expert consensus grades. Diagnosis of high-grade lesions and prediction of hemorrhage control intervention produced accurate results in adult patients.

Blunt splenic injury in adults: Association between volumetric quantitative CT parameters and intervention

BACKGROUND.

Several ordinal grading systems are employed in deciding whether to perform angioembolization or splenectomy following blunt splenic injury. The 2018 AAST Organ Injury Scale (OIS) incorporates vascular lesions but not hemoperitoneum, which is considered in the Thompson classifier. Granular and verifiable quantitative measurements of these features may have a future role in facilitating objective decision-making.

PURPOSE.

To compare performance of CT volumetry-based quantitative modeling to the 1994 and 2018 AAST OIS and Thompson classifier for the following endpoints: decision to perform splenectomy (SPY), and the composite of SPY or angioembolization (AE)

MATERIALS AND METHODS.

Adult BSI patients (age ≥ 18 years) scanned with dual-phase CT prior to intervention at a single level I trauma center from 2017-2019 were included in this retrospective study (n=174). Scoring using 2018 AAST, 1994 AAST, and Thompson systems was performed retrospectively by two radiologists and arbitrated by a third. Endpoints included 1. SPY and 2. The composite of SPY or AE. Logistic regression models were developed from segmented active bleed, contained vascular lesion, splenic parenchymal disruption, and hemoperitoneum volumes. AUCs for ordinal systems and volumetric models were compared.

RESULTS.

Forty-seven BSI patients (27%) underwent SPY, and 87 patients (50%) underwent SPY or AE. Quantitative model AUCs (0.85- SPY, 0.82-composite) were not significantly different from 2018 AAST AUCs (0.81, 0.88, p=0.66, 0.14) for both endpoints, and were significantly improved over Thompson scoring (0.76, p=0.02; 0.77, p=0.04).

CONCLUSION:

Quantitative CT volumetry can be used to model intervention for BSI with accuracy comparable to 2018 AAST scoring and significantly higher than Thompson scoring.

Study Type:

Prognostic

Level of Evidence:

IV

CT volumetry of blunt splenic injury-related features predicts splenectomy and angioembolization in adults and identifies clinically important target features for computer vision and automation research.

A pilot study of deep learning-based CT volumetry for traumatic hemothorax

PURPOSE

We employ nnU-Net, a state-of-the-art self-configuring deep learning-based semantic segmentation method for quantitative visualization of hemothorax (HTX) in trauma patients, and assess performance using a combination of overlap and volume-based metrics. The accuracy of hemothorax volumes for predicting a composite of hemorrhage-related outcomes - massive transfusion (MT) and in-hospital mortality (IHM) not related to traumatic brain injury - is assessed and compared to subjective expert consensus grading by an experienced chest and emergency radiologist.

MATERIALS AND METHODS

The study included manually labeled admission chest CTs from 77 consecutive adult patients with non-negligible (≥ 50 mL) traumatic HTX between 2016 and 2018 from one trauma center. DL results of ensembled nnU-Net were determined from fivefold cross-validation and compared to individual 2D, 3D, and cascaded 3D nnU-Net results using the Dice similarity coefficient (DSC) and volume similarity index. Pearson's r, intraclass correlation coefficient (ICC), and mean bias were also determined for the best performing model. Manual and automated hemothorax volumes and subjective hemothorax volume grades were analyzed as predictors of MT and IHM using AUC comparison. Volume cut-offs yielding sensitivity or specificity ≥ 90% were determined from ROC analysis.

RESULTS

Ensembled nnU-Net achieved a mean DSC of 0.75 (SD: ± 0.12), and mean volume similarity of 0.91 (SD: ± 0.10), Pearson r of 0.93, and ICC of 0.92. Mean overmeasurement bias was only 1.7 mL despite a range of manual HTX volumes from 35 to 1503 mL (median: 178 mL). AUC of automated volumes for the composite outcome was 0.74 (95%CI: 0.58-0.91), compared to 0.76 (95%CI: 0.58-0.93) for manual volumes, and 0.76 (95%CI: 0.62-0.90) for consensus expert grading (p = 0.93). Automated volume cut-offs of 77 mL and 334 mL predicted the outcome with 93% sensitivity and 90% specificity respectively.

CONCLUSION

Automated HTX volumetry had high method validity, yielded interpretable visual results, and had similar performance for the hemorrhage-related outcomes assessed compared to manual volumes and expert consensus grading. The results suggest promising avenues for automated HTX volumetry in research and clinical care.

Blunt splenic injury: Assessment of follow-up CT utility using quantitative volumetry

Purpose:

Trials of non-operative management (NOM) have become the standard of care for blunt splenic injury (BSI) in hemodynamically stable patients. However, there is a lack of consensus regarding the utility of follow-up CT exams and relevant CT features. The purpose of this study is to determine imaging predictors of splenectomy on follow-up CT using quantitative volumetric measurements.

Methods:

Adult patients who underwent a trial of non-operative management (NOM) with follow-up CT performed for BSI between 2017 and 2019 were included (n = 51). Six patients (12% of cohort) underwent splenectomy; 45 underwent successful splenic salvage. Voxelwise measurements of splenic laceration, hemoperitoneum, and subcapsular hematoma were derived from portal venous phase images of admission and follow-up scans using 3D slicer. Presence/absence of pseudoaneurysm on admission and follow-up CT was assessed using arterial phase images. Multivariable logistic regression was used to determine independent predictors of decision to perform splenectomy.

Results:

Factors significantly associated with splenectomy in bivariate analysis incorporated in multivariate logistic regression included final hemoperitoneum volume (p = 0.003), final subcapsular hematoma volume (p = 0.001), change in subcapsular hematoma volume between scans (p = 0.09) and new/persistent pseudoaneurysm (p = 0.003). Independent predictors of splenectomy in the logistic regression were final hemoperitoneum volume (unit OR = 1.43 for each 100 mL change; 95% CI: 0.99–2.06) and new/persistent pseudoaneurysm (OR = 160.3; 95% CI: 0.91–28315.3). The AUC of the model incorporating both variables was significantly higher than AAST grading (0.91 vs. 0.59, p = 0.025). Mean combined effective dose for admission and follow up CT scans was 37.4 mSv.

Conclusion:

Follow-up CT provides clinically valuable information regarding the decision to perform splenectomy in BSI patients managed non-operatively. Hemoperitoneum volume and new or persistent pseudoaneurysm at follow-up are independent predictors of splenectomy.

Evaluation of Traumatic Subdural Hematoma Volume by Using Image Segmentation Assessment Based on Deep Learning

Rapid and accurate evaluations of hematoma volume can guide the treatment of traumatic subdural hematoma. We aim to explore the consistency between the measurement results of traumatic subdural hematoma (TSDH) using a deep learn-based image segmentation algorithm. A retrospective study was conducted on 90 CT images of patients diagnosed with TSDH in our hospital from January 2019 to January 2022. All image data were measured by manual segmentation, convolutional neural networks (CNN) algorithm segmentation, and ABC/2 volume formula. With manual segmentation as the “golden standard,” a consistency test was carried out with CNN algorithm segmentation and ABC/2 volume formula, respectively. The percentage error of CNN algorithm segmentation is less than ABC/2 volume formula. There is no significant difference between CNN algorithm segmentation and manual segmentation (P > 0.05). The area under curve of the ABC/2 volume formula, manual segmentation, and CNN algorithm segmentation is 0.811 (95% CI: 0.717~0.905), 0.840 (95% CI: 0.753~0.928), and 0.832 (95% CI: 0.742~0.922), respectively. From our results, the algorithm based on CNN has a good efficiency in segmentation and accurate calculation of TSDH hematoma volume.

An Extra Set of Intelligent Eyes: Application of Artificial Intelligence in Imaging of Abdominopelvic Pathologies in Emergency Radiology

Imaging in the emergent setting carries high stakes. With increased demand for dedicated on-site service, emergency radiologists face increasingly large image volumes that require rapid turnaround times. However, novel artificial intelligence (AI) algorithms may assist trauma and emergency radiologists with efficient and accurate medical image analysis, providing an opportunity to augment human decision making, including outcome prediction and treatment planning. While traditional radiology practice involves visual assessment of medical images for detection and characterization of pathologies, AI algorithms can automatically identify subtle disease states and provide quantitative characterization of disease severity based on morphologic image details, such as geometry and fluid flow. Taken together, the benefits provided by implementing AI in radiology have the potential to improve workflow efficiency, engender faster turnaround results for complex cases, and reduce heavy workloads. Although analysis of AI applications within abdominopelvic imaging has primarily focused on oncologic detection, localization, and treatment response, several promising algorithms have been developed for use in the emergency setting. This article aims to establish a general understanding of the AI algorithms used in emergent image-based tasks and to discuss the challenges associated with the implementation of AI into the clinical workflow.

Volumetric Markers of Body Composition May Improve Personalized Prediction of Major Arterial Bleeding After Pelvic Fracture: A Secondary Analysis of the Baltimore CT Prediction Model Cohort

METHODS

This work is a retrospective secondary analysis of a single institution cohort used in the development of the Baltimore CT prediction model. The cohort includes 115 consecutive patients that underwent admission contrast-enhanced CT of the abdomen and pelvis for blunt trauma with pelvic ring disruption followed by conventional angiography. Major arterial injury requiring angioembolization served as the outcome variable. Angioembolization was required in 73/115 patients (63% of the cohort). Average age was 46.9 years (±SD 20.4). Body composition measurements were determined as 2-dimensional (2D) or 3-dimensional (3D) parameters and included mid-L3 trabecular bone attenuation, abdominal visceral fat area or volume, and percent muscle fat fraction (as a marker of sarcopenia) measured using segmentation and histogram analysis.

RESULTS

Models incorporating 2D (Model B) or 3D markers (model C) of body composition showed improvement over the original Baltimore model (model A) in all parameters of performance, quality, and fit (area under the receiver-operating curve [AUC], Akaike information criterion, Brier score, Hosmer-Lemeshow test, and adjusted-R2). Area under the receiver-operating curve increased from 0.83 (A), to 0.86 (B), and 0.88 (C). The greatest improvement was seen with 3D parameters.

CONCLUSION

Once automated, quantitative visualization tools providing "free" 3D body composition information can be expected to improve personalized precision diagnostics, outcome prediction, and decision support in patients with bleeding pelvic fractures.

A Coarse-to-Fine Framework for Automated Knee Bone and Cartilage Segmentation Data from the Osteoarthritis Initiative

Knee osteoarthritis (OA) is a degenerative joint disease that is prevalent in advancing age. The pathology of OA disease is still unclear, and there are no effective interventions that can completely alter the OA disease process. Magnetic resonance (MR) image evaluation is sensitive for depicting early changes of knee OA, and therefore important for early clinical intervention for relieving the symptom. Automated cartilage segmentation based on MR images is a vital step in experimental longitudinal studies to follow-up the patients and prospectively define a new quantitative marker from OA progression. In this paper, we develop a deep learning-based coarse-to-fine approach for automated knee bone, cartilage, and meniscus segmentation with high computational efficiency. The proposed method is evaluated using two-fold cross-validation on 507 MR volumes (81,120 slices) with OA from the Osteoarthritis Initiative (OAI)1 dataset. The mean dice similarity coefficients (DSCs) of femoral bone (FB), tibial bone (TB), femoral cartilage (FC), and tibial cartilage (TC) separately are 99.1%, 98.2%, 90.9%, and 85.8%. The time of segmenting each patient is 12 s, which is fast enough to be used in clinical practice. Our proposed approach may provide an automated toolkit to help computer-aided quantitative analyses of OA images.

Added value of deep learning-based liver parenchymal CT volumetry for predicting major arterial injury after blunt hepatic trauma: a decision tree analysis

PURPOSE

In patients presenting with blunt hepatic injury (BHI), the utility of CT for triage to hepatic angiography remains uncertain since simple binary assessment of contrast extravasation (CE) as being present or absent has only modest accuracy for major arterial injury on digital subtraction angiography (DSA). American Association for the Surgery of Trauma (AAST) liver injury grading is coarse and subjective, with limited diagnostic utility in this setting. Volumetric measurements of hepatic injury burden could improve prediction. We hypothesized that in a cohort of patients that underwent catheter-directed hepatic angiography following admission trauma CT, a deep learning quantitative visualization method that calculates % liver parenchymal disruption (the LPD index, or LPDI) would add value to CE assessment for prediction of major hepatic arterial injury (MHAI).

METHODS

This retrospective study included adult patients with BHI between 1/1/2008 and 5/1/2017 from two institutions that underwent admission trauma CT prior to hepatic angiography (n = 73). Presence (n = 41) or absence (n = 32) of MHAI (pseudoaneurysm, AVF, or active contrast extravasation on DSA) served as the outcome. Voxelwise measurements of liver laceration were derived using an existing multiscale deep learning algorithm trained on manually labeled data using cross-validation with a 75-25% split in four unseen folds. Liver volume was derived using a pre-trained whole liver segmentation algorithm. LPDI was automatically calculated for each patient by determining the percentage of liver involved by laceration. Classification and regression tree (CART) analyses were performed using a combination of automated LPDI measurements and either manually segmented CE volumes, or CE as a binary sign. Performance metrics for the decision rules were compared for significant differences with binary CE alone (the current standard of care for predicting MHAI), and the AAST grade.

RESULTS

36% of patients (n = 26) had contrast extravasation on CT. Median [Q1-Q3] automated LPDI was 4.0% [1.0-12.1%]. 41/73 (56%) of patients had MHAI. A decision tree based on auto-LPDI and volumetric CE measurements (CEvol) had the highest accuracy (0.84, 95% CI 0.73-0.91) with significant improvement over binary CE assessment (0.68, 95% CI 0.57-0.79; p = 0.01). AAST grades at different cut-offs performed poorly for predicting MHAI, with accuracies ranging from 0.44-0.63. Decision tree analysis suggests an auto-LPDI cut-off of ≥ 12% for minimizing false negative CT exams when CE is absent or diminutive.

CONCLUSION

Current CT imaging paradigms are coarse, subjective, and limited for predicting which BHIs are most likely to benefit from AE. LPDI, automated using deep learning methods, may improve objective personalized triage of BHI patients to angiography at the point of care.

A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises

Since its renaissance, deep learning has been widely used in various medical imaging tasks and has achieved remarkable success in many medical imaging applications, thereby propelling us into the so-called artificial intelligence (AI) era. It is known that the success of AI is mostly attributed to the availability of big data with annotations for a single task and the advances in high performance computing. However, medical imaging presents unique challenges that confront deep learning approaches. In this survey paper, we first present traits of medical imaging, highlight both clinical needs and technical challenges in medical imaging, and describe how emerging trends in deep learning are addressing these issues. We cover the topics of network architecture, sparse and noisy labels, federating learning, interpretability, uncertainty quantification, etc. Then, we present several case studies that are commonly found in clinical practice, including digital pathology and chest, brain, cardiovascular, and abdominal imaging. Rather than presenting an exhaustive literature survey, we instead describe some prominent research highlights related to these case study applications. We conclude with a discussion and presentation of promising future directions.

Volumetric quantitative measurement of hip effusions by manual versus automated artificial intelligence techniques: An OMERACT preliminary validation study

OBJECTIVE

Preliminary assessment, via OMERACT filter, of manual and automated MRI hip effusion Volumetric Quantitative Measurement (VQM).

METHODS

For 358 hips (93 osteoarthritis subjects, bilateral, 2 time points), 2 radiologists performed manual VQM using custom Matlab software. A Mask R-CNN artificial-intelligence (AI) tool was trained to automatically compute joint fluid volumes.

RESULTS

Manual VQM had excellent inter-observer reliability (ICC 0.96). AI predicted hip fluid volumes with ICC 0.86 (status), 0.58 (change) vs. 2 human readers.

CONCLUSION

Hip joint fluid volumes are reliably assessed by VQM. It is feasible to automate this approach using AI, with promising initial reliability.

Diagnostic value of CT contrast extravasation for major arterial injury after pelvic fracture: A meta-analysis

PURPOSE

We conducted a meta-analysis to determine diagnostic performance of CT intravenous contrast extravasation (CE) as a sign of angiographic bleeding and need for angioembolization after pelvic fractures.

MATERIALS AND METHODS

A systematic literature search combining the concepts of contrast extravasation, pelvic trauma, and CT yielded 206 potentially eligible studies. 23 studies provided accuracy data or sufficient descriptive data to allow 2x2 contingency table construction and provided 3855 patients for meta-analysis. Methodologic quality was assessed using the QUADAS-2 tool. Sensitivity and specificity were synthesized using bivariate mixed-effects logistic regression. Heterogeneity was assessed using the I2-statistic. Sources of heterogeneity explored included generation of scanner (64 row CT versus lower detector row) and use of multiphasic versus single phase scanning protocols.

RESULTS

Overall sensitivity and specificity were 80% (95% CI: 66-90%, I2 = 92.65%) and 93% (CI: 90-96, I2 = 89.34%), respectively. Subgroup analysis showed pooled sensitivity and specificity of 94% and 89% for 64- row CT compared to 69% and 95% with older generation scanners. CE had pooled sensitivity and specificity of 95% and 92% with the use of multiphasic protocols, compared to 74% and 94% with single-phase protocols.

CONCLUSION

The pooled sensitivity and specificity of 64-row CT was 94 and 89%. 64 row CT improves sensitivity of CE, which was 69% using lower detector row scanners. High specificity (92%) can be maintained by incorporating multiphasic scan protocols.

A Multiscale Deep Learning Method for Quantitative Visualization of Traumatic Hemoperitoneum at CT: Assessment of Feasibility and Comparison with Subjective Categorical Estimation

PURPOSE

To evaluate the feasibility of a multiscale deep learning algorithm for quantitative visualization and measurement of traumatic hemoperitoneum and to compare diagnostic performance for relevant outcomes with categorical estimation.

MATERIALS AND METHODS

This retrospective, single-institution study included 130 patients (mean age, 38 years; interquartile range, 25-50 years; 79 men) with traumatic hemoperitoneum who underwent CT of the abdomen and pelvis at trauma admission between January 2016 and April 2019. Labeled cases were separated into five combinations of training (80%) and test (20%) sets, and fivefold cross-validation was performed. Dice similarity coefficients (DSCs) were compared with those from a three-dimensional (3D) U-Net and a coarse-to-fine deep learning method. Areas under the receiver operating characteristic curve (AUCs) for a composite outcome, including hemostatic intervention, transfusion, and in-hospital mortality, were compared with consensus categorical assessment by two radiologists. An optimal cutoff was derived by using a radial basis function-based support vector machine.

RESULTS

Mean DSC for the multiscale algorithm was 0.61 ± 0.15 (standard deviation) compared with 0.32 ± 0.16 for the 3D U-Net method and 0.52 ± 0.17 for the coarse-to-fine method (P < .0001). Correlation and agreement between automated and manual volumes were excellent (Pearson r = 0.97, intraclass correlation coefficient = 0.93). The algorithm produced intuitive and explainable visual results. AUCs for automated volume measurement and categorical estimation were 0.86 and 0.77, respectively (P = .004). An optimal cutoff of 278.9 mL yielded accuracy of 84%, sensitivity of 82%, specificity of 93%, positive predictive value of 86%, and negative predictive value of 83%.

CONCLUSION

A multiscale deep learning method for traumatic hemoperitoneum quantitative visualization had improved diagnostic performance for predicting hemorrhage-control interventions and mortality compared with subjective volume estimation. Supplemental material is available for this article. © RSNA, 2020.