Interobserver reliability of the Tile classification system for pelvic fractures among radiologists and surgeons

Anterior subcutaneous internal fixator (INFIX) versus plate fixation for anterior ring injury in Tile C pelvic fractures: a retrospective study

OBJECTIVES

. The purpose of this study was to compare the reduction effect and clinical outcomes of anterior subcutaneous internal fixation (INFIX) and steel plate-screw internal fixation in the treatment of anterior ring injury in Tile C pelvic fractures.

METHODS

. In this retrospective study, the clinical outcomes of 46 patients treated using INFIX and 44 patients treated with steel plate-screw internal fixation were analyzed and compared. All patients underwent anterior and posterior fixation. The Matta imaging scoring system was used to evaluate the postoperative reduction accuracy; the Majeed scoring system was applied to obtain functional outcomes in clinical follow-up. All potential complications were identified and evaluated accordingly.

RESULTS

Both groups of patients were followed up for a period of 13-36 months, with an average of 27 months. The procedure time and blood loss in the INFIX group were significantly lower than those in the plate group(t = - 2.327, P = 0.023;t = - 4.053, P = 0.000; there was no statistically significant difference in the Majeed score and Matta score between the two groups after surgery (P > 0.05).

CONCLUSIONS

. INFIX treatment for anterior ring injury in Tile C pelvic fractures can achieve good therapeutic effects. Compared to internal fixation with plates and screws, it has advantages such as shorter surgical time and less blood loss. INFIX may be more suitable for obese patients, young women of childbearing age, or patients with urinary system injuries.

Interpretable Severity Scoring of Pelvic Trauma Through Automated Fracture Detection and Bayesian Inference

Pelvic ring disruptions result from blunt injury mechanisms and are potentially lethal mainly due to associated injuries and massive pelvic hemorrhage. The severity of pelvic fractures in trauma victims is frequently assessed by grading the fracture according to the Tile AO/OTA classification in whole-body Computed Tomography (CT) scans. Due to the high volume of whole-body CT scans generated in trauma centers, the overall information content of a single whole-body CT scan and low manual CT reading speed, an automatic approach to Tile classification would provide substantial value, e.g., to prioritize the reading sequence of the trauma radiologists or enable them to focus on other major injuries in multi-trauma patients. In such a high-stakes scenario, an automated method for Tile grading should ideally be transparent such that the symbolic information provided by the method follows the same logic a radiologist or orthopedic surgeon would use to determine the fracture grade. This paper introduces an automated yet interpretable pelvic trauma decision support system to assist radiologists in fracture detection and Tile grading. To achieve interpretability despite processing high-dimensional whole-body CT images, we design a neurosymbolic algorithm that operates similarly to human interpretation of CT scans. The algorithm first detects relevant pelvic fractures on CTs with high specificity using Faster-RCNN. To generate robust fracture detections and associated detection (un)certainties, we perform test-time augmentation of the CT scans to apply fracture detection several times in a self-ensembling approach. The fracture detections are interpreted using a structural causal model based on clinical best practices to infer an initial Tile grade. We apply a Bayesian causal model to recover likely co-occurring fractures that may have been rejected initially due to the highly specific operating point of the detector, resulting in an updated list of detected fractures and corresponding final Tile grade. Our method is transparent in that it provides fracture location and types, as well as information on important counterfactuals that would invalidate the system's recommendation. Our approach achieves an AUC of 0.89/0.74 for translational and rotational instability,which is comparable to radiologist performance. Despite being designed for human-machine teaming, our approach does not compromise on performance compared to previous black-box methods.

Dynamic changes in bleeding sites: Evaluating contrast extravasation on computed tomography and angiography in pelvic fractures

BACKGROUND

Contrast extravasation ("blush") on contrast-enhanced computed tomography (CECT) indicates active bleeding, but diagnostic angiography (AG) following CECT sometimes differs from the CECT findings in patients with pelvic fractures. This study aimed to evaluate the concordance rate between blush on CECT and AG based on pelvic anatomical location.

METHODS

This retrospective single-center study included patients (18 years or older) with pelvic fractures between 2015 and 2023 who had AG after CECT. Angiography was performed in patients with blush or hematoma on CECT, or unstable hemodynamics due to pelvic fracture. Pelvic bleeding regions were categorized separately into left and right as anterior internal iliac artery and posterior internal iliac artery based on pelvic arterial anatomy. The concordance rate between blush on CECT and AG was assessed using the k statistic.

RESULTS

A total of 87 patients with pelvic fractures were included. Among these, 75 (86%) had blush on CECT, and 12 (14%) had no blush on CECT. The concordance rate was 83% (95% confidence interval, 0.03-0.57; k = 0.30) when assessed on an individual patient basis and 53% (95% confidence interval, 0.24-0.44; k = 0.34) when evaluated by anatomical regions on a unilateral half of the pelvis. The concordance rate for the anatomical hemipelvis evaluation was significantly lower compared with the individual evaluation ( p < 0.001). Thirty-three percent of patients with unilateral blush on CECT showed new contralateral blushes during AG.

CONCLUSION

The location of active bleeding often differs between CECT and AG, indicating that bleeding may change intermittently and at various sites over time. This suggests the importance of evaluating both sides of the pelvic arteries during AG rather than solely relying on CECT findings.

LEVEL OF EVIDENCE

Prognostic and Epidemiological; Level IV.

Reliability of interobserver and intraobserver assessments of the Tile versus Young-Burgess classifications for pelvic ring injuries

Tile and Young-Burgess classification systems have been widely used to classify pelvic ring injuries in clinical settings; however, these two systems are rarely mentioned in the literature despite their reproducibility and practical application. The purpose of this study was to assess the interobserver and intraobserver reliability of the Tile and Young-Burgess classification of pelvic ring injuries. Eight hundred seven patients with pelvic fractures who presented to two I-level trauma centres between January 2016 and June 2020 were recruited for this study. The anteroposterior, inlet, and outlet of the pelvis were observed on radiographic images, and CT scans were performed on each patient. The anteroposterior, inlet, and outlet of the pelvis were radiographed at each visit, and CT scans were performed. Four observers were recruited to classify the injury type according to the Young-Burgess and Tile classification systems. Interobserver and intraobserver reliability (kappa coefficient, κ) were applied to assess the reproducibility of the two classification systems. A total of 174 patients (21.56%) could not be classified with the Young-Burgess classification, and all patients could be classified according to the Tile system. The mean κ value of the interobserver reliability was 0.558 (range, 0.444 to 0.653) for the Tile classification, indicating moderate agreement, whereas the mean κ value was 0.637 (range, 0.560 to 0.713), indicating substantial agreement, when the Young-Burgess classification was used. The mean κ value for intraobserver reliability was 0.597 (range, 0.567 to 0.653) for the Tile classification and 0.776 (range, 0.748 to 0.803) for the Young-Burgess classification. More pelvic ring injuries could be classified with the Tile classification than could be classified with the Young-Burgess classification. However, the reproducibility of the Young-Burgess classification was greater than that of the Tile classification.

2023 EULAR recommendations on imaging in diagnosis and management of crystal-induced arthropathies in clinical practice

OBJECTIVE

To formulate evidence-based recommendations and overarching principles on the use of imaging in the clinical management of crystal-induced arthropathies (CiAs).

METHODS

An international task force of 25 rheumatologists, radiologists, methodologists, healthcare professionals and patient research partners from 11 countries was formed according to the EULAR standard operating procedures. Fourteen key questions on the role of imaging in the most common forms of CiA were generated. The CiA assessed included gout, calcium pyrophosphate deposition disease and basic calcium phosphate deposition disease. Imaging modalities included conventional radiography, ultrasound, CT and MRI. Experts applied research evidence obtained from four systematic literature reviews using MEDLINE, EMBASE and CENTRAL. Task force members provided level of agreement (LoA) anonymously by using a Numerical Rating Scale from 0 to 10.

RESULTS

Five overarching principles and 10 recommendations were developed encompassing the role of imaging in various aspects of patient management: making a diagnosis of CiA, monitoring inflammation and damage, predicting outcome, response to treatment, guided interventions and patient education. Overall, the LoA for the recommendations was high (8.46-9.92).

CONCLUSIONS

These are the first recommendations that encompass the major forms of CiA and guide the use of common imaging modalities in this disease group in clinical practice.

Performance of three predictive scores to avoid delayed diagnosis of significant blunt bowel and mesenteric injury: A 12-year retrospective cohort study

BACKGROUND

Avoiding missed diagnosis and therapeutic delay for significant blunt bowel and mesenteric injuries (sBBMIs) after trauma is still challenging despite the widespread use of computed tomography (CT). Several scoring tools aiming at reducing this risk have been published. The purpose of the present work was to assess the incidence of delayed (>24 hours) diagnosis for sBBMI patients and to compare the predictive performance of three previously published scores using clinical, radiological, and laboratory findings: the Bowel Injury Prediction Score (BIPS) and the scores developed by Raharimanantsoa Score (RS) and by Faget Score (FS).

METHODS

A population-based retrospective observational cohort study was conducted; it included adult trauma patients after road traffic crashes admitted to Lausanne University Hospital, Switzerland, between 2008 and 2019 (n = 1,258) with reliable information about sBBMI status (n = 1,164) and for whom all items for score calculation were available (n = 917). The three scores were retrospectively applied on all patients to assess their predictive performance.

RESULTS

The incidence of sBBMI after road traffic crash was 3.3% (38 of 1,164), and in 18% (7 of 38), there was a diagnostic and treatment delay of more than 24 hours. The diagnostic performances of the FS, the RS, and the BIPS to predict sBBMI, expressed as the area under the receiver operating characteristic curve, were 95.3% (95% confidence interval [CI], 92.7-97.9%), 89.2% (95% CI, 83.2-95.3%), and 87.6% (95% CI, 81.8-93.3%) respectively.

CONCLUSION

The present study confirms that diagnostic delays for sBBMI still occur despite the widespread use of abdominal CT. When CT findings during the initial assessment are negative or equivocal for sBBMI, using a score may be helpful to select patients for early diagnostic laparoscopy. The FS had the best individual diagnostic performance. However, the BIPS or the RS, relying on clinical and laboratory variables, may be helpful to select patients for early diagnostic laparoscopy when there are unspecific CT signs of bowel or mesenteric injury.

LEVEL OF EVIDENCE

Prognostic and Epidemiological; Level III.

Earthquakes from a radiological perspective: what is demanded from the radiologists, and what can we do? A pictorial review

Earthquakes are among the most destructive and unpredictable natural disasters. Various diseases and ailments, such as bone fractures, organ and soft-tissue injuries, cardiovascular diseases, lung diseases, and infectious diseases, can develop in the aftermath of severe earthquakes. Digital radiography, ultrasound, computed tomography, and magnetic resonance imaging are significant imaging modalities utilized for the quick and reliable assessment of earthquake-related ailments to facilitate the planning of suitable therapy. This article examines the usual radiological imaging characteristics observed in individuals from quake-damaged regions and summarizes the strengths and functionality of imaging modalities. In such circumstances, where quick decision-making processes are life-saving and essential, we hope this review will be a practical reference for readers.

Prevalence and Predictors of Long-term Opioid Use After Pelvic Fractures

BACKGROUND

Opioids are effective in short-term pain treatment; however, their long-term effectiveness is unconfirmed. Many patients are exposed to opioids after pelvic injuries with little known about persistent use afterward. We assessed the prevalence and predictors of long-term opioid use following pelvic fractures.

MATERIALS AND METHODS

This retrospective study enrolled 277 patients with acute pelvic fractures over five years. Daily and total morphine milligram equivalents (MME) were calculated. The primary outcome was long-term opioid use (LOU) defined as ongoing opioid use 60-90 days post-discharge. The secondary outcome was intermediate-term opioid use (IOU) defined as ongoing opioid use 30-60 days post-discharge. Univariable and logistic regression analyses were performed.

RESULTS

Median (interquartile range) total inpatient opioid MME was 422 (157-1667) with a median daily MME of 69 (26-145). Long-term opioid use occurred in 16%, and IOU occurred in 29%. Univariable analysis found that total and daily inpatient opioid use were each significantly associated with LOU (median MME, 1241 vs 371; median MMEs, 127.7 vs 59.2, respectively) and IOU (median MME, 1140 vs 326; median MMEs, 111.8 vs 57.9, respectively). Logistic regression analysis found daily inpatient MME ≥50 (odds ratio [OR] 3.027, 95% confidence interval [CI] 1.059-8.652]) and pelvic fracture type (Tile B/C) (OR 2.992 [CI 1.324-6.763])were independent predictors of LOU.

CONCLUSION

Total and daily inpatient opioid use were significantly associated with LOU and IOU. Patients who received ≥50 MME/inpatient day had a higher likelihood of LOU. This study seeks to inform clinical decisions for pain management to prevent adverse outcomes.

Performance Assessment of Out-of-Hospital Use of Pelvic Circumferential Compression Devices for Severely Injured Patients in Switzerland: A Nationwide Retrospective Cross-Sectional Study

BACKGROUND

Patients with severe pelvic fractures carry a greater risk of severe bleeding, and pelvic compression devices (PCCD) are used to stabilize the pelvis on the pre-hospital scene. The aim of this study was to describe the use of PCCD in the pre-hospital setting on a nationwide scale (Switzerland) and determine the sensitivity, specificity and rates of over- and under-triage of the current application practices. The secondary objective was to identify pre-hospital factors associated with unstable pelvic fractures.

METHODS

Retrospective cross-sectional study using anonymized patient data (1 January 2015-31 December 2020) from the Swiss Trauma Registry (STR). Based on AIS scores, patients were assigned a unique principal diagnosis among three categories (unstable pelvic fracture-stable pelvic fracture-other) and assessed for use or not of PCCD. Secondarily, patient characteristics, initial pre-hospital vital signs, means of pre-hospital transport and trauma mechanism were also extracted from the database.

RESULTS

2790 patients were included for analysis. A PCCD was used in 387 (13.9%) patients. In the PCCD group, 176 (45.5%) had an unstable pelvic fracture, 52 (13.4%) a stable pelvic fracture and 159 (41.1%) an injury unrelated to the pelvic region. In the group who did not receive a PCCD, 214 (8.9%) had an unstable pelvic fracture, 182 (7.6%) a stable pelvic fracture and 2007 (83.5%) an injury unrelated to the pelvic region. The nationwide sensitivity of PCCD application was 45.1% (95% CI 40.1-50.2), the specificity 91.2% (95% CI 90-92.3), with both over- and under-triage rates of 55%. The prevalence of unstable fractures in our population was 14% (390/2790). We identified female sex, younger age, lower systolic blood pressure, higher shock index, pedestrian hit and fall ≥3 m as possible risk factors for an unstable pelvic fracture.

CONCLUSIONS

Our results demonstrate a nationwide both over- and under-triage rate of 55% for out-of-hospital PCCD application. Female gender, younger age, lower blood pressure, higher shock index, pedestrian hit and fall >3 m are possible risk factors for unstable pelvic fracture, but it remains unclear if those parameters are relevant clinically to perform pre-hospital triage.

Correlation between traumatic pelvic ring injuries and sexual dysfunctions: a multicentric retrospective study

PURPOSE

Among the functional impairments associated with pelvic ring injuries (PRI), sexual dysfunction (SD) is a common clinical issue. The aim of this study is to investigate correlations between traumatic PRI, genitourinary, and sexual dysfunctions, for a proper multidisciplinary treatment.

METHODS

We performed an observational, multicentric study, from January 2020 to 2022. We conducted a follow-up after surgery at three, six, 12, and 24 months by measuring the Female Sexual Functioning Index (FSFI), the International Index of Erectile Function (IIEF), the Arizona Sexual Experience Scale (ASEX), the Majeed Score, and the SF-12. Descriptive statistics was conducted on T-test, Whelc's test, and one-way ANOVA which were performed when appropriate.

RESULTS

A total of 76 patients (mean age 42.17 ± 15 years) were included in the study and allocated into three groups (A, B, and C). Tile A group revealed good sexual outcomes, similar to that of healthy patients. Tile B group demonstrated worsen SD than the previous group. In Tile C group, there was a longer average duration of the orthopaedic surgery when compared to group B. However, in terms of SDs, statistical significance could not be demonstrated between groups C and B.

CONCLUSIONS

We observed a progressive spontaneous recovery of sexual function, corresponding to each PRI group. Moreover, men classified as B2 had milder SDs than B1 male patients.

Concordance of neonatal critical condition data between secondary databases: Florida and Texas birth certificate Linkage with medicaid analytic extract

BACKGROUND

Limited information is available about neonates' critical conditions data quality. The study aim was to measure the agreement regarding presence of neonatal critical conditions between Medicaid Analytic eXtract claims data and Birth Certificate (BC) records.

METHODS

Claims data files of neonates born between 1999-2010 and their mothers were linked to birth certificates in the states of Texas and Florida. In claims data, neonatal critical conditions were identified using medical encounter claims records within the first 30 days postpartum, while in birth certificates, the conditions were identified based on predetermined variables. We calculated the prevalence of cases within each data source that were identified by its comparator, in addition to calculating overall agreement and kappa statistics.

RESULTS

The sample included 558,224 and 981,120 neonates in Florida and Texas, respectively. Kappa values show poor agreement (< 20%) for all critical conditions except neonatal intensive care unit (NICU) admission, which showed moderate (> 50%) and substantial (> 60%) agreement in Florida and Texas, respectively. claims data resulted in higher prevalences and capture of a larger proportion of cases than the BC, except for assisted ventilation.

CONCLUSIONS

Claims data and BC showed low agreement on neonatal critical conditions except for NICU admission. Each data source identified cases most of which the comparator failed to capture, with higher prevalences estimated within claims data except for assisted ventilation.

Histological, radiological and clinical analysis of the supraspinatus tendon and muscle in rotator cuff tears

BACKGROUND

Macroscopic alterations of the affected rotator cuff (RC) are undoubtedly linked to microscopic changes, but they may underestimate the actual degree of the disease. Moreover, it remains unclear whether preoperative structural RC changes may alter clinical outcomes.

METHODS

Supraspinatus tendon and muscle samples were collected from 47 patients undergoing RC surgery. Tendons were evaluated histologically according to the Bonar score; fatty infiltration and muscle atrophy were quantified using a software for biomedical image analysis (ImageJ) in percentage of area affected in the observed muscle section. Preoperative shoulder ROM and pain were evaluated. Radiological muscle atrophy was evaluated with the Tangent Sign and Occupation Ratio; fatty infiltration was assessed according to the Goutallier classification. Correlations between histological, radiological and clinical outcomes were assessed. Statistics were performed using the Spearman correlation coefficient. Intraobserver and interobserver agreement was calculated.

RESULTS

Histopathologic fatty infiltration (r = 0.007, p = 0.962), muscle atrophy (r = 0.003, p = 0.984) and the total Bonar score (r = 0.157, p = 0.292) were not correlated to preoperative shoulder pain. Muscle atrophy showed a significant but weak negative correlation with the preoperative movement of abduction (r = -0.344, p = 0.018). A significant but weak positive correlation was found between muscle atrophy and the total Bonar score (r = 0.352, p = 0.015). No correlation between histological and radiological evaluation was found for both fatty infiltration (r = 0.099, p = 0.510) and muscle atrophy (Tangent Sign: r = -0.223, p = 0.131; Occupation Ratio: r = -0.148, p = 0.319). Our histological evaluation showed a modal value of 3 (out of 3) for fatty infiltration and an equal modal value of 2 and 3 (out of 3) for muscle atrophy. In contrast, the modal value of the Goutallier score was 1 (out of 4) and 28 patients out of 47 showed a negative Tangent sign. At histology, intraobserver agreement ranged from 0.59 to 0.81 and interobserver agreement from 0.57 to 0.64. On the MRI intraobserver agreement ranged from 0.57 to 0.71 and interobserver agreement ranged from 0.53 to 0.65.

CONCLUSIONS

Microscopic muscle atrophy appeared to negatively correlate with the movement of abduction leading to functional impairment. Shoulder pain did not show any relationship with microscopic changes. Radiological evaluation of the supraspinatus muscle alterations seemed to underestimate the degree of the same abnormalities evaluated at histology.

A retrospective analysis of minimally invasive internal fixation versus nonoperative conservative management of pelvic ring fragility fractures and the elderly

OBJECTIVE

We aimed to investigate the adoption of treatment patterns for hip fractures combined with minimally invasive surgical treatment of fragility fractures of the pelvis in older individuals and reviewed and analysed the treatment efficacy and feasibility.

METHODS AND DATA

From September 2017 to February 2021, 135 older individuals with fragility fractures of the pelvis were admitted to our hospital. We retrospectively analysed patients who received surgical or conservative treatments. The general preoperative data, including sex, age, disease duration, cause of injury, AO/OTA type, BMI, bone mineral density, time from injury to admission, time from injury to surgery, ASA classification, number of underlying diseases, mean bed rest, clinical fracture healing, VAS score and Majeed functional score, were recorded.

RESULTS

The mean follow-up time for all 135 patients was 10.5 ± 3.6 months. Among 135 patients, 95 survived, and 11 and 29 patients passed after the surgical (mortality rate = 17.74%) and conservative (mortality rate = 39.73%) treatments, respectively. The average follow-up time for the 95 surviving patients was 14.5 ± 1.8 months. The Majeed and VAS scores for the operation group were significantly better than those of the conservative group. The bed rest and fracture healing times were also shorter in the surgical treatment group than in the conservative group.

CONCLUSION

The use of a minimally invasive surgical treatment combined with the geriatric hip fracture treatment model to treat fragility fractures of the pelvis improved the quality of life in older patients.

Evolution of the AO Spine Sacral and Pelvic Classification System: a systematic review

OBJECTIVE

The purpose of this study was to describe the genesis of the AO Spine Sacral and Pelvic Classification System in the context of historical sacral and pelvic grading systems.

METHODS

A systematic search of MEDLINE, EMBASE, Google Scholar, and Cochrane databases was performed consistent with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to identify all existing sacral and pelvic fracture classification systems.

RESULTS

A total of 49 articles were included in this review, comprising 23 pelvic classification systems and 17 sacral grading schemes. The AO Spine Sacral and Pelvic Classification System represents both the evolutionary product of these historical systems and a reinvention of classic concepts in 5 ways. First, the classification introduces fracture types in a graduated order of biomechanical stability while also taking into consideration the neurological status of patients. Second, the traditional belief that Denis central zone III fractures have the highest rate of neurological deficit is not supported because this subgroup often includes a broad spectrum of injuries ranging from a benign sagittally oriented undisplaced fracture to an unstable "U-type" fracture. Third, the 1990 Isler lumbosacral system is adopted in its original format to divide injuries based on their likelihood of affecting posterior pelvic or spinopelvic stability. Fourth, new discrete fracture subtypes are introduced and the importance of bilateral injuries is acknowledged. Last, this is the first integrated sacral and pelvic classification to date.

CONCLUSIONS

The AO Spine Sacral and Pelvic Classification is a universally applicable system that redefines and reorders historical fracture morphologies into a rational hierarchy. This is the first classification to simultaneously address the biomechanical stability of the posterior pelvic complex and spinopelvic stability, while also taking into consideration neurological status. Further high-quality controlled trials are required prior to the inclusion of this novel classification within a validated scoring system to guide the management of sacral and pelvic injuries.

Design and evaluation of an intelligent reduction robot system for the minimally invasive reduction in pelvic fractures

Introduction

Pelvic fracture is a severe high-energy injury with the highest disability and mortality of all fractures. Traditional open surgery is associated with extensive soft tissue damages and many complications. Minimally invasive surgery potentially mitigates the risks of open surgical procedures and is becoming a new standard for pelvic fracture treatment. The accurate reduction has been recognized as the cornerstone of minimally invasive surgery for pelvic fracture. At present, the closed reduction in pelvic fractures is limited by the current sub-optimal 2D intra-operative imaging (fluoroscopy) and by the high forces of soft tissue involved in the fragment manipulation, which might result in fracture malreduction. To overcome these shortcomings and facilitate pelvic fracture reduction, we developed an intelligent robot-assisted fracture reduction (RAFR) system for pelvic fracture.

Methods

The presented method is divided into three parts. The first part is the preparation of 20 pelvic fracture models. In the second part, we offer an automatic reduction algorithm of our robotic reduction system, including Intraoperative real-time 3D navigation, reduction path planning, control and fixation, and robotic-assisted fracture reduction. In the third part, image registration accuracy and fracture reduction accuracy were calculated and analyzed.

Results

All 20 pelvic fracture bone models were reduced by the RAFR system; the mean registration error E1 of the 20 models was 1.29 ± 0.57 mm. The mean reduction error E2 of the 20 models was 2.72 ± 0.82 mm. The global error analysis of registration and reduction results showed that higher errors are mainly located at the edge of the pelvis, such as the iliac wing.

Conclusion

The accuracy of image registration error and fracture reduction error in our study was excellent, which could reach the requirements of the clinical environment. Our study demonstrated the precision and effectiveness of our RAFR system and its applicability and usability in clinical practice, thus paving the way toward robot minimally invasive pelvic fracture surgeries.

An Interpretable Approach to Automated Severity Scoring in Pelvic Trauma

Pelvic ring disruptions result from blunt injury mechanisms and are often found in patients with multi-system trauma. To grade pelvic fracture severity in trauma victims based on whole-body CT, the Tile AO/OTA classification is frequently used. Due to the high volume of whole-body trauma CTs generated in busy trauma centers, an automated approach to Tile classification would provide substantial value, e. g., to prioritize the reading queue of the attending trauma radiologist. In such scenario, an automated method should perform grading based on a transparent process and based on interpretable features to enable interaction with human readers and lower their workload by offering insights from a first automated read of the scan. This paper introduces an automated yet interpretable pelvic trauma decision support system to assist radiologists in fracture detection and Tile grade classification. The method operates similarly to human interpretation of CT scans and first detects distinct pelvic fractures on CT with high specificity using a Faster-RCNN model that are then interpreted using a structural causal model based on clinical best practices to infer an initial Tile grade. The Bayesian causal model and finally, the object detector are then queried for likely co-occurring fractures that may have been rejected initially due to the highly specific operating point of the detector, resulting in an updated list of detected fractures and corresponding final Tile grade. Our method is transparent in that it provides finding location and type using the object detector, as well as information on important counterfactuals that would invalidate the system's recommendation and achieves an AUC of 83.3%/85.1% for translational/rotational instability. Despite being designed for human-machine teaming, our approach does not compromise on performance compared to previous black-box approaches.

An Automated Deep Learning Method for Tile AO/OTA Pelvic Fracture Severity Grading from Trauma whole-Body CT

Admission trauma whole-body CT is routinely employed as a first-line diagnostic tool for characterizing pelvic fracture severity. Tile AO/OTA grade based on the presence or absence of rotational and translational instability corresponds with need for interventions including massive transfusion and angioembolization. An automated method could be highly beneficial for point of care triage in this critical time-sensitive setting. A dataset of 373 trauma whole-body CTs collected from two busy level 1 trauma centers with consensus Tile AO/OTA grading by three trauma radiologists was used to train and test a triplanar parallel concatenated network incorporating orthogonal full-thickness multiplanar reformat (MPR) views as input with a ResNeXt-50 backbone. Input pelvic images were first derived using an automated registration and cropping technique. Performance of the network for classification of rotational and translational instability was compared with that of (1) an analogous triplanar architecture incorporating an LSTM RNN network, (2) a previously described 3D autoencoder-based method, and (3) grading by a fourth independent blinded radiologist with trauma expertise. Confusion matrix results were derived, anchored to peak Matthews correlation coefficient (MCC). Associations with clinical outcomes were determined using Fisher's exact test. The triplanar parallel concatenated method had the highest accuracies for discriminating translational and rotational instability (85% and 74%, respectively), with specificity, recall, and F1 score of 93.4%, 56.5%, and 0.63 for translational instability and 71.7%, 75.7%, and 0.77 for rotational instability. Accuracy of this method was equivalent to the single radiologist read for rotational instability (74.0% versus 76.7%, p = 0.40), but significantly higher for translational instability (85.0% versus 75.1, p = 0.0007). Mean inference time was < 0.1 s per test image. Translational instability determined with this method was associated with need for angioembolization and massive transfusion (p = 0.002-0.008). Saliency maps demonstrated that the network focused on the sacroiliac complex and pubic symphysis, in keeping with the AO/OTA grading paradigm. A multiview concatenated deep network leveraging 3D information from orthogonal thick-MPR images predicted rotationally and translationally unstable pelvic fractures with accuracy comparable to an independent reader with trauma radiology expertise. Model output demonstrated significant association with key clinical outcomes.