Research Presentation Session

RPS 1117 - Abdomen and brain

Lectures

1
RPS 1117 - Adrenal glands enhancement in computed tomography as a predictor of 24-hour mortality in critically ill patients

RPS 1117 - Adrenal glands enhancement in computed tomography as a predictor of 24-hour mortality in critically ill patients

05:49R. Winzer, Dresden / DE

Purpose:

To assess the hyperattenuation of adrenal glands on contrast-enhanced computed tomography scans as a predictor of hospital mortality in critical care unit patients.

Methods and materials:

Seventy-three patients (67.5 ±12.3 years of age) were included in this retrospective analysis. All patients underwent contrast-enhanced CT due to life-threatening indications. Attenuation of the adrenal glands in the portal venous phase was ROI-based assessed quantitatively by documenting Hounsfield units (HU) by two radiologists. From ROC analysis, Youden’s J statistic was derived with death within 24 (48 and 72) hours as a classifier. Interrater agreement was assessed using intraclass correlation coefficient for measurements of the HU in adrenal glands and weighted kappa (k) after classifying patients with hyperdense adrenal enhancement (= enhancement above the threshold by ROC analysis).

Results:

Fourteen patients (19%) died within 24 hours. In the portal venous phase HU means of both adrenal glands differed significantly between patients who died within 24 hours and those who survived (135.7 ± 46 HU vs 74.7 ± 24 HU; p=0.001). For portal venous phase, data ROI-based ROC analysis yielded for a Youden index of 0.79/0.85, a sensitivity of 100%/100% and a specificity of 79%/85% (p<.0001), a cut-off point of 93/95 HU for prediction of 24-hour mortality for rater 1 and 2, respectively. The AUC was 0.92/0.93. ICC of 0.99 and weighted kappa (k) of 0.84 indicated a high interrater agreement.

Conclusion:

Venous hyperattenuation of adrenal glands on contrast-enhanced CT can predict 24-hour mortality quite well and can serve as a reproducible prognostic marker for the patient's outcome.

Limitations:

Retrospective study.

Ethics committee approval

Ethics committee approval is given.

Funding:

No funding was received for this work.

2
RPS 1117 - Do emergency physicians appropriately request a head CT in patients with new-onset seizure?

RPS 1117 - Do emergency physicians appropriately request a head CT in patients with new-onset seizure?

05:27I. García Tuells, Murcia / ES

Purpose:

To determine how Emergency Head Computed Tomography (EHCT) is requested in patients with new-onset seizure in our environment. EHCT would be appropriate if the new-onset seizure was focal, associated with neurological deficit or in certain contexts.

Methods and materials:

Physicians from 3 hospitals (H1, H2, H3), regularly involved in emergency on-call duty, answered a five-question survey on EHCT appropriateness for new-onset seizures. Fisher's test was used to compare the answers (A) among hospitals, resident/staff physician, residence year and medical specialty.

Results:

We obtained 39 surveys [H1 12 (20.77%), H2 18 (46.15%), H3 6 (23.08%)] from 31 residents (79.49%) [14 (45.16%) ≥3 years of training] and 8 staff physicians (20.51%). 31/39 (79.49%) were Family and Community Medicine Specialists (FCMS). 32/39 (82%) would request (A4+A5) EHCT whatever the kind of seizure (Q1). That was the most common response in FCMS (27/31 -87.1%- vs 5/8 -62%-, P=0.043) and in staff physicians (8/8 -100%- vs 24/31 -77.4%- P=0.042). 9/39 (23.07%) and 8/39 (20.51%) requested an EHCT (A4+A5) according to Radiology and Emergency Departments guidelines (Q4) or Emergency Department guidelines (Q5). Residents ≤2nd year (15/17 -88.23 - vs 6/14 -42.86%- P = 0.029) were not aware (A2+A3) of those guidelines (Q5). For 31/39 (79.45%) of the participants, the EHCT results did not modify (A2+A3) the clinical attitude (Q2) and 27/39 (69.20%) were not sure (A3) about EHCT detected relevant alterations in those patients (Q3). The answers were similar among hospitals.

Conclusion:

On-call physicians routinely request CT scans for new-onset seizure ignoring current evidence. It is mandatory to go into quality improvement cycles promoted by radiology departments to improve current practices.

Limitations:

Limitations were a potential Hawthorne effect and the small sample size.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

3
RPS 1117 - The performance of artificial intelligence in the detection of intracranial haemorrhage on head computed tomography with clinical workflow integration

RPS 1117 - The performance of artificial intelligence in the detection of intracranial haemorrhage on head computed tomography with clinical workflow integration

05:53N. Watte, Laken / BE

Purpose:

To evaluate the performance of an artificial intelligence (AI) tool using a deep learning algorithm for detecting intracranial haemorrhage (ICH) on head computed tomographic (CT) exams with clinical workflow integration at the emergency department.

Methods and materials:

We retrospectively collected a dataset at the emergency department, containing 500 consecutive head CT exams and their clinical reports, between Sept 1, 2019, and Oct 1, 2019. All CT exams were automatically pseudo-anonymised and transferred for ICH detection by an AI tool (Aidoc, Israel). We registered the number of studies that were successfully processed by AI, and we calculated the diagnostic performance by the positive (PPV) and negative (NPV) predictive values, sensitivity, and specificity. The original clinical radiology report and consensus by independent supervision were considered as the gold standard. Concordance was assessed by Cohen’s k statistic.

Results:

During a one-month implementation phase, the algorithm created a report for 388 studies (77,6%) during real-time radiology workflow; 22,4% failed, due to multiple possible technical issues, not further subject to this study. From the 388 processed studies, 31 (7,9%), were labelled by AI as having ICH. The expert readers detected 37 (9,5%) ICHs. The concordance between AI and expert reading was substantial (kappa=0,65). Algorithm performance for ICH revealed a 98% NPV and 61% PPV. The false-negative and false-positive rates were 6/337 (1,7%) and 20/37 (54%), respectively. The sensitivity and specificity were 84%.

Conclusion:

In a clinical emergency setting, 77% of all head CT exams could be automatically evaluated by AI. With high specificity and negative predictive value, the AI tool shows the potential to rule out ICH. The positive predictive value remains moderate.

Limitations:

We currently investigated 500 cases. Further research with 3000 CT exams is ongoing.

Ethics committee approval

n/a

Funding:

n/a

4
RPS 1117 - Diagnostic accuracy of multidetector CT in colonic ischaemia in the emergency department

RPS 1117 - Diagnostic accuracy of multidetector CT in colonic ischaemia in the emergency department

06:19G. Addeo, Firenze / IT

Purpose:

To evaluate the diagnostic accuracy of MDCT (multidetector computed tomography), performed in an acute care setting in the diagnosis of primary CI (colonic ischaemia), with laparoscopic or laparotomic evaluation or surgery as the reference standard.

Methods and materials:

The abdominal MDCT scans of 277 patients with clinical-laboratory suspicion of CI were retrospectively reviewed, 46 patients were excluded for secondary CI. The final study population consisted of 231 patients. CT was classified as positive for CI when the radiologist's interpretation showed a definitive diagnosis of CI; clear absence of CI was classified as a negative result. If the radiologist missed the relevant findings and failed to suspect CI, the CT report was defined as incorrect. The MDCT findings were true positive, true negative, false positive, and false negative considering laparoscopic or laparotomic evaluation or endoscopy as the reference standard.

Results:

In 41% of 231 patients included, CI was pathologically confirmed on the use of surgery. On the basis of the emergency radiology reports there were 30% true positive cases, 3,5% false positive, 11% false negative and 56% true negative. We calculated sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy of MDCT obtaining: 73,4% sensitivity; 94% specificity; a PPV and a NPV of 90% and of 84%, respectively; 86% diagnostic accuracy.

Conclusion:

Despite the successful CT imaging results in CI diagnosis, high colonic distention was the major cause leading to more false negative results even after intravenous administration of the contrast agent. Radiologists' experience and expertise have an important impact on their performance.

Limitations:

Our investigation is in itself limited by its retrospective nature.

Ethics committee approval

The Hospital Institutional Review Board approved the study.

Funding:

No funding was received for this work.

5
RPS 1117 - Accuracy of Single-Pass Split-Bolus CT for detecting vascular injury in the spleen: a retrospective study in 111 patients with blunt splenic trauma

RPS 1117 - Accuracy of Single-Pass Split-Bolus CT for detecting vascular injury in the spleen: a retrospective study in 111 patients with blunt splenic trauma

06:11F. Berger, Toronto / CA

Purpose:

To evaluate the diagnostic performance of Single-Pass Split-Bolus CT (SPSB-CT) for the detection of all splenic vascular injuries (SVI) and clinically relevant splenic vascular injuries in blunt trauma patients with proven splenic injury.

Methods and materials:

Six-year single Level 1 trauma centre retrospective analysis of all blunt trauma patients with splenic injury on SPSB-CT. Patients were identified through electronic records. Data at the presentation and during 3 months follow-up were used. Two radiologists and one trauma surgeon, using all clinical and imaging data, defined the reference standard for the presence of SVI (AV fistula, pseudoaneurysms, active contrast extravasation) and clinically relevant SVI causing management changes or death. Four blinded radiologists evaluated initial SPSB-CT for SVI with binary (yes/no) and confidence (1-5) scores. Pooled sensitivity and specificity and observer agreement (Fleiss Kappa) statistics were calculated with 95% confidence intervals.

Results:

111 patients were included (male: 76%, mean age: 35, range 9-81) with 108 thoracoabdominal SPSB-CT (median DLP 455 mGycm), and 3 abdominopelvic SPSB-CT (mean DLP 243 mGycm). Mean aortic and hepatic enhancement was 310 HU and 109 HU. Incidence of SVI was 37/111 (33.3%), incidence of clinically relevant SVI was 27/111 (24.3%). For SVI, sensitivity was 85.8% (range 74.8% - 89.2%) and specificity was 89.9% (range 87.8% - 91.9%). For clinically relevant SVI, sensitivity was 88.9% (range 85.2% - 92.6%) and specificity was 81.9% (range 79.8% - 83.3%). Observer agreement was 88.1% with a Fleiss kappa of 0.76. At a 3 month follow-up, two patients had delayed splenic haemorrhage. No deaths related to missed SVI by SPSB-CT.

Conclusion:

In blunt trauma splenic injury patients, SPSB-CT adequately detects splenic vascular injury with good inter-observer agreement.

Limitations:

Retrospective, single-centre study.

Ethics committee approval

Ethics approval and informed consents was obtained.

Funding:

No funding was received for this work.

6
RPS 1117 - Variability in non-contrast head CT image quality in Ireland: opportunities for tarameter optimisation and standardisation to improve stroke care

RPS 1117 - Variability in non-contrast head CT image quality in Ireland: opportunities for tarameter optimisation and standardisation to improve stroke care

05:46J. Hynes, Dublin / IE

Purpose:

Non-contrast computed tomography (NCCT) of the brain is a critical tool in the investigation of suspected acute ischaemic stroke. The recognition of subtle differences in attenuation values between normal brain parenchyma and regions of ischaemic tissue is crucial to accurate and timely diagnosis. Our study aims to identify which combination of CT acquisition parameters maximises the ability to detect these early changes. We also investigated the variation in imaging protocols for suspected acute stroke at different institutions in the regional hospital network.

Methods and materials:

Uniformity images of the CATPHAN phantom were obtained under various combinations of kV, mAs, reconstruction kernel, slice thickness and strength of iterative reconstruction. Images were analysed by an in-house algorithm, based on the statistical method of Chao et. al., (2000) to objectively measure low contrast detectability. Data is presented as contrast-detail diagrams, i.e., the minimum separation in Hounsfield Units (HU) required to distinguish a low contrast object of a given equivalent diameter from the background. 12 CT scanners made by 3 different manufacturers across 6 hospitals were surveyed.

Results:

Increasing mAs, slice thickness and utilising higher levels of iterative reconstruction objectively improve low contrast detectability in NCCT brain. Deviations in protocols across manufacturers and hospitals were found to result in variations in image quality.

Conclusion:

Variations in acquisition protocol produce differences in image quality which may contribute to missed and/or delayed findings. Optimising low contrast detectability and ensuring standardisation between linked institutions offers the potential to make a meaningful improvement in modern stroke care.

Limitations:

Standardisation will inevitably be limited by the variation in manufacturers, radiologists’ preferences and scanner performance.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

7
RPS 1117 - Identification of lodged projectiles on CT based on their shape, caliber measurements and dual-energy material differentiation

RPS 1117 - Identification of lodged projectiles on CT based on their shape, caliber measurements and dual-energy material differentiation

05:46D. Gascho, Zurich / CH

Purpose:

The purpose of this study was to assess the visual classification of bullets and the accuracy of calibre estimations on computed tomograpy, as well as the feasibility of dual-energy-based material differentiation.

Methods and materials:

This postmortem study includes fatal gunshot wounds (n=10) with a projectile or bullet fragment lodged inside the body. The decedents underwent CT with 120 kVp. Two additional scans with tube voltages of 120 and 140 kVp (CT dose index: both 9 mGy) were performed over the anatomic region where the projectile was lodged. The shape of the projectile was assessed by an expert in ballistics using volume renderings. The calibres were estimated by measuring the diameter of the projectile on aligned multi-planar reconstructions. Two radiologists (experienced / little experienced in ballistics) performed the measurements. For each projectile, the dual-energy index (DEI) was calculated. The DEI was based on CT numbers obtained from ROI measurements at 120 and 140 kVp (Fig.1C&D).

Results:

Non-fragmented bullets could be identified as semi-jacketed hollow-point (SJHP) bullets or full metal jacket (FMJ) bullets. The SJHP bullets demonstrated characteristic mushrooming on CT. The experienced radiologist was close to the actual calibre according to the measurements of the diameters of non-fragmented bullets. All lead bullets, whether fragmented or not, clearly differed from a bullet made of copper/zinc in their DEI.

Conclusion:

CT can provide information regarding the type of ammunition used, which is crucial for the assessment of injuries. However, visual assessments and calibre estimations are only feasible for non-fragmented projectiles. The DEI-based material differentiation is not limited to non-fragmented projectiles and allowed a clear distinction between lead and copper/zinc bullets.

Limitations:

This study is limited by the small number of cases.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

8
RPS 1117 - Thin slices and maximum intensity projection reconstructions increase sensitivity to hyperdense artery sign in acute ischemic stroke

RPS 1117 - Thin slices and maximum intensity projection reconstructions increase sensitivity to hyperdense artery sign in acute ischemic stroke

06:13J. Rosskopf, Ulm / DE

Purpose:

To investigate the effect of different cranial nonenhanced CT (NECT) image reconstructions on correct detection of hyperdense artery sign (HAS) which is a frequently rising challenge particularly also for observers with lower practice level on NECT.

Methods and materials:

Twenty-five of 100 patients’ NECT image data presented with HAS. Sixteen observers with lower practice level on NECT evaluated independently and in a randomised order the three image reconstructions of each data set with thin slice 0.75 mm, thick slab 5 mm and 5 mm maximum intensity projection (MIP) and rated the likelihood of the presence of HAS in the middle cerebral artery.

Results:

MIP and thin slice image reconstructions yielded significantly higher sensitivities for the correct detection of HAS than thick slab reconstructions (73%/73% vs 45%; p < 0.05). Sensitivity to HAS did not differ between using thin slice and MIP reconstructions (73% vs 73%). The interobserver reliability was moderate (κ, 0.4). Ranking of HAS detectability did not correlate with image reconstructions (p > 0.05 of ICC in a mixed model analysis).

Conclusion:

MIP and thin slices reconstructions increased the sensitivity to HAS (73%) whereas thick slab reconstructions seemed to be rather less appropriate (45%). These findings might be important also in particular for observers with lower practice level on NECT.

Limitations:

The observers in our study might have been biased in detecting clots by the appearance of early signs of infarction.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

9
RPS 1117 - Diffusion-weighted imaging can add value to emergency MRI for the diagnosis of acute appendicitis among pregnant women

RPS 1117 - Diffusion-weighted imaging can add value to emergency MRI for the diagnosis of acute appendicitis among pregnant women

05:48Y. Wong, Gueishan / TW

Purpose:

We investigated whether or not diffusion-weighted imaging (DWI) can add value to emergency non-enhanced MRI for the correct diagnosis of acute appendicitis among pregnant women.

Methods and materials:

From January 2018 to August 2019, 44 consecutive emergency MRI without contrast enhancement were performed to pregnant women for assessing acute right lower abdominal condition. Their median age was 33.0 years (IQR 27.0, 36.0) and median gestation age was 16 weeks (IQR 10.3, 24.9). All MRI examinations were retrospectively reviewed for acute appendicitis by two readers independently. The charts were reviewed for final diagnosis. The sensitivities, specificities, positive predictive value (PPV) and negative predictive value (NPV) of the variables for final diagnosis of acute appendicitis were computed.

Results:

Among 44 patients, the final diagnosis of acute appendicitis was confirmed in 17 (38.6%). The final diagnosis of the other 27 patients was not acute appendicitis. They comprised 4 patients who underwent surgeries for other abdominal conditions and 23 patients who did not undergo surgery but were discharged without complications. Of 44 MRI examinations, 20 had DWI pulse sequence. Twenty four MRI did not have DWI. The final diagnosis was significantly associated with DWI (p<0.001), T2WI (p<0.001) and fat-saturated T2WI (p<0.001). The sensitivities, specificities, PPV and NPV of MRI overall diagnosis for acute appendicitis were 100.0% (95%CI 100.0, 100.0), 100.0% (95%CI 100.0, 100.0), 100.0%, 100.0% (with DWI) and 100.0% (95%CI 100.0, 100.0), 82.4% (95%CI 64.2, 100.5), 70.0%, 100.0% (without DWI).

Conclusion:

DWI can add value to non-enhanced MRI for a correct diagnosis of acute appendicitis for pregnant women.

Limitations:

Limitations were selection bias and a small number of subjects. The research was done at a single institution and discrepancies were resolved between 2 readers.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

10
RPS 1117 - Diagnostic accuracy of multidetector CT in primary acute mesenteric ischemia

RPS 1117 - Diagnostic accuracy of multidetector CT in primary acute mesenteric ischemia

05:07M. Lanzetta, Florence / IT

Purpose:

The purpose was to evaluate the diagnostic accuracy of multidetector CT in detecting primary acute mesenteric ischemia (AMI), considering laparoscopy, laparotomy, or other diagnoses confirmed by clinical-laboratoristic data as the reference standard.

Methods and materials:

Initial MDCT scans in 231 patients with acute abdomen associated with metabolic acidosis and/or hyperlactatemia were retrospectively reviewed. CT was classified as positive or negative from the radiologist’s report and considered true positive (TP) or false positive if laparoscopy/laparotomy or clinical status and serial MDCTs confirmed the diagnosis of AMI or not. A negative CT was considered true negative when laparoscopy/laparotomy excluded AMI or when CT findings oriented for another diagnosis were confirmed by clinical-instrumental data. Negative CT was considered false negative (FN) if laparoscopy/laparotomy confirmed the diagnosis of AMI. Moreover, all MDCTs were retrospectively evaluated by a skilled radiologist for the presence of vascular, bowel wall, mesenteric-peritoneal findings, and the involvement of other organs.

Results:

Out of all 231 patients who underwent contrast-enhanced MDCT for acute abdomen and abnormal laboratory tests, 60 turned out to have AMI. MDCT shows a sensitivity of 82%, a specificity of 96%, a positive predictive value of 87%, a negative predictive value of 94% and a diagnostic accuracy of 92% in the diagnosis of AMI. Absent/decreased enhancement of the bowel wall was detectable in all patients with AMI; absent/decreased enhancement of the bowel wall, and colon involvement were more frequent in FN compared to TP; paper thin wall and peritoneal findings were more frequent in intestinal injury requiring resection.

Conclusion:

MDCT plays a fundamental role in the diagnosis of AMI, making it possible to identify findings more correlated with ischemic damage requiring intestinal resection.

Limitations:

Retrospective design from a single centre, lack of control group.

Ethics committee approval

Ethics committee approval was obtained.

Funding:

No funding was received for this work.

PEP Subscription Required

This course is only accessible for ESR Premium Education Package subscribers.