S 7 - My educational or social project at my university

Purpose:

Although Monte Carlo simulation by using the specification of the linear accelerator (LINAC) is the most common method to determine the energy spectrum, it consumes a lot of computing time and the treatment head specification is not always easy to obtain. Therefore, a simplified method was discussed in this study to estimate the energy spectrum that can reproduce the depth dose agreed with the measured data. To estimate the energy spectrum, a monoenergetic depth dose library was provided. Characteristics of the energy spectrum as a function of distance from the central axis were also described in this study.

Methods and materials:

The monoenergetic depth dose of 0.1MeV to 11MeV was simulated using the DOSRZnrc code with a history of 1000000000 to obtain better statistical accuracy. The energy spectrum was reconstructed using the monoenergetic depth dose library and a GRG (generalised reduced gradient) non-linear function was performed repeatedly until a good agreement between the calculated and measured depth dose was obtained.

Results:

The total relative difference of the reconstructed and measured dose distribution was less than 1%, which provides good accuracy in the energy spectrum estimation.

Conclusion:

A simplified method to estimate the energy spectrum incident on the surface of a water phantom was successfully implemented, which can be used as a fast energy spectrum estimation technique to obtain an accurate calculation of dose distribution.

Limitations:

n/a

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To retrospectively assess the rate of missed lung cancer on radiographs/CTs to identify the type of errors that occur and correlate the results to international guidelines for recommended miss rates for radiological lung cancer diagnosis.

Methods and materials:

A manual search in PACS of CT thorax examinations from 01/10/2017-31/03/2019 was performed. CTs for cancer follow-up were excluded. Cases with lung cancers/indeterminate lesions were collected. Prior examinations (cut-off 2yrs) were searched for and each case classified:

1. Lesion not visible/no prior examinations/correct follow-up.

2. Lesion not reported.

3. Lesions reported but misinterpreted.

4. Lesions reported correctly but insufficiently conveyed to a clinician.

5. Lesions reported and conveyed correctly, but insufficient clinical follow-up.

6. Technical issues.

The time-lapse due to missed diagnosis was recorded.

Results:

2,266 CTs were initially reviewed and 50 were found to have lung cancer or indeterminate lesions.

The radiological diagnosis was correct in 41 (82%) cases.

35 (70%) were from group 1.

4 (8%) were from group 2.

5 (10%) were from group 3

6 (12%) were from group 5.

No cases were from group 4/6.

Of the 4 missed lesions, 2 were located in the left hilum and 1 each in the right and left upper lobe. Of the 5 misinterpreted lesions, 2 were considered an infection, 2 fibrosis, and 1 calcification.

Of the 6 without sufficient follow-up, 2 were elderly with advanced cancer and unable to undergo treatment, 2 postponed treatment themselves, and 2 were truly not followed up despite radiologists' recommendation.

The time-lapse ranged from 2-24 months.

Conclusion:

The overall missed rate of lung cancer fulfilled international recommendations (>75%). Misinterpretation was a major issue as well as misses in the hilum. Education and training need to focus on that.

Limitations:

As CT examinations for cancer follow-ups were excluded, new cancers occuring there could have been missed.

Ethics committee approval

Ethics committee review and data protection regulations were followed.

Funding:

No funding was received for this work.

Purpose:

To investigate the performance of a deep learning-based algorithm for fully automated quantification of left ventricular volumes and function in cardiac MRI.

Methods and materials:

We retrospectively analysed CMR examinations of 50 patients (74% men, median age 57 years). The most common indications were known or suspected ischaemic heart disease, myocarditis, or cardiomyopathies. Fully automated volumetric analysis of left ventricular volumes and function was performed using a deep learning-based algorithm. The analysis was subsequently corrected by a cardiovascular imaging expert as necessary. Manual volumetric analysis was performed by two radiology trainees. The time required for analysis was recorded. Volumetric results were compared using Bland-Altman statistics.

Results:

The fully automated volumetric analysis was successfully completed in all patients in a median of 8.4 seconds. With expert review and corrections, the analysis required a median of 110 seconds. The median time required for manual analysis was 3.5 minutes for a cardiovascular imaging fellow and 9 minutes for a radiology resident. Compared with the expert-corrected results, the fully automated algorithm showed a mean deviation of -2.1% for end-diastolic volume, +3% for end-systolic volume, -10% for stroke volume, and -7.5% for ejection fraction. No corrections were necessary for 10 patients (20%). In the remainder, corrections were made on a median of 2 slices. The most common correction was related to inclusion or exclusion of slices at the apex or base of the left ventricle.

Conclusion:

The deep learning-based algorithm allows for a fully automated analysis of left ventricular volumes and function with good accuracy. Although manual corrections are required in most patients for precise results, this approach is very time-efficient compared to manual analysis.

Limitations:

A limited number of 50 cases.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To determine the accuracy and diagnostic performance of regadenoson-induced stress perfusion cardiac magnetic resonance (CMR) imaging in routine clinical practice.

Methods and materials:

403 consecutive individuals with known or suspected coronary artery disease
underwent conventional stress/rest perfusion CMR in a 1.5-Tesla MR system
(MAGNETOM Aera, Siemens Healthineers) for clinical indication. An intravenous bolus of regadenoson (0.4mg) was employed as a stressor. Stress perfusion CMR studies were visually assessed for perfusion deficits. CMR results were compared with invasive coronary angiography, considering positive 50% stenosis in at least one coronary vessel.

Results:

A total of 402 individuals (305 males, mean age 64±11.8 years) were retrospectively evaluated. One patient could not complete the examination due to severe chest pain requiring nitroglycerin. Most frequently, the test was performed in patients with chest pain and prior coronary revascularisation (34.7%). Stress-CMR resulted as positive in 123 patients (30.6%) and negative in 279 (69.4%). Invasive coronary angiography was performed in 102 individuals. Compared with coronary angiography, regadenoson stress perfusion CMR showed a sensitivity of 94.5% (95% CI, 86.6%-98.5%), specificity of 82.8% (64.2%-94.1%), positive predictive value of 93.2% (86.1%-96.8%), negative predictive value of 85.7% (69.5%-94%), and accuracy of 91.2% (83.9%-95.9%).

Conclusion:

Regadenoson stress perfusion CMR has a high diagnostic accuracy for the detection of significant coronary artery disease in routine clinical practice.

Limitations:

A retrospective study.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Sonography is often the first imaging technique undergraduate medical students become familiar with. In German medical school, the curriculum is no defined space for practical skills training in imaging. Until 2017, there was only one available sonography device for training without any form of instruction for all medical students between year 3 and 6 at the Johannes-Gutenberg University Mainz.
This was the motivation for establishing a new peer-to-peer student sonography course (Sono-For-Klinik, Mainz).

Methods and materials:

In early 2017, we trained 22 student tutors with the help of the Departments of Radiology and Internal Medicine. Furthermore, we wrote a script with over 180 pages and created several teaching materials. Course rooms with electronic equipment and sonography devices were provided by the Rudolf-Frey-Lernklinik Mainz. Moreover, we linked in with external clinical partners who provided us with images and expertise. A standardised questionnaire was used for the evaluation of the course. The gained practical skills of the students were verified with objective structured clinical examinations.

Results:

The first course was able to start in summer, 2017, with 60 students. The number of participants reached 151 in the next semester. Today, 1,040 students have participated in the course and 42 tutors have been trained. New sub courses (e.g. with a focus on echocardiography or FAST) have been created. We were certified as an official course by the DEGUM (German Society for Sonography) in 2018.

Conclusion:

The need for practical imaging skills is huge among undergraduate students. Since the establishment of the course, all of our course capacities have been occupied. Due to self-motivation, it was possible to feed this need and to establish our educational project with over 1,000 participants in 2.5 years.

Limitations:

The retrospective design.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Gadolinium use is not without risk and economic cost. With patient safety of paramount importance and in this time of increasing health economic scrutiny, consideration must be given to alternative safer and less-expensive diagnostic and surveillance imaging techniques without sacrificing outcomes. This study's purpose was to evaluate whether T2-weighted imaging (T2-MRI) can provide an accurate alternative to gold-standard T1-weighted post-gadolinium imaging (Gd-MRI) in the assessment of non-operated cranial meningiomata.

Methods and materials:

90 meningiomata (30 skull base, 30 infratentorial, and 30 supratentorial) were identified from a tertiary neuroscience centre database. Two neuroradiologists independently measured the single maximal diameter of meningioma on either axial T2-MRI or Gd-MRI without reference to the other sequence. Observers were blind to previous measurements and clinical reports. This was repeated after a washout period (2 weeks) to permit intra and interobserver correlation. Bland–Altman plots evaluated the level of agreement (set at +/-1 mm) between meningioma measurements obtained on T2W-MRI and Gd-MRI.

Results:

Neuroradiologists showed excellent intra and interobserver concordance on both Gd-MRI and T2-MRI (intra and inter ICC>0.9, irrespective of meningioma location). On Bland-Altman plots, the mean difference in measurements of supratentorial, infratentorial, and skull base meningiomata using T2-MRI and Gd-MRI were 0.46 mm, 0.78 mm, and 1.64 mm, respectively.

Conclusion:

T2-MRI should be considered as an alternative to Gd-MRI in the surveillance of non-operated meningiomata located in the supra and infratentorial compartments. Change to the surveillance regime may provide health and economic benefits. Gd-MRI is still indicated in the assessment of skull base meningiomata. Skull base tumours are often intimately related to other complex anatomical structures and the use of Gd-MRI may enhance local soft-tissue contrast in order to better define tumour extent.

Limitations:

A single-centre, consultant-only study with a small sample and differing protocols.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

The purpose of this study was to evaluate the efficiency of using different 3D-printed models and the CTs of the patients from which the models were created for teaching sectional anatomy to medical students. Our hypothesis was that in order to understand sectional or radiological anatomy one must first create a three-dimensional reconstruction of that structure in their mind. By creating a physical 3D reconstruction using the patient's own CT, we can make sure that the model we created is more accurate and that the learning process would, therefore, be much easier.

Methods and materials:

Two 3D-printing technologies (Fused deposition modelling and stereolithography) were used to create the following models: scoliosis, pelvic bone, coronary arteries, 2 aorta models, heart, and a tumour model. During 5 hands-on workshops with a total of 50 student participants, we presented this subject and we also provided a 5 question questionnaire at the end that we used to interpret the results.

Results:

80% of the students completely agreed that the 3D-printed models were useful in understanding sectional anatomy. 76% would use a similar model of another structure for the same purpose. 82% would recommend using a 3D-printed model to another student or colleague. 76% of the students agreed that a lot of credibility to this study method was attributed to the models being an accurate representation of the patients anatomical structure and over 96% of them would like to see 3D-printed models being implemented in university courses.

Conclusion:

In conclusion, the questionnaire results showed that 3D-printed models can be a learning bridge from conventional to sectional anatomy.

Limitations:

This learning method is completely dependent on the lecturer's public speaking and teaching skills.

Ethics committee approval

n/a

Funding:

No funding was received for this work.