CTiR 7 - Clinical Trials in Radiology 2

Purpose:

Proper de-identification of medical images while maintaining research-relevant information and connection between cross-modality image studies is complex and cumbersome. We aimed to develop a medical image anonymisation system that integrates currently available open-source tools to overcome the de-identification difficulties in a large multi-center cross-modality AI imaging project.

Methods and materials:

Cardiovascular images including cardiac CT, nuclear myocardial perfusion images, invasive coronary angiography, coronary fractional flow reserve, coronary optical coherence tomography, and intravascular ultrasound from eight institutes were retrospectively collected to build the database. The images were de-identified with an on-site standalone anonymiser then sent to the Core Lab for main de-identification process. The main de-identification processor is composed of an open-source DICOM server with a custom server-side script, a MIRC clinical trial processor with custom pipelines, and python scripts for special anonymisation and processing needs. Each component runs in a separate container but connect to the same virtual network. De-identification profile and server settings are versioned and tested in a separate container before publishing.

Results:

A total of 12 million multi-center cross-modality cardiovascular images from 2000 patients were de-identified successfully while preserving research-relevant information and connection between cross-modality image studies. A special de-identification process for specific type of image annotation was developed and integrated into the main de-identification process.

Conclusion:

We successfully built a medical image anonymisation system in a multi-center cross-modality cardiovascular imaging database by integrating available tools.

Limitations:

(1) Lack of real time progress monitoring and feedback for end users.
(2) Requires manual manipulation to redo de-identification with new profile.

Ethics committee approval

Research Committee of National Taiwan University Hospital.

Funding:

Ministry of Science and Technology, Taiwan.

Purpose:

To determine if the addition of abbreviated breast MRI (A-MRI) to surveillance mammography (MG) impacted patient anxiety and if it improved cancer detection rate (CDR) in breast cancer survivors.

Methods and materials:

A prospective parallel unblinded controlled trial was performed in a tertiary care academic center on women with a personal history of breast cancer who were randomised into two groups: surveillance with MG or MG plus A-MRI. The primary outcome was anxiety measured via four validated questionnaires at the time of enrollment, after notification of imaging results, and 6 months later, and compared between the two groups. Other parameters including recall, biopsy rates, and CDR were compared between MG and A-MRI.

Results:

199 patients were included, 105 women in A-MRI plus MG and 94 in MG groups. MRI recall rate was 26.6% (28/105), biopsy rate was 20% (21/105), with CDR 5.7% (6/105) - 6 cancers found by MRI alone. Compared to MRI alone, the rates of MG recall, biopsy and CDR were significantly lower: 4.5% (9/199, p<0.00001), 3.0% (6/199, p<0.00001), 0.5% (1/199, p<0.0039) respectively - 1 DCIS found in a patient who had MG only. Penn State Worry Questionnaire scores were similar for both groups and did not change over time, both moderately severe.

Conclusion:

Compared to mammography alone, A-MRI had significantly higher cancer detection in breast cancer survivors. MRI was not associated with a “reassuring effect”, but showed no increase in anxiety despite a higher rate of recalls and biopsies. Further work is needed to ensure an acceptable benefit-to-harm ratio.

Limitations:

Potential for selection bias due to oncologist preference.

Ethics committee approval

Approved by the institutional review board and written informed consent was obtained.

Funding:

Support was provided in part by the Ottawa Hospital Cancer Program.

Purpose:

To evaluate the performance of ultrasound (US) in diagnosing patients with pancreatic ductal adenocarcinoma (PDAC) in our region.

Methods and materials:

Patients diagnosed with PDAC between 2014-15 were identified by the Cancer Registry data. Their US requisitions, images, and reports were reviewed independently by a radiology resident and abdominal radiologist, and finalised via consensus. Examinations were excluded if a pancreatic lesion was known at the time of US.

The following data elements were extracted: clinical suspicion of PDAC, comment on image quality, detection of tumour, location, size, detection of secondary signs and liver metastases, suspicion of neoplasm raised, and follow-up recommendations. US were graded as true-positive, indeterminate, or false-negative. One-way ANOVA and chi-square tests were performed according to these subgroups.

Results:

113 US examinations on 107 patients (53 men, 54 women; mean 70 +/- 13 years) were included. Cancer was suspected clinically in 49 patients. 35/113 (31.0%) examinations commented on image quality, however only 49/113 (43.4%) visualised the tumour. There were 50 true-positives, 40 indeterminates, and 23 false-negatives. There was no difference in age, weight, tumour location, or size across subgroups. However, sex distribution, clinical suspicion of PDAC, comment on US quality, visualisation of tumour, detection of secondary signs, and liver metastases were significantly different (p<0.005). The number of examinations reporting suspicion of neoplasm or recommending follow-up also varied significantly (p<0.0001).

Conclusion:

Assessment of PDAC with US is unreliable in our region, with a large proportion of indeterminate and false-negative studies. These results have important implications for radiologists reporting US and referring physicians relying on US for patient work-up.

Limitations:

Retrospective study. Operator dependent nature of US.

Ethics committee approval

REB approval, PHIA adherence.

Funding:

No funding was received for this work.

Purpose:

To compare the diagnostic accuracy of dynamic contrast-enhanced computed tomography (DCE-CT) and ¹⁸Fluorine-Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography (PET/CT) for the diagnosis of solitary pulmonary nodules.

Methods and materials:

In this prospective multicentre trial, 380 participants with an incidentally detected solitary pulmonary nodule (8-30mm), and no recent history of malignancy, underwent DCE-CT and PET/CT. All patients underwent either biopsy with histological diagnosis or completed 2 years of CT follow-up. PET/CT was considered positive if it met one of the following criteria: nodule tracer uptake equal to or greater than the mediastinum with irregular/spiculate morphology on CT or evidence of distant metastases. DCE-CT was considered positive if the degree of nodule enhancement was ≥20 Hounsfield units. The primary outcome measures were the comparative diagnostic accuracies of DCE-CT and PET/CT.

Results:

Of 380 participants recruited, 312 (47% female, 68.1±9.0 years-old) completed the study. The median pulmonary nodule diameter was 14mm (IQR 11-19), with a 61% rate of malignancy. The sensitivity and specificity for DCE-CT was 95.2% (95%CI 91.2;97.5) and 29.3% (95%CI 22.0;37.8), and for PET/CT was 72.5% (95%CI 65.7;78.4) and 80.5% (95%CI 72.6;86.5). The area under the receiver operator characteristic curve (AUROC) was 0.63 (95%CI 0.58;0.67) for DCE-CT and 0.76 (95%CI 0.72;0.82) for PET/CT (p<0.001). Combining the tests resulted in a significant increase in the diagnostic accuracy over PET/CT alone (AUROC=0.89 (95%CI 0.86;0.93),p<0.001).

Conclusion:

PET/CT has a superior diagnostic accuracy to DCE-CT for the diagnosis of solitary pulmonary nodules. However, combining the two techniques improves the diagnostic accuracy over either test alone.

Limitations:

Results may not be generalisable to screening detected nodules.

Ethics committee approval

REC-12/SW/0206.

Funding:

NIHR-HTA (09/22/117).

Purpose:

To compare the safety and efficacy of bed-side ultrasound-guided (US) Trocar versus US/fluoroscopy-guided Seldinger techniques for percutaneous cholecystostomy (PC).

Methods and materials:

This is a prospective, two-center, non-inferiority study comparing bed-side US-guided Trocar PC (group T; 53 patients, mean age:74.31±16.19 years, male:28) versus US/fluoroscopy-guided Seldinger PC (group S; 52 patients, mean age: 79.92±13.38, male:26) in consecutive patients undergoing PC in two large tertiary university hospitals. Primary endpoints were technical success and procedure-related complications rates. Secondary endpoints included procedural duration, pain assessment, and clinical success in up to 3 months of follow-up.

Results:

PC was successfully performed in all 105 patients (100%). Clinical success was similar between the two study groups (86.8% group T vs.76.9% group S; p=0.09). Mean procedural time was significantly lower in group T (1.77±1.62min vs. 4.88±2.68min group S; p<0.0001). Significantly more procedure-related complications were noted in group S compared to group T (11.5% vs. 1.9%; p=0.02). In group S, bile leak was 7.7%, abscess formation 1.9%, and gallbladder rupture 1.9%. No procedure-related death was noted. Minor bleeding occurred in one patient (1.9%) in group T, which auto-resolved. At 12 hours, pain score was significantly lower for patients in group T (0.78 ± 1 vs. 3.12 ± 1.36; p=0.0001).

Conclusion:

Bed-side, US-guided Trocar PC was equally effective with less procedure-related complications, required less procedural time, and resulted in decreased post-procedural pain, compared to fluoroscopically-guided PC using the Seldinger technique.

Limitations:

Relatively small number of patients included. Each center performed only one method for PC. Patients were not blinded to the study’s method.

Ethics committee approval

Institutional board approval was obtained and an informed consent explaining the risks and benefits of the procedure was obtained in all cases.

Funding:

No funding was received for this work.