Deep learning for classification of thyroid nodules on ultrasound: validation on an independent dataset.

OBJECTIVES: The purpose is to apply a previously validated deep learning algorithm to a new thyroid nodule ultrasound image dataset and compare its performances with radiologists. METHODS: Prior study presented an algorithm which is able to detect thyroid nodules and then make malignancy classifications with two ultrasound images. A multi-task deep convolutional neural network was trained from 1278 nodules and originally tested with 99 separate nodules. The results were comparable with that of radiologists. The algorithm was further tested with 378 nodules imaged with ultrasound machines from different manufacturers and product types than the training cases. Four experienced radiologists were requested to evaluate the nodules for comparison with deep learning. RESULTS: The Area Under Curve (AUC) of the deep learning algorithm and four radiologists were calculated with parametric, binormal estimation. For the deep learning algorithm, the AUC was 0.69 (95% CI: 0.64-0.75). The AUC of radiologists were 0.63 (95% CI: 0.59-0.67), 0.66 (95% CI:0.61-0.71), 0.65 (95% CI: 0.60-0.70), and 0.63 (95%CI: 0.58-0.67). CONCLUSION: In the new testing dataset, the deep learning algorithm achieved similar performances with all four radiologists. The relative performance difference between the algorithm and the radiologists is not significantly affected by the difference of ultrasound scanner.

Can procedure time for paracentesis be optimized based on bottle selection?

PURPOSE: The purpose of our study was to assess if plastic containers could decrease the overall procedure time for paracentesis relative to more commonly used glass containers. METHODS: In this IRB exempt study, initial pilot data comparing filling time of glass and plastic containers in an ex vivo setting under identical conditions revealed power calculations that n = 37 patients per group would be needed to achieve standard deviation (SD) = 60 s, difference (diff) = 40 s, two-tailed alpha-level 0.05, and power 80%. Total of 43 patients (93 containers) were enrolled and randomized to glass or plastic bottles at enrollment. Timing of bottle filling was assessed using standardized sonographic screen captures. RESULTS: An interim look at statistics at n = 20 patients indicated that original conjectures from pilot data were conservative and smaller sample size was sufficient to stop the study and conduct the analyses. Specifically, SD = 54 s, diff = 49 s, two-tailed alpha-level 0.05, and power 80% required n = 21 patients per group. Plastic containers had a statistically significantly lower average filling time per bottle (162.7 ± 53.3 s) compared to glass (212.2 ± 50.4 s) (p = 0.003). Viscosity was calculated for each specimen and did not affect the statistical significance of the results (p = 0.32). CONCLUSION: Plastic containers have 50 s time savings of per bottle filling time as compared to glass bottles as theorized based on their faster flow rate. This holds true in both an ex vivo setting and in patients and can have important downstream impacts on patient throughput, provider efficiency and system wide cost savings.

Automated Structured Reporting for Thyroid Ultrasound: Effect on Reporting Errors and Efficiency.

PURPOSE: To compare the effectiveness of different reporting templates using the ACR Thyroid Imaging Reporting and Data System (TI-RADS) for thyroid ultrasound. METHODS: In this retrospective study, four radiologists implemented ACR TI-RADS while dictating 20 thyroid ultrasounds for each of four different templates: free text, minimally structured, fully structured, fully structured and automated (embedded software automatically sums TI-RADS points, correlates with nodule size, and inserts appropriate recommendation into report impression). In total, 80 reports were constructed per template type. Frequencies of different errors related to ACR TI-RADS were recorded: errors in point assignment, point addition, risk-level assignment, and recommendation. Reporting times were recorded, and a survey about using the template was administered. Differences in error rates were compared using χ2 and Fisher's exact tests, and differences in reporting times were compared using Kruskal-Wallis tests. RESULTS: Across all readers, errors were identified in 27.5% of reports (22 of 80) for the free text template, 28.8% (23 of 80) for the minimally structured template, 18.8% (15 of 80) for the fully structured template, and 0% (0 of 80) for the fully structured and automated template (P < .0001). Frequency of each error type (number assignment, addition, TR categorization, recommendation) decreased across the four templates (P < .0005 to P < .005). Median reporting times for the less complex templates were 210 to 240 seconds, whereas the median automated template reporting time was 180 seconds (P = .41). Radiologists subjectively preferred using the automated template. CONCLUSION: A structured reporting template for thyroid ultrasound that automatically executed steps of ACR TI-RADS resulted in fewer reporting errors for radiologists.

Preprocedural computed tomography before cardiac implanted electronic device lead extraction: Indication, technique, and approach to interpretation.

Cardiac implantable electronic devices (CIEDs) frequently need to be extracted due to infection, hardware failure, and other causes. The extraction of the CIED is typically performed using percutaneous methods. While these procedures are mostly performed without incident there is a small risk of significant complications. Dedicated imaging pre-CEID removal to include the central veins and heart with multidetector computed tomography (MDCT) can be utilized to evaluate the lead course and termination, the integrity of the central veins and cardiac chambers, and identify potential complications that may alter the lead extraction procedure as well as reimplantation of subsequent leads. Indications for preprocedural imaging, the technique of dedicated preprocedural lead extraction MDCT, and the approach to the interpretation of the images is discussed in this review.

Moving Radiology Workflow to the Electronic Health Record: Quantitative and Qualitative Experience From a Large Academic Medical Center.

RATIONALE AND OBJECTIVES: To objectively and subjectively evaluate a large, academic radiology department's transition to electronic health record (EHR) centered workflow. MATERIALS AND METHODS: Multiple metrics were compared from before and after the move to EHR-driven workflow. Examination ordering and reading priority data were obtained for 30 days both before and after the transition. Sixteen radiologists were observed opening a computed tomography (CT) examination, and time to open, mouse clicks, and keystrokes were recorded. Information available to the radiologist during interpretation was also compared. Additionally, a 12 question survey was sent out to the residents and faculty both before and after the transition. RESULTS: Implementation of an eight-level reading priority system increased worklist granularity and improved identification of more urgent studies to read. Radiologists opened CT studies in picture archiving and communications system-driven workflow in 52.4 ± 16.9 seconds using 9.5 ± 3.9 clicks and 6.3 ± 2.9 keystrokes, compared to 17.3 ± 9.5 seconds, 4.8 ± 1.5 clicks, and 0.1 ± 0.3 keystrokes in EHR-driven workflow (p < 0.001 for each measure). More information was available to the radiologist during examination interpretation, and 54.7% of radiologists rated the ease of use of the new system as good or very good (compared to 4.2% for the old system, p < 0.001). CONCLUSION: Transitioning to an EHR-driven workflow at a large academic medical center improved efficiency, was favorable to radiologists, and enhanced examination prioritization.

DIfferential Subsampling With Cartesian Ordering With Respiratory Triggering Versus Conventional Liver Acquisition With Volume Acquisition: A Multiple Reader Preference Study.

OBJECTIVE: The aim of this study was to compare respiratory-triggered DIfferential Subsampling with Cartesian Ordering (rtDISCO) and breath-held Liver Acquisition with Volume Acquisition (LAVA) image quality. METHODS: In this institutional review board-approved, Health Insurance Portability and Accountability Act-compliant prospective study, 25 subjects underwent T1 imaging with rtDISCO and LAVA before and after intravenous contrast. Three readers scored individual series and side-by-side comparisons for motion and noise. Eight clinical tasks were qualitatively assessed. RESULTS: As individual series, readers rated rtDISCO images as more degraded by motion on both precontrast (mean rtDISCO score, 2.7; LAVA, 1.6; P < 0.001) and postcontrast images (rtDISCO, 2.4; LAVA, 1.8; P < 0.001). Readers preferred LAVA images based on motion on both precontrast (mean preference, -1.2; P < 0.001) and postcontrast images (mean preference, -0.7; P < 0.001) on side-by-side assessment. There was no preference between sequences for 6 of 8 clinical tasks on postcontrast images. CONCLUSIONS: Readers preferred LAVA with respect to motion but not noise; there was no preference in most of the tested clinical tasks.

Imaging-guided percutaneous thrombin injection for the treatment of iatrogenic femoral artery pseudoaneurysms.

PURPOSE: To evaluate factors that may affect successful ultrasound-guided percutaneous thrombin injection of iatrogenic femoral artery pseudoaneurysms (PSA). MATERIALS AND METHODS: This was an IRB-approved, HIPAA-compliant retrospective study of 326 consecutive subjects (138 males, 188 females; mean age 68 years, range 18-95) who underwent thrombin injection for treatment of femoral PSA; follow-up ultrasound was available in 145 subjects. The number of PSA lobes and dimensions, pre-procedure laboratory values (international normalized ratio [INR], activated partial thromboplastin time [aPTT], platelet count), and concomitant anticoagulation therapy were recorded. RESULTS: Technical success was achieved in 98.2% (320/326) of subjects. Primary effectiveness (complete thrombosis at 24 h) was achieved in 74.5% (108/145). Twenty-five subjects underwent repeat thrombin injection, successful in 21 subjects, for a total effectiveness rate of 97.0% (129/133). No imaging factor was associated with technique failure, including number of lobes (p = 0.898), largest dimension (p = 0.344), or volume (p = 0.697). No statistically significant difference in pre-procedure INR, aPTT, or platelet count was found between subjects with CT and those with IT (p > 0.138). Anticoagulation therapy was associated with incomplete thrombosis (35.5% [38/107] for CT vs. 63.9% [23/26] for IT; p = 0.002). CONCLUSION: Imaging-guided percutaneous thrombin injection has high technical success and effectiveness rates for the treatment of iatrogenic femoral artery PSA. Anticoagulation therapy was the only factor associated with incomplete thrombosis.

Structured reporting of CT enterography for inflammatory bowel disease: effect on key feature reporting, accuracy across training levels, and subjective assessment of disease by referring physicians.

PURPOSE: To compare the content and accuracy of structured reporting (SR) versus non-structured reporting (NSR) for computed tomographic enterography (CTE) of inflammatory bowel disease (IBD). MATERIALS AND METHODS: This IRB-approved, HIPAA-compliant, retrospective study included 30 adult subjects (15 male, 15 female; mean age 41.9 years) with IBD imaged with CTE. Nine radiologists (3 faculty, 3 abdominal imaging fellows, and 3 senior radiology residents) independently interpreted all examinations using both NSR and SR, separated by four weeks. Reports were assessed for documentation of 15 key reporting features and a subset of 5 features was assessed for accuracy. Thirty faculty reports (15 NSR [5 per reader] and 15 SR [5 per reader]) were randomly selected for review by three referring physicians, who independently rated quality metrics for each report. RESULTS: NSR documented the presence or absence of 8.2 ± 2.2 key features, while SR documented 14.6 ± 0.5 features (p < 0.001). SR resulted in increased documentation of 13 of 15 features including stricture (p < 0.001), fistula (p < 0.001), fluid collection (p = 0.003), and perianal disease (p < 0.001). Among a subset of five features, accuracy for diagnosing multifocal disease was minimally increased when using SR (76% NSR vs. 83% SR; p = 0.01), but accuracy for other features was not affected by report type. Referring physicians significantly preferred SR based on ease of information extraction (p < 0.01). CONCLUSION: Structured reporting of CTE for IBD improved documentation of key reporting features for trainees and faculty, though there was minimal impact on accuracy. Referring physicians subjectively preferred the structured reports.

Use of Preprocedural MDCT for Cardiac Implantable Electric Device Lead Extraction: Frequency of Findings That Change Management.

OBJECTIVE: Five percent of cardiac implantable electric devices (CIEDs) are removed each year. Percutaneous extraction is preferred but can be complicated if the leads adhere to the vasculature or perforate. The goal of this study is to assess the frequency of findings on dedicated MDCT that alter preprocedural planning for percutaneous CIED extraction. MATERIALS AND METHODS: One hundred patients with CIEDs who underwent MDCT before percutaneous lead extraction were analyzed. Major findings that could preclude percutaneous removal, including lead course and termination, were distinguished from moderately significant findings that could alter but not preclude percutaneous removal, including endofibrosis of leads to the vasculature, lead termination abnormalities, central vein stenosis, or thrombus. Incidental findings were characterized separately. Findings were correlated with preprocedural decisions, the extraction procedure performed, and procedural outcomes. RESULTS: Twenty-six women and 74 men with 125 right ventricular leads, 84 right atrial leads, and 26 coronary venous leads were evaluated. Major findings were present in 7% of patients, including six patients with lead perforation and one with a lead coursing outside a tricuspid annuloplasty ring. Moderately significant findings of endothelial fibrosis were found in 78% of patients. The central veins were narrowed or occluded in 42% of patients, and thrombus was present in 2% of patients. Thirty-six percent of patients had incidental findings, and 4% of patients had unexpected findings requiring immediate inpatient attention. CONCLUSION: MDCT performed before CIED lead extraction is able to identify major and moderately significant findings that can alter either percutaneous extraction or preprocedural planning. The use of dedicated preprocedural MDCT can help to stratify patient risk, guide decision making by the proceduralist, and identify non-catheter-related findings that affect patient management.