What’s new in Cardiothoracic Imaging – November 2020

3 years ago

Subtracted Computed Tomography Angiography in the Evaluation of Coronary Arteries with Severe Calcification or Stents Using a 320-Row Computed Tomography Scanner

Chen, Chun-Chi MD; Wu, Patricia Wanping MD; Tsay, Pei-Kwei PhD; Wang, Chun-Chieh MD; Toh, Cheng-Hong MD, PhD; Wan, Yung-Liang MD.

Journal of Thoracic Imaging: September 2020 – Volume 35 – Issue 5 – p 317-325.

In an interesting study, researchers from the Chang Gung University in Taoyuan, Taiwan compared the efficacy of subtracted coronary CT angiography (SCCTA) with non-subtracted coronary CT angiography (NSCCTA) in patients with densely calcified coronary arteries and those with pre-existing coronary stents. The idea behind this trial originated from the observed poor diagnostic performance of CCTA in patients with coronary stents or severely calcified coronary arteries (defined as patients with a calcium score [CS] > 400) secondary to extensive streak artifact. Patients were screened for a multitude of coronary artery disease (CAD) risk factors including BP, LDL, BG, BMI, smoking history and family history of CAD. 39 total patients were enrolled into two different groups “stented” (n=12) and “heavily calcified” (n=20) to undergo SCCTA and NSCCTA prior to interventional coronary angiography (ICA) for validation of results. All examinations were performed on a 320 row Toshiba CT scanner and were performed in a prospective manner with a breath hold during a single heartbeat. For the SCCTA, a non-contrasted examination was first performed, followed by the contrasted portion; these data were then post processed on an offline workstation allowing for the creation of multiplanar reformats, maximum intensity projections (MIPS) and for volume rendering. The proximal, middle and distal segments of the right, left and left circumflex coronary arteries were evaluated by subtraction and non-subtraction techniques for image quality (using a four-point system from non-diagnostic, to poor, to diagnostic, to excellent), the presence of in-stent restenosis (ISR) or significant stenosis (defined as luminal narrowing >50%); any nonassessable segments were classified as ISR or stenosis. 30 stented segments in 12 subjects and 202 calcified segments in 20 patients were evaluated and sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive values (NPV) were estimated for the performance of SCCTA and NSCCTA in both of these two groups compared to findings on ICA. 3/30 of the stented segments and 32/202 of the calcified regions were determined to represent ISR or significant stenosis. SCCTA proved statistically superior to NCCTA in specificity, accuracy and PPV for the stented and heavily calcified groups with an impressive specificity of 100% and 97.6% compared to 55.6% and 76.5%, respectively. SCCTA also proved to produce higher quality images with a score of 3.27 ± 0.69 vs 2.59 ± 0.67 (P<.001). Other statistically significant differences included the highest total CS between stented groups (453.9 ± 507.8) and the calcified group (2023.2 ± 1144.9), P<.001. The heavily calcified group also incidentally exhibited an increased systolic blood pressure (152.0 ± 17.7) when compared to their stented compatriots (129.6 ± 18.7, P< .002). This suggests that SCCTA generates higher quality images that allow for the superior specificity, PPV and accuracy in diagnosing ISR and CAD in patients with prior stenting or severely calcified coronary arteries. SCCTA has some notable drawbacks including increased radiation dose compared to standard dual energy CCTAs and the need for a high level of patient cooperation to aid in images with high enough quality that are capable of subtraction.

 

Deep Learning–based Automatic Detection Algorithm for Reducing Overlooked Lung Cancers on Chest Radiographs

Sowon Jang, Hwayoung Song, Yoon Joo Shin, Junghoon Kim, Jihang Kim, Kyung Won Lee, Sung Soo Lee, Woojoo Lee, Seungjae Lee, and Kyung Hee Lee.

Radiology 2020 296:3, 652-661.

Researchers from Seoul, South Korea sought to evaluate the efficacy of a deep learning –based automatic detection algorithm (DLAD) in helping to identify subtle malignancies on chest radiograph. The authors combed through 2897 radiographs taken between January 2010 and December 2014 to identify 118 patients who had subtle visible lung cancers on previous chest radiographs; data were expanded to include 147 radiographs to include prior radiographs if the nodules were missed. A healthy control group of 234 patients was randomly selected from the 145,387 chest radiographs with an interpretation as “no active lesion” or “within normal limits” from the same time period. The 147 studies with nodules were combined with the 234 healthy control studies for a total of 381, this group was then randomly split into equal cohorts A, B, C, and D containing 95 or 96 chest radiographs. 6 chest radiologists and 3 radiology residents were then tasked to read each cohort twice, once with the assistance of DLAD and once without the DLAD with a 4-week washout period in between. The DLAD used is a commercially available convoluted neural network (CNN) which was created by Lunit (a South Korean company) utilizing 43,292 chest radiographs with 25 layers and 8 residual connections. The DLAD works by generating a color map over the radiograph that correlates to a lesion by using a 15% activation value and generating a probability value between 0 and 100%. The radiologists were tasked with replicating this function by identifying subtle areas of concern, placing a region of interest (ROI), assigning a confidence value (0-100) and determining whether or not to recommend a follow up CT scan. Operators’ performance with and without the DLAD were assessed by generating an alternative free-response receiver operating characteristic curve (AUC) using RJafroc software. Operators were also assessed on their relative sensitivity and specificity both with and without the DLAD. DLAD resulted in statistically significant increased sensitivity, specificity and AUC for both chest Radiologists and Radiology residents. Interestingly, the improved performance was greatest amongst residents compared to attending Radiologists. Most importantly, more subtle lesions were identified with the use of DLAD than without and the average reading time was decreased, although this was not statistically significant. There are multiple problematic areas in this study, mainly the relatively small sample size with a fairly large number of true positive films like increase readers’ specificity and does not accurately reflect clinical practice. Also, since the observers are reading each film twice, recall bias may have an untoward effect on results (even with a 4-week washout period). This article seems to reflect what others have suggested, in that the use of DLAD may augment the Radiologist’s ability to detect subtle cancerous lesions on chest radiograph, an area that is currently responsible for a large swath of medicolegal liability.

 

Myocardial Fibrosis and Inflammation in Liver Cirrhosis: MRI Study of the Liver-Heart Axis

Alexander Isaak, Michael Praktiknjo, Christian Jansen, Anton Faron, Alois M. Sprinkart, Claus C. Pieper, Johannes Chang, Rolf Fimmers, Carsten Meyer, Darius Dabir, Daniel Thomas, Jonel Trebicka, Ulrike Attenberger, Daniel Kuetting, and Julian A. Luetkens.

Radiology 2020 297:1, 51-61.

Researchers from the University Hospital Bonn in Germany designed a prospective trial to evaluate the association of liver cirrhosis and myocardial inflammation and fibrosis. Cirrhotic cardiomyopathy (CCM) has been shown to occur frequently in the cirrhotic patient population and represents a significant source of morbidity and mortality in a fragile patient population. This prospective study utilized multiparametric cardiac and hepatic MRI in 42 patients with known cirrhosis and portal hypertension without a significant cardiac history and 18 healthy controls referred for MRI to rule out structural causes of heart disease and non-specific cardiac symptoms. These 60 patients underwent combined cardiac and hepatic MRI exams performed using a single scan via 1.5T whole body MRI. Cardiac MRI was EKG gated for steady state face procession cine images. Myocardial T1 and T2 relaxation times were obtained via cardiac mapping and hematocrit-corrected extracellular volume (ECV) was determined using pre and post contrast T1 images. Late gadolinium enhancement (LGE) was assessed qualitatively by consensus agreement between two radiologists. Hepatic images were evaluated by one radiologist blinded to the cardiac findings. T1 and T2 relaxation times and ECV were obtained. Liver stiffness was estimated using hepatic elastography. During statistical analysis, the 42 cirrhotic patients were stratified by Childs-Pugh score into Class A (11), B (20), and C (11). No differences in left ventricular ejection fraction (LVEF) or LV end diastolic volume index (LVEDV) were observed. The cirrhotic group did exhibit statistically significant increases in heart rate (79 bpm ± 13 vs 66 bpm ± 12 P<.001) and cardiac indices (3.7 L/min/m2 ± 0.9 vs 2.9 L/min/m2 ± 0.5, P<.001). Myocardial T1 relaxation times, ECV and the presence of LGE were all statistically significantly elevated in the cirrhotic group compared to the healthy controls as well. These effects were also related to the patients’ Child-Pugh score, with patients who have more severe liver disease exhibiting statistically significantly increased T1 relaxation times, ECV and LGE; suggesting an association between the grade of liver disease and extent of myocardial fibrosis and inflammation. The longitudinal strain in cirrhotic patients was decreased in the cirrhotic patient population, suggesting a subclinical component of systolic dysfunction in this group. MR-elastography-based stiffness was the only reliable variable to be independently associated with LGE, increased myocardial T1 relaxation time, and ECV. Researchers found that by implementing a cutoff of 9.2kPa in liver stiffness, a sensitivity of 77.3% and a specificity of 75% for the presence of non-ischemic LGE. Since LGE is a marker for increased adverse cardiovascular outcomes in patients with non-ischemic cardiomyopathy (NICM) and all-cause mortality, it is suggested that patient’s with hepatic fibrosis may benefit from closer cardiac follow up. This study nicely demonstrates how the function of the heart and liver are intimately entwined, and that as liver function is impaired, serious cardiac dysfunction follows. There were several limitations to this study, mainly that the radiologists were not entirely blinded from hepatic findings, since part of the liver parenchyma is in view during short axis evaluation of the heart. Other limitations include the study’s sample size and the absence of radiologic-pathologic correlation of the presence of liver or heart fibrosis.

 

Calcium Scoring: A personalized probability assessment predicts the need for additional or alternative testing to coronary CT angiography

Judit Simon, Lili Száraz, Bálint Szilveszter, Alexisz Panajotu, Ádám Jermendy, Andrea Bartykowszki, Melinda Boussoussou, Borbála Vattay, Zsófia Dóra Drobni, Béla Merkely, Pál Maurovich-Horvat, and Márton Kolossváry

European Radiology, 2020; 30(10): 5499–5506

Researchers from the MTA-SE Cardiovascular Imaging Research Group and Semmelweis University in Hungary further wanted to evaluate the usefulness of the calcium score (CACS), in combination with anthropometrics and risk factors, for predicting if alternative diagnostic testing would be beneficial if coronary CTA showed nondiagnostic images or significant stenosis. The researchers retrospectively reviewed 4120 patients, which included 276 patients with nondiagnostic CTA in at least 1 coronary segment, of which 1073 patients had at least a >50% stenosis.  The researchers used multiple models to test their hypothesis.  Model 1 used included BMI, pulse, and arrhythmia.  Model 2 included the components of model 1 as well as including age, gender, and type of angina.  Model 3 included the components of model 2 and included CACS.  Model 1 was shown to have weak diagnostic accuracy (AUC 0.56), but when included with model 2 that discriminatory power improved (AUC 0.72, p <0.001), and additionally when adding CACS (model 3) that discriminatory power improved again (AUC 0.84, p <0.001).  Odds ratios were also established for multiple factors including: any type of arrhythmia increased odds of downstream testing by 2.12; atypical angina 1.29; and typical angina 1.64 (all p <0.001). When comparing the results of the 3 models with non-diagnostic CTA image quality, there was no significant model that showed increased discriminatory power (p >0.05).  In patients with significant coronary obstruction visualized on CTA, model 1 was shown to not have significant discriminatory power, however when combined with model 2 and model 3, diagnostic power improved (AUC 0.74 and p <0.001, and AUC 0.87, p <0.001 respectively).  Additionally, the researchers evaluated the classical different grades of calcium scores (1-10; 11-100; 101-400, 401-1000, >1000) and the relation to the type of angina, HR, and arrhythmia.  Due to the importance of each of these factors in regards to cardiac risk, the researchers discovered that there was no single cutoff value for CACS that could predict the need for further cardiac diagnostic testing. However, they have developed tables incorporating each factor which allows for guidance to determine if further downstream diagnostic testing is needed. The study has several limitations.  The study design was retrospective and data acquired from a single center and a single vendor.  Fractional flow reserve was not incorporated and the results of the downstream testing were not part of the study design.

Overall, the researchers conclude that routine evaluation of CACS before CTA may be advised as it helps to identify patients in which further evaluation will be necessary.

 

Observer Performance for Detection of Pulmonary Nodules at Chest CT over a Large Range of Radiation Dose Levels

Joel G. Fletcher , David L. Levin, Anne-Marie G. Sykes, Rebecca M. Lindell, Darin B. White, Ronald S. Kuzo, Vighnesh Suresh, Lifeng Yu, Shuai Leng, David R. Holmes III, Akitoshi Inoue, Matthew P. Johnson, Rickey E. Carter, Cynthia H. McCollough

Radiology. 2020 Sep 29;200969

Researchers for the Mayo Clinic wanted to investigate the role of low dose radiation Chest CT with that of routine dose Chest CT to evaluate indeterminate pulmonary nodules of 5 mm or greater (indeterminate defined as no definite benign features).  This was a retrospective case control study involving 83 patients that underwent routine dose Chest CT for pulmonary nodule evaluation.  The routine scans were acquired via single X ray source at 120 kV and automatic exposure control setting of 70 quality reference mAs (QRM), 0.5 second rotation time, and 128 x 0.6-mm detector configuration.  Subsequently the scans underwent filtered back projection (FBP) or iterative reconstruction (IR) to produce lower dose data sets using the original acquisition. Overall, the researchers found that per patient and per nodule sensitivity drops with declining radiation dose. However, specificity remains virtually unchanged despite a 25 fold difference in QRM between the routine dose and re-engineered low radiation dose (ranging from 70 QRM and 2.5 QRM).  Additionally, it was shown that sensitivity for subsolid nodules is significantly decreased for all dose configurations less than 10 QRM (p <0.05).  Multiple issues were raised with decreasing radiation dose including lower overall diagnostic image quality, greater image noise, and worsening sharpness.

Overall, the results show that radiation dose could be reduced by 1/7 of a routine chest CT, and ⅓ of the low dose screening CT, while still yielding noninferior performance for indeterminate pulmonary nodules.  Alternatively, part solid and subsolid nodules were shown to have decreased detection at radiation dose levels below that of the standard low dose screening CT.  Of note, the comparison of IR to FBP showed that IR has subjectively better image quality for lower doses, however the data showed no difference in observer performance. Although further research needs to be completed, further decreasing radiation dose for annual screenings should provide long term benefit.

The limitations of the study include a retrospective case control study design with limited (3) readers, who did not examine the non-axial multiplanar reformats.  The extrapolation from subspecialized radiologists to general practitioners is also a potential limitation.  The cohort had a greater proportion of subsolid nodules which may not reflect the true population prevalence.

 

CT-Based Biomarkers for Prediction of Chronic Thromboembolic Pulmonary Hypertension After an Acute Pulmonary Embolic Event

Giovanni Lorenz, Mnahi Bin Saeedan, Jennifer Bullen, Frederikus A Klok, Lucia J M Kroft, Lilian J Meijboom, Gustavo A Heresi, Apichaya Sripariwuth, Rahul D Renapurkar

American Journal of Roentgenology 2020 Oct;215(4):800-806.

Researchers from the Cleveland Clinic, Leiden University in The Netherlands, and Naresuan University in Thailand investigated the role of CT-based markers and their predictive value for developing chronic thromboembolic pulmonary hypertension (CTEPH) after acute pulmonary embolism.  CTEPH is categorized as group 4 disorders in the updated clinical classification of pulmonary hypertension (P-HTN).  CTEPH is a treatable form of P-HTN, typically with pulmonary thromboendarterectomy, however early diagnosis is imperative to success in treatment as delay worsens prognosis and perioperative mortality rate.  This study was performed retrospectively to evaluate patients that had developed CTEPH and re-examined initial presenting Chest CT to assess if markers could be determined to predict eventual CTEPH development.  Exclusion criteria included patients with signs of chronic PE at baseline or other significant lung disease which could contribute to P-HTN. Subjective grading of thrombus was examined in the main pulmonary artery, left and right main pulmonary arteries, as well as all 5 interlobar branches. Branches were graded as no obstruction, incomplete obstruction, and complete obstruction.  Additional markers evaluated include: Walsh score (quantitative measurement of abnormalities in segmental PA), right ventricle diameter (RVD), left ventricle diameter (LVD), RVD-LVD ratio, RA diameter, PA diameter, right heart thrombus, pericardial effusion, lung infarction, and mosaic attenuation. Univariable analysis of these markers was then performed and multiple factors were shown to increase probability of CTEPH development.  Occlusive central or lobar clot on initial imaging, Walsh score, lung infarction, and mosaic attenuation were all found to be associated with future CTEPH development (P <0.001). Pericardial effusion (P <0.017) was also found to be associated with CTEPH development.    Additionally, other signs such as RVD-LVD ratio, diameter of the PA or RA, and right heart thrombus had no association with development of future CTEPH after acute PE.

The limitations of the study include a retrospective design and possible selection bias due to exclusion of certain patients from the sample.  Patients included in the study were from the PE response team that have a submassive/massive PE that may result in potential bias.  Ventilation-perfusion scans or additional functional tests were not evaluated that have better sensitivity to evaluate perfusion defects as compared to CT.

Overall, this study shows that CT-based biomarkers can play an important role in identifying patients at higher risk for development of CTEPH after an acute PE.

References
  • Share