Diagnostic Assessment of Deep Learning Algorithms for Frozen Tissue Section Analysis in Women with Breast Cancer.

PURPOSE: Assessing the metastasis status of the sentinel lymph nodes (SLNs) for hematoxylin and eosin-stained frozen tissue sections by pathologists is an essential but tedious and time-consuming task that contributes to accurate breast cancer staging. This study aimed to review a challenge competition (HeLP 2019) for the development of automated solutions for classifying the metastasis status of breast cancer patients. Materials and Methods: A total of 524 digital slides were obtained from frozen SLN sections: 297 (56.7%) from Asan Medical Center (AMC) and 227 (43.4%) from Seoul National University Bundang Hospital (SNUBH), South Korea. The slides were divided into training, development, and validation sets, where the development set comprised slides from both institutions and training and validation set included slides from only AMC and SNUBH, respectively. The algorithms were assessed for area under the receiver operating characteristic curve (AUC) and measurement of the longest metastatic tumor diameter. The final total scores were calculated as the mean of the two metrics, and the three teams with AUC values greater than 0.500 were selected for review and analysis in this study. RESULTS: The top three teams showed AUC values of 0.891, 0.809, and 0.736 and major axis prediction scores of 0.525, 0.459, and 0.387 for the validation set. The major factor that lowered the diagnostic accuracy was micro-metastasis. CONCLUSION: In this challenge competition, accurate deep learning algorithms were developed that can be helpful for making a diagnosis on intraoperative SLN biopsy. The clinical utility of this approach was evaluated by including an external validation set from SNUBH.

[Beyond Coronary CT Angiography: CT Fractional Flow Reserve and Perfusion]. / ì „ì‚°í™”ë‹¨ì¸µì´¬ì˜ ê´€ìƒë™ë§¥ì¡°ì˜ìˆ : ë¶„íší˜ˆë¥˜ì˜ˆë¹„ë ¥ê³¼ 심근관류 영상.

Cardiac CT has been proven to provide diagnostic and prognostic evaluation of coronary artery disease for cardiovascular risk stratification and treatment decision-making based on rapid technological development and various research evidence. Coronary CT angiography has emerged as a gateway test for coronary artery disease that can reduce invasive angiography due to its high negative predictive value, but the diagnostic specificity is relatively low. However, coronary CT angiography is likely to overcome its limitations through functional evaluation to identify the hemodynamic significance of coronary artery disease by analyzing myocardial perfusion and fractional flow reserve through cardiac CT. Recently, studies have been actively conducted to incorporate artificial intelligence to make this more objective and reproducible. In this review, functional imaging techniques of cardiac computerized tomography are explored.

In-vitro and In-Vivo Assessment of 4D Flow MRI Reynolds Stress Mapping for Pulsatile Blood Flow.

Imaging hemodynamics play an important role in the diagnosis of abnormal blood flow due to vascular and valvular diseases as well as in monitoring the recovery of normal blood flow after surgical or interventional treatment. Recently, characterization of turbulent blood flow using 4D flow magnetic resonance imaging (MRI) has been demonstrated by utilizing the changes in signal magnitude depending on intravoxel spin distribution. The imaging sequence was extended with a six-directional icosahedral (ICOSA6) flow-encoding to characterize all elements of the Reynolds stress tensor (RST) in turbulent blood flow. In the present study, we aimed to demonstrate the feasibility of full RST analysis using ICOSA6 4D flow MRI under physiological conditions. First, the turbulence analysis was performed through in vitro experiments with a physiological pulsatile flow condition. Second, a total of 12 normal subjects and one patient with severe aortic stenosis were analyzed using the same sequence. The in-vitro study showed that total turbulent kinetic energy (TKE) was less affected by the signal-to-noise ratio (SNR), however, maximum principal turbulence shear stress (MPTSS) and total turbulence production (TP) had a noise-induced bias. Smaller degree of the bias was observed for TP compared to MPTSS. In-vivo study showed that the subject-variability on turbulence quantification was relatively low for the consistent scan protocol. The in vivo demonstration of the stenosis patient showed that the turbulence analysis could clearly distinguish the difference in all turbulence parameters as they were at least an order of magnitude larger than those from the normal subjects.

Challenge for Diagnostic Assessment of Deep Learning Algorithm for Metastases Classification in Sentinel Lymph Nodes on Frozen Tissue Section Digital Slides in Women with Breast Cancer.

PURPOSE: Assessing the status of metastasis in sentinel lymph nodes (SLNs) by pathologists is an essential task for the accurate staging of breast cancer. However, histopathological evaluation of SLNs by a pathologist is not easy and is a tedious and time-consuming task. The purpose of this study is to review a challenge competition (HeLP 2018) to develop automated solutions for the classification of metastases in hematoxylin and eosin-stained frozen tissue sections of SLNs in breast cancer patients. MATERIALS AND METHODS: A total of 297 digital slides were obtained from frozen SLN sections, which include post-neoadjuvant cases (n = 144, 48.5%) in Asan Medical Center, South Korea. The slides were divided into training, development, and validation sets. All of the imaging datasets have been manually segmented by expert pathologists. A total of 10 participants were allowed to use the Kakao challenge platform for six weeks with two P40 GPUs. The algorithms were assessed in terms of the AUC (area under receiver operating characteristic curve). RESULTS: The top three teams showed 0.986, 0.985, and 0.945 AUCs for the development set and 0.805, 0.776, and 0.765 AUCs for the validation set. Micrometastatic tumors, neoadjuvant systemic therapy, invasive lobular carcinoma, and histologic grade 3 were associated with lower diagnostic accuracy. CONCLUSION: In a challenge competition, accurate deep learning algorithms have been developed, which can be helpful in making frozen diagnosis of intraoperative SLN biopsy. Whether this approach has clinical utility will require evaluation in a clinical setting.

Reference parameters for left ventricular wall thickness, thickening, and motion in stress myocardial perfusion CT: Global and regional assessment.

PURPOSE: To use adenosine-induced stress CT myocardial perfusion imaging (CT-MPI) to determine normal reference values for left ventricle (LV) wall thickness (WT) and motion parameters. METHODS: This study included 106 Korean subjects (52 men and 54 women) who underwent CT-MPI due to chest pain, but were not found to have any detectable mild or severe coronary artery disease or myocardial perfusion defect. The following quantitative parameters were assessed on the CT-MPI according to a 17-segment model: LV myocardial thickness at end-systolic (WTES) and end-diastolic (WTED) phases, systolic wall thickening (SWT), and wall motion (WM). The associations of the measured parameters with the subjects' demographic characteristics and comorbidities were also analyzed. RESULTS: Septal wall (7.2 mm) and basal-level (7.7 mm) LV myocardium demonstrated significantly higher WT (p < 0.001). SWT was highest in lateral (77.8%, p < 0.014) and apical (78.9%, p = 0.009) myocardium, while lateral (7.7 mm) and basal (6.7 mm) myocardium exhibited the greatest WM (p < 0.001). WT was significantly higher in men and younger (<60 years) subjects (all, p < 0.001). Hypertensive individuals presented with significantly higher SWT (79.9%, p = 0.024). LV WT exhibited statistically significant correlations (all positive, except for age) with age, height, weight, body surface area, body mass index, and systolic blood pressure (all, p < 0.010). CONCLUSIONS: The present study provides CT-MPI reference values for LV myocardial WT, SWT, and WM measured on an adult Korean population. Knowledge of such normal reference measurements would be beneficial for the efficient interpretation of CT-MPI examinations in populations of Asian ethnicity.

Ischemic burden assessment of myocardial perfusion CT, compared with SPECT using semi-quantitative and quantitative approaches.

BACKGROUND: We aimed to compare the myocardial ischemic burden assessed using semi-quantitative and quantitative analysis of computed tomography-myocardial perfusion imaging (CT-MPI) with that of single-photon emission computed tomography (SPECT). METHODS: From 2011 to 2013, 97 patients who underwent CT-MPI and SPECT were evaluated. The extent and severity of perfusion defects were assessed on a 5-point scale using a standard 16-segment model, and were expressed as summed stress score (SSS) and summed difference score (SDS). Receiver operating characteristic (ROC) curves for quantitative parameters were generated for the diagnosis of abnormal perfusion defect (SSS ≥ 4) and presence of ischemia (SDS ≥ 2) on SPECT. RESULTS: On CT-MPI, 298 (19.2%) of the 1552 segments showed perfusion abnormalities during stress, whereas perfusion abnormalities were shown in 179 (11.5%) segments on SPECT-MPI. On a per-person basis, there was good agreement, with intraclass correlation coefficients of 0.78 for SSS and 0.72 for SDS. A significant reduction of attenuation in stress and myocardial perfusion reserve index, along with an increase in % defect volume of CT-MPI, were demonstrated as the degree of perfusion defect or ischemia on SPECT increased. On the ROC curves, % defect volume on CT-MPI demonstrated the highest area under the curve: 0.91 for abnormal perfusion defect and 0.89 (all p < 0.001) for the presence of ischemia on SPECT. CONCLUSIONS: Semi-quantitative analysis of CT-MPI showed good accordance with SPECT. A quantitative approach for CT-MPI, especially % defect volume, may provide additional value in the identification of myocardial perfusion abnormalities. CLINICAL TRIAL REGISTRATION: https://clinicaltrials.gov/ct2/show/NCT01696006.

Machine learning assessment of myocardial ischemia using angiography: Development and retrospective validation.

BACKGROUND: Invasive fractional flow reserve (FFR) is a standard tool for identifying ischemia-producing coronary stenosis. However, in clinical practice, over 70% of treatment decisions still rely on visual estimation of angiographic stenosis, which has limited accuracy (about 60%-65%) for the prediction of FFR < 0.80. One of the reasons for the visual-functional mismatch is that myocardial ischemia can be affected by the supplied myocardial size, which is not always evident by coronary angiography. The aims of this study were to develop an angiography-based machine learning (ML) algorithm for predicting the supplied myocardial volume for a stenosis, as measured using coronary computed tomography angiography (CCTA), and then to build an angiography-based classifier for the lesions with an FFR < 0.80 versus ≥ 0.80. METHODS AND FINDINGS: A retrospective study was conducted using data from 1,132 stable and unstable angina patients with 1,132 intermediate lesions who underwent invasive coronary angiography, FFR, and CCTA at the Asan Medical Center, Seoul, Korea, between 1 May 2012 and 30 November 2015. The mean age was 63 ± 10 years, 76% were men, and 72% of the patients presented with stable angina. Of these, 932 patients (assessed before 31 January 2015) constituted the training set for the algorithm, and 200 patients (assessed after 1 February 2015) served as a test cohort to validate its diagnostic performance. Additionally, external validation with 79 patients from two centers (CHA University, Seongnam, Korea, and Ajou University, Suwon, Korea) was conducted. After automatic contour calibration using the caliber of guiding catheter, quantitative coronary angiography was performed using the edge-detection algorithms (CAAS-5, Pie-Medical). Clinical information was provided by the Asan BiomedicaL Research Environment (ABLE) system. The CCTA-based myocardial segmentation (CAMS)-derived myocardial volume supplied by each vessel (right coronary artery [RCA], left anterior descending [LAD], left circumflex [LCX]) and the myocardial volume subtended to a stenotic segment (CAMS-%Vsub) were measured for labeling. The ML for (1) predicting vessel territories (CAMS-%LAD, CAMS-%LCX, and CAMS-%RCA) and CAMS-%Vsub and (2) identifying the lesions with an FFR < 0.80 was constructed. Angiography-based ML, employing a light gradient boosting machine (GBM), showed mean absolute errors (MAEs) of 5.42%, 8.57%, and 4.54% for predicting CAMS-%LAD, CAMS-%LCX, and CAMS-%RCA, respectively. The percent myocardial volumes predicted by ML were used to predict the CAMS-%Vsub. With 5-fold cross validation, the MAEs between ML-predicted percent myocardial volume subtended to a stenotic segment (ML-%Vsub) and CAMS-%Vsub were minimized by the elastic net (6.26% ± 0.55% for LAD, 5.79% ± 0.68% for LCX, and 2.95% ± 0.14% for RCA lesions). Using all attributes (age, sex, involved vessel segment, and angiographic features affecting the myocardial territory and stenosis degree), the ML classifiers (L2 penalized logistic regression, support vector machine, and random forest) predicted an FFR < 0.80 with an accuracy of approximately 80% (area under the curve [AUC] = 0.84-0.87, 95% confidence intervals 0.71-0.94) in the test set, which was greater than that of diameter stenosis (DS) > 53% (66%, AUC = 0.71, 95% confidence intervals 0.65-0.78). The external validation showed 84% accuracy (AUC = 0.89, 95% confidence intervals 0.83-0.95). The retrospective design, single ethnicity, and the lack of clinical outcomes may limit this prediction model's generalized application. CONCLUSION: We found that angiography-based ML is useful to predict subtended myocardial territories and ischemia-producing lesions by mitigating the visual-functional mismatch between angiographic and FFR. Assessment of clinical utility requires further validation in a large, prospective cohort study.

Impact of a Geometric Correction for Proximal Flow Constraint on the Assessment of Mitral Regurgitation Severity Using the Proximal Flow Convergence Method.

BACKGROUND: Overestimation of the severity of mitral regurgitation (MR) by the proximal isovelocity surface area (PISA) method has been reported. We sought to test whether angle correction (AC) of the constrained flow field is helpful to eliminate overestimation in patients with eccentric MR. METHODS: In a total of 33 patients with MR due to prolapse or flail mitral valve, both echocardiography and cardiac magnetic resonance image (CMR) were performed to calculate regurgitant volume (RV). In addition to RV by conventional PISA (RVPISA), convergence angle (α) was measured from 2-dimensional Doppler color flow maps and RV was corrected by multiplying by α/180 (RVAC). RV measured by CMR (RVCMR) was used as a gold standard, which was calculated by the difference between total stroke volume measured by planimetry of the short axis slices and aortic stroke volume by phase-contrast image. RESULTS: The correlation between RVCMR and RV by echocardiography was modest [RVCMR vs. RVPISA (r = 0.712, p < 0.001) and RVCMR vs. RVAC (r = 0.766, p < 0.001)]. However, RVPISA showed significant overestimation (RVPISA - RVCMR = 50.6 ± 40.6 mL vs. RVAC - RVCMR = 7.7 ± 23.4 mL, p < 0.001). The overall accuracy of RVPISA for diagnosis of severe MR, defined as RV ≥ 60 mL, was 57.6% (19/33), whereas it increased to 84.8% (28/33) by using RVAC (p = 0.028). CONCLUSION: Conventional PISA method tends to provide falsely large RV in patients with eccentric MR and a simple geometric AC of the proximal constraint flow largely eliminates overestimation.

In vivo assessment of aortic root geometry in normal controls using 3D analysis of computed tomography.

AIMS: Understanding normal asymmetry in the aortic root could aid in the development of new surgical repair techniques or devices with improved haemodynamic performance. The purpose of this study was to assess geometric asymmetry and age-related changes in the normal aortic root using 3D computed tomography. METHODS AND RESULTS: The institutional review board approved this retrospective study of 130 normal subjects (mean age, 51.4 years; 58 men). Specialized 3D software measured individual cusp sinus volumes (CSVs), cusp surface areas (CSAs), and intercommissural distances (ICDs). Age-related aortic root changes were evaluated with simple correlation, ANOVA test among age groups, and multivariable linear regression analyses. The CSV and CSA of left coronary cusp (LCC) were significantly smaller than those of right coronary cusp (RCC) and non-coronary cusp (NCC) (both, P < 0.001) in all age groups. The mean ratios of RCC or NCC-to-LCC were 1.38 and 1.36 for CSV, 1.19 and 1.20 for CSA, and 1.21 and 1.06 for ICD, respectively. The CSV and ICD increased in older age with weak-to-moderate correlation coefficients in both men and women. By multivariable linear regression, CSVs and ICDs of all cusps showed a positive correlation with age (P < 0.05), and the female gender was associated with a smaller size of the CSV and CSA. CONCLUSIONS: The LCC was significantly smaller than the other two cusps, and the aortic root size increased with age.

Association between C-reactive Protein and type of coronary arterial plaque in asymptomatic patients: assessment with coronary CT angiography.

PURPOSE: To determine whether C-reactive protein is associated with the type of coronary plaque seen at computed tomographic (CT) angiography. MATERIALS AND METHODS: The institutional review board approved this retrospective study, and the need for informed consent was waived. C-reactive protein levels were measured in 2653 asymptomatic subjects (mean age ± standard deviation, 54.7 years ± 9.2; 1811 men) who underwent self-referred coronary CT angiography as part of a general health checkup. The presence of coronary plaque, plaque type (calcified, mixed calcified, or noncalcified), stenosis degree, and number of involved segments were evaluated. Subjects with one type of plaque (calcified plaque, mixed plaque, and noncalcified plaque groups) and two or more types of plaque (multiple lesions group) were analyzed separately. Multivariate logistic regression analysis was used to evaluate the association between increasing C-reactive protein levels and plaque type. RESULTS: Coronary plaque was found in 1150 of the 2653 subjects (43.3%): calcified plaque (n = 604, 22.8%), mixed plaque (n = 67, 2.5%), noncalcified plaque (n = 208, 7.8%), and multiple lesions (n = 271, 10.2%). The C-reactive protein cutoff value of the fourth quartile was 1.2 mg/L (11.4 nmol/L), and all types of coronary plaque were increased in the higher quartile of the C-reactive protein levels. Multivariate logistic regression analysis showed that a higher C-reactive protein level was an independent predictor for the presence of noncalcified plaque (fourth vs first quartile group, odds ratio = 1.70, P = .025) and significant (50% and higher) coronary stenosis (odds ratio = 1.76, P = .020) after adjustment for traditional risk factors for coronary artery disease. CONCLUSION: C-reactive protein is associated with noncalcified coronary arterial plaque, as seen at coronary CT angiography in asymptomatic patients after adjustment for traditional risk factors.

False-negative results at multi-detector row CT colonography: multivariate analysis of causes for missed lesions.

PURPOSE: To determine causes of false-negative results at multi-detector row computed tomographic (CT) colonography and determine presumptive causes with logistic regression analysis. MATERIALS AND METHODS: Institutional review board approval and informed consent were obtained. The study included 394 colonic segments in 31 men and 25 women at high risk for colorectal cancer (mean age +/- standard deviation, 60.2 years +/- 9.3 for men and 56.8 years +/- 13.3 for women). Multi-detector row CT colonography and colonoscopy (reference standard) were performed in a blinded manner, and the results were compared. CT colonographic findings were interpreted in consensus by two readers using a primary two-dimensional with three-dimensional problem-solving approach. Adequacy of colonic preparation and distention was recorded. Sensitivity and specificity were obtained with 95% confidence intervals (CIs). Lesions missed at CT colonography were retrospectively reassessed to identify why they were missed, and, if the causes were not apparent, logistic regression analysis was performed to determine the presumptive causes. RESULTS: Colonic preparation and distention were optimal in 17 patients (30%) but suboptimal in 37 (66%) and poor enough to make the results nondiagnostic in two (4%). Twenty-nine of 63 lesions were missed at CT colonography. When all flat, sessile, and pedunculated lesions (n = 63) were included, sensitivities were 75% (nine of 12; 95% CI: 48%, 100%) for lesions 10 mm or larger and 79% (19 of 24; 95% CI: 65%, 93%) for those 6 mm or larger. When only sessile and pedunculated lesions (n = 60) were included, corresponding sensitivities were 100% (nine of nine; 73%, 100%) and 90% (19 of 21; 78%, 100%), respectively. All three missed lesions larger than 10 mm were flat, and all three flat lesions were missed. Two 3-mm high lesions, including one invasive adenocarcinoma, were misinterpreted as feces at blinded image review; one 1-mm high tubular adenoma with adenocarcinoma foci could not be visualized even in retrospect. Sessile or pedunculated polyps 5 mm or smaller were significantly more likely to be missed than those 6 mm or larger (adjusted odds ratio, 11.6; P = .027). CONCLUSION: Aside from inadequate bowel preparation and/or distention, flat lesions and small polyps are the two main causes for missed lesions at multi-detector row CT colonography.