AI analytics can be used as imaging biomarkers for predicting invasive upgrade of ductal carcinoma in situ

OBJECTIVES

To evaluate whether the quantitative abnormality scores provided by artificial intelligence (AI)-based computer-aided detection/diagnosis (CAD) for mammography interpretation can be used to predict invasive upgrade in ductal carcinoma in situ (DCIS) diagnosed on percutaneous biopsy.

METHODS

Four hundred forty DCIS in 420 women (mean age, 52.8 years) diagnosed via percutaneous biopsy from January 2015 to December 2019 were included. Mammographic characteristics were assessed based on imaging features (mammographically occult, mass/asymmetry/distortion, calcifications only, and combined mass/asymmetry/distortion with calcifications) and BI-RADS assessments. Routine pre-biopsy 4-view digital mammograms were analyzed using AI-CAD to obtain abnormality scores (AI-CAD score, ranging 0-100%). Multivariable logistic regression was performed to identify independent predictive mammographic variables after adjusting for clinicopathological variables. A subgroup analysis was performed with mammographically detected DCIS.

RESULTS

Of the 440 DCIS, 117 (26.6%) were upgraded to invasive cancer. Three hundred forty-one (77.5%) DCIS were detected on mammography. The multivariable analysis showed that combined features (odds ratio (OR): 2.225, p = 0.033), BI-RADS 4c or 5 assessments (OR: 2.473, p = 0.023 and OR: 5.190, p < 0.001, respectively), higher AI-CAD score (OR: 1.009, p = 0.007), AI-CAD score ≥ 50% (OR: 1.960, p = 0.017), and AI-CAD score ≥ 75% (OR: 2.306, p = 0.009) were independent predictors of invasive upgrade. In mammographically detected DCIS, combined features (OR: 2.194, p = 0.035), and higher AI-CAD score (OR: 1.008, p = 0.047) were significant predictors of invasive upgrade.

CONCLUSION

The AI-CAD score was an independent predictor of invasive upgrade for DCIS. Higher AI-CAD scores, especially in the highest quartile of ≥ 75%, can be used as an objective imaging biomarker to predict invasive upgrade in DCIS diagnosed with percutaneous biopsy.

CRITICAL RELEVANCE STATEMENT

Noninvasive imaging features including the quantitative results of AI-CAD for mammography interpretation were independent predictors of invasive upgrade in lesions initially diagnosed as ductal carcinoma in situ via percutaneous biopsy and therefore may help decide the direction of surgery before treatment.

KEY POINTS

• Predicting ductal carcinoma in situ upgrade is important, yet there is a lack of conclusive non-invasive biomarkers. • AI-CAD scores-raw numbers, ≥ 50%, and ≥ 75%-predicted ductal carcinoma in situ upgrade independently. • Quantitative AI-CAD results may help predict ductal carcinoma in situ upgrade and guide patient management.

Unenhanced Breast MRI With Diffusion-Weighted Imaging for Breast Cancer Detection: Effects of Training on Performance and Agreement of Subspecialty Radiologists

OBJECTIVE

To investigate whether reader training improves the performance and agreement of radiologists in interpreting unenhanced breast magnetic resonance imaging (MRI) scans using diffusion-weighted imaging (DWI).

MATERIALS AND METHODS

A study of 96 breasts (35 cancers, 24 benign, and 37 negative) in 48 asymptomatic women was performed between June 2019 and October 2020. High-resolution DWI with b-values of 0, 800, and 1200 sec/mm² was performed using a 3.0-T system. Sixteen breast radiologists independently reviewed the DWI, apparent diffusion coefficient maps, and T1-weighted MRI scans and recorded the Breast Imaging Reporting and Data System (BI-RADS) category for each breast. After a 2-h training session and a 5-month washout period, they re-evaluated the BI-RADS categories. A BI-RADS category of 4 (lesions with at least two suspicious criteria) or 5 (more than two suspicious criteria) was considered positive. The per-breast diagnostic performance of each reader was compared between the first and second reviews. Inter-reader agreement was evaluated using a multi-rater κ analysis and intraclass correlation coefficient (ICC).

RESULTS

Before training, the mean sensitivity, specificity, and accuracy of the 16 readers were 70.7% (95% confidence interval [CI]: 59.4-79.9), 90.8% (95% CI: 85.6-94.2), and 83.5% (95% CI: 78.6-87.4), respectively. After training, significant improvements in specificity (95.2%; 95% CI: 90.8-97.5; P = 0.001) and accuracy (85.9%; 95% CI: 80.9-89.8; P = 0.01) were observed, but no difference in sensitivity (69.8%; 95% CI: 58.1-79.4; P = 0.58) was observed. Regarding inter-reader agreement, the κ values were 0.57 (95% CI: 0.52-0.63) before training and 0.68 (95% CI: 0.62-0.74) after training, with a difference of 0.11 (95% CI: 0.02-0.18; P = 0.01). The ICC was 0.73 (95% CI: 0.69-0.74) before training and 0.79 (95% CI: 0.76-0.80) after training (P = 0.002).

CONCLUSION

Brief reader training improved the performance and agreement of interpretations by breast radiologists using unenhanced MRI with DWI.

Learnability of Thyroid Nodule Assessment on Ultrasonography: Using a Big Data Set

OBJECTIVE

The aims of the work described here were to evaluate the learnability of thyroid nodule assessment on ultrasonography (US) using a big data set of US images and to evaluate the diagnostic utilities of artificial intelligence computer-aided diagnosis (AI-CAD) used by readers with varying experience to differentiate benign and malignant thyroid nodules.

METHODS

Six college freshmen independently studied the "learning set" composed of images of 13,560 thyroid nodules, and their diagnostic performance was evaluated after their daily learning sessions using the "test set" composed of images of 282 thyroid nodules. The diagnostic performance of two residents and an experienced radiologist was evaluated using the same "test set." After an initial diagnosis, all readers once again evaluated the "test set" with the assistance of AI-CAD.

RESULTS

Diagnostic performance of almost all students increased after the learning program. Although the mean areas under the receiver operating characteristic curves (AUROCs) of residents and the experienced radiologist were significantly higher than those of students, the AUROCs of five of the six students did not differ significantly compared with that of the one resident. With the assistance of AI-CAD, sensitivity significantly increased in three students, specificity in one student, accuracy in four students and AUROC in four students. Diagnostic performance of the two residents and the experienced radiologist was better with the assistance of AI-CAD.

CONCLUSION

A self-learning method using a big data set of US images has potential as an ancillary tool alongside traditional training methods. With the assistance of AI-CAD, the diagnostic performance of readers with varying experience in thyroid imaging could be further improved.

Machine learning‑based radiomics models for prediction of locoregional recurrence in patients with breast cancer

Locoregional recurrence (LRR) is the predominant pattern of relapse after definitive breast cancer treatment. The present study aimed to develop machine learning (ML)-based radiomics models to predict LRR in patients with breast cancer by using preoperative magnetic resonance imaging (MRI) data. Data from patients with localized breast cancer that underwent preoperative MRI between January 2013 and December 2017 were collected. Propensity score matching (PSM) was performed to adjust for clinical factors between patients with and without LRR. Radiomics features were obtained from T2-weighted with and without fat-suppressed MRI and contrast-enhanced T1-weighted with fat-suppressed MRI. In the present study five ML models were designed, three base models (support vector machine, random forest, and logistic regression) and two ensemble models (voting model and stacking model) composed of the three base models, and the performance of each base model was compared with the stacking model. After PSM, 28 patients with LRR and 86 patients without LRR were included. Of these 114 patients, 80 patients were randomly selected to train the models, and the remaining 34 patients were used to evaluate the performance of the trained models. In total, 5,064 features were obtained from each patient, and 47-51 features were selected by applying variance threshold and least absolute shrinkage and selection operator. The stacking model demonstrated superior performance in area under the receiver operating characteristic curve (AUC), with an AUC of 0.78 compared to a range of 0.61 to 0.70 for the other models. An external validation study to investigate the efficacy of the stacking model of the present study was initiated and is still ongoing (Korean Radiation Oncology Group 2206).

Corrigendum: Diffusion-Weighted Magnetic Resonance Imaging for Preoperative Evaluation of Patients With Breast Cancer: Protocol of a Prospective, Multicenter, Observational Cohort Study

This corrects the article on p. 292 in vol. 26, PMID: 37272245.

Diffusion-Weighted Magnetic Resonance Imaging for Preoperative Evaluation of Patients With Breast Cancer: Protocol of a Prospective, Multicenter, Observational Cohort Study

PURPOSE

Detection of multifocal, multicentric, and contralateral breast cancers in patients affects surgical management. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) can identify additional foci that were initially undetected by conventional imaging. However, its use is limited owing to low specificity and high false-positive rate. Multiparametric MRI (DCE-MRI + diffusion-weighted [DW] MRI) can increase the specificity. We aimed to describe the protocols of our prospective, multicenter, observational cohort studies designed to compare the diagnostic performance of DCE-MRI and multiparametric MRI for the diagnosis of multifocal, multicentric cancer and contralateral breast cancer in patients with newly diagnosed breast cancer.

METHODS

Two studies comparing the performance of DCE-MRI and multiparametric MRI for the diagnosis of multifocal, multicentric cancer (NCT04656639) and contralateral breast cancer (NCT05307757) will be conducted. For trial NCT04656639, 580 females with invasive breast cancer candidates for breast conservation surgery whose DCE-MRI showed additional suspicious lesions (breast imaging reporting and data system [BI-RADS] category ≥ 4) on DCE-MRI in the ipsilateral breast will be enrolled. For trial NCT05307757, 1098 females with invasive breast cancer whose DCE-MRI showed contralateral lesions (BI-RADS category ≥ 3 or higher on DCE-MRI) will be enrolled. Participants will undergo 3.0-T DCE-MRI and DW-MRI. The diagnostic performance of DCE-MRI and multiparametric MRI will be compared. The receiver operating characteristic curve, sensitivity, specificity, positive predictive value, and characteristics of the detected cancers will be analyzed. The primary outcome is the difference in the receiver operating characteristic curve between DCE-MRI and multiparametric MRI interpretation. Enrollment completion is expected in 2024, and study results are expected to be presented in 2026.

DISCUSSION

This prospective, multicenter study will compare the performance of DCE-MRI versus multiparametric MRI for the preoperative evaluation of multifocal, multicentric, and contralateral breast cancer and is currently in the patient enrollment phase.

TRIAL REGISTRATION

ClinicalTrials.gov Identifier: NCT04656639, NCT05307757. Registered on April 1 2022.

Diagnosis of thyroid micronodules on ultrasound using a deep convolutional neural network

To assess the performance of deep convolutional neural network (CNN) to discriminate malignant and benign thyroid nodules < 10 mm in size and compare the diagnostic performance of CNN with those of radiologists. Computer-aided diagnosis was implemented with CNN and trained using ultrasound (US) images of 13,560 nodules ≥ 10 mm in size. Between March 2016 and February 2018, US images of nodules < 10 mm were retrospectively collected at the same institution. All nodules were confirmed as malignant or benign from aspirate cytology or surgical histology. Diagnostic performances of CNN and radiologists were assessed and compared for area under curve (AUC), sensitivity, specificity, accuracy, positive predictive value, and negative predictive value. Subgroup analyses were performed based on nodule size with a cut-off value of 5 mm. Categorization performances of CNN and radiologists were also compared. A total of 370 nodules from 362 consecutive patients were assessed. CNN showed higher negative predictive value (35.3% vs. 22.6%, P = 0.048) and AUC (0.66 vs. 0.57, P = 0.04) than radiologists. CNN also showed better categorization performance than radiologists. In the subgroup of nodules ≤ 5 mm, CNN showed higher AUC (0.63 vs. 0.51, P = 0.08) and specificity (68.2% vs. 9.1%, P < 0.001) than radiologists. Convolutional neural network trained with thyroid nodules ≥ 10 mm in size showed overall better diagnostic performance than radiologists in the diagnosis and categorization of thyroid nodules < 10 mm, especially in nodules ≤ 5 mm.

Changes in Automated Mammographic Breast Density Can Predict Pathological Response After Neoadjuvant Chemotherapy in Breast Cancer

OBJECTIVE

Mammographic density is an independent risk factor for breast cancer that can change after neoadjuvant chemotherapy (NCT). This study aimed to evaluate percent changes in volumetric breast density (ΔVbd%) before and after NCT measured automatically and determine its value as a predictive marker of pathological response to NCT.

MATERIALS AND METHODS

A total of 357 patients with breast cancer treated between January 2014 and December 2016 were included. An automated volumetric breast density (Vbd) measurement method was used to calculate Vbd on mammography before and after NCT. Patients were divided into three groups according to ΔVbd%, calculated as follows: Vbd (post-NCT - pre-NCT)/pre-NCT Vbd × 100 (%). The stable, decreased, and increased groups were defined as -20% ≤ ΔVbd% ≤ 20%, ΔVbd% < -20%, and ΔVbd% > 20%, respectively. Pathological complete response (pCR) was considered to be achieved after NCT if there was no evidence of invasive carcinoma in the breast or metastatic tumors in the axillary and regional lymph nodes on surgical pathology. The association between ΔVbd% grouping and pCR was analyzed using univariable and multivariable logistic regression analyses.

RESULTS

The interval between the pre-NCT and post-NCT mammograms ranged from 79 to 250 days (median, 170 days). In the multivariable analysis, ΔVbd% grouping (odds ratio for pCR of 0.420 [95% confidence interval, 0.195-0.905; P = 0.027] for the decreased group compared with the stable group), N stage at diagnosis, histologic grade, and breast cancer subtype were significantly associated with pCR. This tendency was more evident in the luminal B-like and triple-negative subtypes.

CONCLUSION

ΔVbd% was associated with pCR in breast cancer after NCT, with the decreased group showing a lower rate of pCR than the stable group. Automated measurement of ΔVbd% may help predict the NCT response and prognosis in breast cancer.

Radiomics Analysis of Gray-Scale Ultrasonographic Images of Papillary Thyroid Carcinoma > 1 cm: Potential Biomarker for the Prediction of Lymph Node Metastasis

Purpose

This study aimed to investigate radiomics analysis of ultrasonographic images to develop a potential biomarker for predicting lymph node metastasis in papillary thyroid carcinoma (PTC) patients.

Materials and Methods

This study included 431 PTC patients from August 2013 to May 2014 and classified them into the training and validation sets. A total of 730 radiomics features, including texture matrices of gray-level co-occurrence matrix and gray-level run-length matrix and single-level discrete two-dimensional wavelet transform and other functions, were obtained. The least absolute shrinkage and selection operator method was used for selecting the most predictive features in the training data set.

Results

Lymph node metastasis was associated with the radiomics score (p < 0.001). It was also associated with other clinical variables such as young age (p = 0.007) and large tumor size (p = 0.007). The area under the receiver operating characteristic curve was 0.687 (95% confidence interval: 0.616-0.759) for the training set and 0.650 (95% confidence interval: 0.575-0.726) for the validation set.

Conclusion

This study showed the potential of ultrasonography-based radiomics to predict cervical lymph node metastasis in patients with PTC; thus, ultrasonography-based radiomics can act as a biomarker for PTC.

Application of Artificial Intelligence Computer-Assisted Diagnosis Originally Developed for Thyroid Nodules to Breast Lesions on Ultrasound

As thyroid and breast cancer have several US findings in common, we applied an artificial intelligence computer-assisted diagnosis (AI-CAD) software originally developed for thyroid nodules to breast lesions on ultrasound (US) and evaluated its diagnostic performance. From January 2017 to December 2017, 1042 breast lesions (mean size 20.2 ± 11.8 mm) of 1001 patients (mean age 45.9 ± 12.9 years) who underwent US-guided core-needle biopsy were included. An AI-CAD software that was previously trained and validated with thyroid nodules using the convolutional neural network was applied to breast nodules. There were 665 benign breast lesions (63.0%) and 391 breast cancers (37.0%). The area under the receiver operating characteristic curve (AUROC) of AI-CAD to differentiate breast lesions was 0.678 (95% confidence interval: 0.649, 0.707). After fine-tuning AI-CAD with 1084 separate breast lesions, the diagnostic performance of AI-CAD markedly improved (AUC 0.841). This was significantly higher than that of radiologists when the cutoff category was BI-RADS 4a (AUC 0.621, P < 0.001), but lower when the cutoff category was BI-RADS 4b (AUC 0.908, P < 0.001). When applied to breast lesions, the diagnostic performance of an AI-CAD software that had been developed for differentiating malignant and benign thyroid nodules was not bad. However, an organ-specific approach guarantees better diagnostic performance despite the similar US features of thyroid and breast malignancies.

Sarcopenia increases the risk of major organ or vessel invasion in patients with papillary thyroid cancer

While sarcopenia is associated with poor overall survival and cancer-specific survival in solid cancer patients, the impact of sarcopenia on clinicopathologic features that can influence conventional papillary thyroid cancer (PTC) prognosis remains unclear. To investigate the impact of sarcopenia on aggressive clinicopathologic features in PTC patients, prospectively collected data on 305 patients who underwent surgery for PTC with preoperative staging ultrasonography and bioelectrical impedance analysis were retrospectively analyzed. Nine sarcopenia patients with preoperative sarcopenia showed more patients aged 55 or older (p = 0.022), higher male proportion (p < 0.001), lower body-mass index (p = 0.015), higher incidence of major organ or vessel invasion (p = 0.001), higher T stage (p = 0.002), higher TNM stage (p = 0.007), and more tumor recurrence (p = 0.023) compared to the non-sarcopenia patients. Unadjusted and adjusted logistic regression analyses showed that sarcopenia (odds ratio (OR) 9.936, 95% confidence interval (CI) 2.052–48.111, p = 0.004), tumor size (OR 1.048, 95% CI 1.005–1.093, p = 0.027), and tumor multiplicity (OR 3.323, 95% CI 1.048–10.534, p = 0.041) significantly increased the risk of T4 cancer. Sarcopenia patients showed significantly lower disease-free survival probability compared to non-sarcopenia patients. Therefore, preoperative sarcopenia in PTC patients should raise clinical suspicion for a more locally advanced disease and direct appropriate management and careful follow-up.

US, Mammography, and Histopathologic Evaluation to Identify Low Nuclear Grade Ductal Carcinoma in Situ

Background Low nuclear grade ductal carcinoma in situ (DCIS) identified at biopsy can be upgraded to intermediate to high nuclear grade DCIS at surgery. Methods that confirm low nuclear grade are needed to consider nonsurgical approaches for these patients. Purpose To develop a preoperative model to identify low nuclear grade DCIS and to evaluate factors associated with low nuclear grade DCIS at biopsy that was not upgraded to intermediate to high nuclear grade DCIS at surgery. Materials and Methods In this retrospective study, 470 women (median age, 50 years; interquartile range, 44-58 years) with 477 pure DCIS lesions at surgical histopathologic evaluation were included (January 2010 to December 2015). Patients were divided into the training set (n = 330) or validation set (n = 147) to develop a preoperative model to identify low nuclear grade DCIS. Features at US (mass, nonmass) and at mammography (morphologic characteristics, distribution of microcalcification) were reviewed. The upgrade rate of low nuclear grade DCIS was calculated, and multivariable regression was used to evaluate factors for associations with low nuclear grade DCIS that was not upgraded later. Results A preoperative model that included lesions manifesting as a mass at US without microcalcification and no comedonecrosis at biopsy was used to identify low nuclear grade DCIS, with a high area under the receiver operating characteristic curve of 0.97 (95% CI: 0.94, 1.00) in the validation set. The upgrade rate of low nuclear grade DCIS at biopsy was 38.8% (50 of 129). Ki-67 positivity (odds ratio, 0.04; 95% CI: 0.0003, 0.43; P = .005) was inversely associated with constant low nuclear grade DCIS. Conclusion The upgrade rate of low nuclear grade ductal carcinoma in situ (DCIS) at biopsy to intermediate to high nuclear grade DCIS at surgery occurred in more than a third of patients; low nuclear grade DCIS at final histopathologic evaluation could be identified if the mass was viewed at US without microcalcifications and had no comedonecrosis at histopathologic evaluation of biopsy. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Rahbar in this issue.

Automatic Detection and Segmentation of Breast Cancer on MRI Using Mask R-CNN Trained on Non-Fat-Sat Images and Tested on Fat-Sat Images

RATIONALE AND OBJECTIVES

Computer-aided methods have been widely applied to diagnose lesions on breast magnetic resonance imaging (MRI). The first step was to identify abnormal areas. A deep learning Mask Regional Convolutional Neural Network (R-CNN) was implemented to search the entire set of images and detect suspicious lesions.

MATERIALS AND METHODS

Two DCE-MRI datasets were used, 241 patients acquired using non-fat-sat sequence for training, and 98 patients acquired using fat-sat sequence for testing. All patients have confirmed unilateral mass cancers. The tumor was segmented using fuzzy c-means clustering algorithm to serve as the ground truth. Mask R-CNN was implemented with ResNet-101 as the backbone. The neural network output the bounding boxes and the segmented tumor for evaluation using the Dice Similarity Coefficient (DSC). The detection performance, and the trade-off between sensitivity and specificity, was analyzed using free response receiver operating characteristic.

RESULTS

When the precontrast and subtraction image of both breasts were used as input, the false positive from the heart and normal parenchymal enhancements could be minimized. The training set had 1469 positive slices (containing lesion) and 9135 negative slices. In 10-fold cross-validation, the mean accuracy = 0.86 and DSC = 0.82. The testing dataset had 1568 positive and 7264 negative slices, with accuracy = 0.75 and DSC = 0.79. When the obtained per-slice results were combined, 240 of 241 (99.5%) lesions in the training and 98 of 98 (100%) lesions in the testing datasets were identified.

CONCLUSION

Deep learning using Mask R-CNN provided a feasible method to search breast MRI, localize, and segment lesions. This may be integrated with other artificial intelligence algorithms to develop a fully automatic breast MRI diagnostic system.

Using ultrasonographic features to predict the outcomes of patients with small papillary thyroid carcinomas: a retrospective study implementing the 2015 ATA patterns and ACR TI-RADS categories

BACKGROUND

The aim of this study was to evaluate whether risk stratification systems using ultrasonographic (US) features show associations with the outcomes of patients with small papillary thyroid carcinomas (PTCs).

METHODS

This retrospective study received institutional review board approval. From March 2007 to February 2010, 775 patients who underwent surgery for small PTCs (10-20 mm) were included. Based on preoperative US features, PTCs were categorized according to the 2015 American Thyroid Association (ATA) guideline and the American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS). The associations of clinicopathological and US features with postoperative patient outcomes were evaluated.

RESULTS

In total, 61 patients had high-volume central lymph node metastasis (CLNM, 7.9%) and 100 patients had lateral lymph node metastasis (LLNM, 12.9%). In univariable analyses, a high number of suspicious US features and higher ACR TI-RADS point totals were significantly associated with both high-volume CLNM (P=0.001, each) and LLNM (P<0.001, each). In multivariable analyses of preoperative features, a higher number of suspicious US features and higher ACR TI-RADS point totals were independently associated with high-volume CLNM (odds ratio [OR], 1.516 and 1.201; P=0.002 and P=0.001, respectively) and LLNM (OR, 1.763 and 1.293; all P<0.001). Individual US features, ATA categories, and ACR TI-RADS point totals were not significantly associated with recurrence or distant metastasis.

CONCLUSION

The number of suspicious US features and the ACR TI-RADS point total are potential risk factors for cervical lymph node metastasis in patients with small PTCs.

Convolutional Neural Network to Stratify the Malignancy Risk of Thyroid Nodules: Diagnostic Performance Compared with the American College of Radiology Thyroid Imaging Reporting and Data System Implemented by Experienced Radiologists

BACKGROUND AND PURPOSE

Comparison of the diagnostic performance for thyroid cancer on ultrasound between a convolutional neural network and visual assessment by radiologists has been inconsistent. Thus, we aimed to evaluate the diagnostic performance of the convolutional neural network compared with the American College of Radiology Thyroid Imaging Reporting and Data System (TI-RADS) for the diagnosis of thyroid cancer using ultrasound images.

MATERIALS AND METHODS

From March 2019 to September 2019, seven hundred sixty thyroid nodules (≥10 mm) in 757 patients were diagnosed as benign or malignant through fine-needle aspiration, core needle biopsy, or an operation. Experienced radiologists assessed the sonographic descriptors of the nodules, and 1 of 5 American College of Radiology TI-RADS categories was assigned. The convolutional neural network provided malignancy risk percentages for nodules based on sonographic images. Sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were calculated with cutoff values using the Youden index and compared between the convolutional neural network and the American College of Radiology TI-RADS. Areas under the receiver operating characteristic curve were also compared.

RESULTS

Of 760 nodules, 176 (23.2%) were malignant. At an optimal threshold derived from the Youden index, sensitivity and negative predictive values were higher with the convolutional neural network than with the American College of Radiology TI-RADS (81.8% versus 73.9%, P = .009; 94.0% versus 92.2%, P = .046). Specificity, accuracy, and positive predictive values were lower with the convolutional neural network than with the American College of Radiology TI-RADS (86.1% versus 93.7%, P < .001; 85.1% versus 89.1%, P = .003; and 64.0% versus 77.8%, P < .001). The area under the curve of the convolutional neural network was higher than that of the American College of Radiology TI-RADS (0.917 versus 0.891, P = .017).

CONCLUSIONS

The convolutional neural network provided diagnostic performance comparable with that of the American College of Radiology TI-RADS categories assigned by experienced radiologists.

Implications of US radiomics signature for predicting malignancy in thyroid nodules with indeterminate cytology

OBJECTIVES

The purpose of this study was to evaluate the role of the radiomics score using US images to predict malignancy in AUS/FLUS and FN/SFN nodules.

METHODS

One hundred fifty-five indeterminate thyroid nodules in 154 patients who received initial US-guided FNA for diagnostic purposes were included in this retrospective study. A representative US image of each tumor was acquired, and square ROIs covering the whole nodule were drawn using the Paint program of Windows 7. Texture features were extracted by in-house texture analysis algorithms implemented in MATLAB 2019b. The LASSO logistic regression model was used to choose the most useful predictive features, and ten-fold cross-validation was performed. Two prediction models were constructed using multivariable logistic regression analysis: one based on clinical variables, and the other based on clinical variables with the radiomics score. Predictability of the two models was assessed with the AUC of the ROC curves.

RESULTS

Clinical characteristics did not significantly differ between malignant and benign nodules, except for mean nodule size. Among 730 candidate texture features generated from a single US image, 15 features were selected. Radiomics signatures were constructed with a radiomics score, using selected features. In multivariable logistic regression analysis, higher radiomics score was associated with malignancy (OR = 10.923; p < 0.001). The AUC of the malignancy prediction model composed of clinical variables with the radiomics score was significantly higher than the model composed of clinical variables alone (0.839 vs 0.583).

CONCLUSIONS

Quantitative US radiomics features can help predict malignancy in thyroid nodules with indeterminate cytology.

Comparison of diagnostic performance of the ACR and Kwak TIRADS applying the ACR TIRADS' size thresholds for FNA

OBJECTIVE

To investigate the diagnostic performances and unnecessary fine-needle aspiration (FNA) rates of two point-scale based TIRADS and compare them with a modified version using the ACR TIRADS' size thresholds.

METHODS

Our Institutional Review Board approved this retrospective study and waived the requirement for informed consent. A total of 2106 thyroid nodules 10 mm or larger in size in 2084 patients with definitive cytopathologic findings were included. Ultrasonography categories were assigned according to each guideline. We applied the ACR TIRADS' size thresholds for FNA to the Kwak TIRADS and defined it as the modified Kwak TIRADS (mKwak TIRADS). Diagnostic performances and unnecessary FNA rates were evaluated for both the original and modified guidelines.

RESULTS

Of the original guidelines, the ACR TIRADS had higher specificity, accuracy, and area under the receiver operating characteristic curve (AUC) (63.1%, 68.9%, and 0.748, respectively). When the size threshold of the ACR TIRADS was applied to the Kwak TIRADS, the resultant mKwak TIRADS had higher specificity, accuracy, and AUC (64.7%, 70.3%, and 0.765, respectively) than the ACR TIRADS. The mKwak TIRADS also had a lower unnecessary FNA rate than the ACR TIRADS (54.8% and 56.4%, respectively). The false-negative rate of the Kwak TIRADS was the lowest (1.9%) among all TIRADS.

CONCLUSION

The modified Kwak TIRADS incorporating the size thresholds of the ACR TIRADS showed higher diagnostic performance and a lower unnecessary FNA rate than the original point-scale based TIRADS.

KEY POINTS

• Of the original guidelines, the ACR TIRADS had the highest specificity, accuracy, and area under the receiver operating characteristic curve (AUC) (63.1%, 68.9%, and 0.748, respectively). • When the size threshold of the ACR TIRADS was applied to the Kwak TIRADS, the resultant modified version of Kwak TIRADS had higher specificity, accuracy, and AUC (64.7%, 70.3%, and 0.765, respectively) than the ACR TIRADS. • The false-negative rate of the Kwak TIRADS was the lowest (1.9%) among all TIRADS.

Diffusion-Weighted Magnetic Resonance Imaging for Breast Cancer Screening in High-Risk Women: Design and Imaging Protocol of a Prospective Multicenter Study in Korea

Purpose

Interest in unenhanced magnetic resonance imaging (MRI) screening for breast cancer is growing due to concerns about gadolinium deposition in the brain and the high cost of contrast-enhanced MRI. The purpose of this report is to describe the protocol of the Diffusion-Weighted Magnetic Resonance Imaging Screening Trial (DWIST), which is a prospective, multicenter, intraindividual comparative cohort study designed to compare the performance of mammography, ultrasonography, dynamic contrast-enhanced (DCE) MRI, and diffusion-weighted (DW) MRI screening in women at high risk of developing breast cancer.

Methods

A total of 890 women with BRCA mutation or family history of breast cancer and lifetime risk ≥ 20% are enrolled. The participants undergo 2 annual breast screenings with digital mammography, ultrasonography, DCE MRI, and DW MRI at 3.0 T. Images are independently interpreted by trained radiologists. The reference standard is a combination of pathology and 12-month follow-up. Each image modality and their combination will be compared in terms of sensitivity, specificity, accuracy, positive predictive value, rate of invasive cancer detection, abnormal interpretation rate, and characteristics of detected cancers. The first participant was enrolled in April 2019. At the time of manuscript submission, 5 academic medical centers in South Korea are actively enrolling eligible women and a total of 235 women have undergone the first round of screening. Completion of enrollment is expected in 2022 and the results of the study are expected to be published in 2026.

Discussion

DWIST is the first prospective multicenter study to compare the performance of DW MRI and conventional imaging modalities for breast cancer screening in high-risk women. DWIST is currently in the patient enrollment phase.

Trial Registration

ClinicalTrials.gov Identifier: NCT03835897

Follow-Up Intervals for Breast Imaging Reporting and Data System Category 3 Lesions on Screening Ultrasound in Screening and Tertiary Referral Centers

Objective

To assess the appropriate follow-up interval, and rate and timepoint of cancer detection in women with Breast Imaging Reporting and Data System (BI-RADS) 3 lesions on screening ultrasonography (US) according to the type of institution.

Materials and Methods

A total of 1451 asymptomatic women who had negative or benign findings on screening mammogram, BI-RADS 3 assessment on screening US, and at least 6 months of follow-up were included. The median follow-up interval was 30.8 months (range, 6.8–52.9 months). The cancer detection rate, cancer detection timepoint, risk factors, and clinicopathological characteristics were compared between the screening and tertiary centers. Nominal variables were compared using the chi-square or Fisher's exact test and continuous variables were compared using the independent t test or Mann-Whitney U test.

Results

In 1451 women, 19 cancers (1.3%) were detected; two (0.1%) were diagnosed at 6 months and 17 (1.2%) were diagnosed after 12.3 months. The malignancy rates were both 1.3% in the screening (9 of 699) and tertiary (10 of 752) centers. In the screening center, all nine cancers were invasive cancers and diagnosed after 12.3 months. In the tertiary center, two were ductal carcinomas in situ and eight were invasive cancers. Two of the invasive cancers were diagnosed at 6 months and the remaining eight cancers newly developed after 13.1 months.

Conclusion

One-year follow-up rather than 6-month follow-up may be suitable for BI-RADS 3 lesions on screening US found in screening centers. However, more caution is needed regarding similar findings in tertiary centers where 6-month follow-up may be more appropriate.

Cytopathologic criteria and size should be considered in comparison of fine-needle aspiration vs. core-needle biopsy for thyroid nodules: results based on large surgical series

PURPOSE

To evaluate and compared the diagnostic performances of FNA and CNB using various cytopathologic criteria and size subgroups to see how the comparison results differ accordingly.

METHODS

From May 2012 to May 2019, 8187 thyroid nodules in 8139 patients who had undergone preoperative US-guided FNA or CNB at outside clinics were included in this retrospective study (mean size: 11.9 ± 9.5 mm). Preoperative US-FNA was performed in 7496 (91.6%) nodules and US-CNB was performed in 691 (8.4%) nodules. Propensity score matching was used to compare the sensitivities between FNA and CNB in diagnosis of malignancy and neoplasm according to different cytologic test criteria.

RESULTS

Of the 8187 thyroid nodules, 7833 (95.7%) were malignant and 354 (4.3%) were benign. Mean size of the thyroid nodules in the CNB group was significantly larger than the FNA group, 15.7 ± 12.7 mm vs. 11.6 ± 9.0 mm, respectively (P < 0.001). After matching, sensitivity in the CNB group were significantly higher in the total population, and in subgroups <10 mm for criteria 1 and 2 (all P < 0.05, respectively). No significant differences were seen between the sensitivities of FNA and CNB for nodules ≥10 mm regardless of criteria in diagnosis of malignancy or neoplasm (all P > 0.05, respectively).

CONCLUSIONS

Results comparing sensitivities between FNA and CNB differ according to the different cytopathologic criteria used for calculation. CNB has significantly higher sensitivity to FNA in subcentimeter nodules when using criteria 1 or 2. Diagnostic sensitivities did not show significant differences for nodules ≥10 mm regardless of the cytopathologic criteria used, that should be considered in selecting biopsy methods.

Artificial intelligence to predict the BRAFV600E mutation in patients with thyroid cancer

Purpose

To investigate whether a computer-aided diagnosis (CAD) program developed using the deep learning convolutional neural network (CNN) on neck US images can predict the BRAF^V600E mutation in thyroid cancer.

Methods

469 thyroid cancers in 469 patients were included in this retrospective study. A CAD program recently developed using the deep CNN provided risks of malignancy (0–100%) as well as binary results (cancer or not). Using the CAD program, we calculated the risk of malignancy based on a US image of each thyroid nodule (CAD value). Univariate and multivariate logistic regression analyses were performed including patient demographics, the American College of Radiology (ACR) Thyroid Imaging, Reporting and Data System (TIRADS) categories and risks of malignancy calculated through CAD to identify independent predictive factors for the BRAF^V600E mutation in thyroid cancer. The predictive power of the CAD value and final multivariable model for the BRAF^V600E mutation in thyroid cancer were measured using the area under the receiver operating characteristic (ROC) curves.

Results

In this study, 380 (81%) patients were positive and 89 (19%) patients were negative for the BRAF^V600E mutation. On multivariate analysis, older age (OR = 1.025, p = 0.018), smaller size (OR = 0.963, p = 0.006), and higher CAD value (OR = 1.016, p = 0.004) were significantly associated with the BRAF^V600E mutation. The CAD value yielded an AUC of 0.646 (95% CI: 0.576, 0.716) for predicting the BRAF^V600E mutation, while the multivariable model yielded an AUC of 0.706 (95% CI: 0.576, 0.716). The multivariable model showed significantly better performance than the CAD value alone (p = 0.004).

Conclusion

Deep learning-based CAD for thyroid US can help us predict the BRAF^V600E mutation in thyroid cancer. More multi-center studies with more cases are needed to further validate our study results.

Annual Trends in Ultrasonography-Guided 14-Gauge Core Needle Biopsy for Breast Lesions

Objective

To examine time trends in ultrasonography (US)-guided 14-gauge core needle biopsy (CNB) for breast lesions based on the lesion size, Breast Imaging-Reporting and Data System (BI-RADS) category, and pathologic findings.

Materials and Methods

We retrospectively reviewed consecutive US-guided 14-gauge CNBs performed from January 2005 to December 2016 at our institution. A total of 22,297 breast lesions were included. The total number of biopsies, tumor size (≤ 10 mm to > 40 mm), BI-RADS category (1 to 5), and pathologic findings (benign, high risk, ductal carcinoma in situ [DCIS], invasive cancer) were examined annually, and the malignancy rate was analyzed based on the BI-RADS category.

Results

Both the total number of US scans and US-guided CNBs increased while the proportion of US-guided CNBs to the total number of US scans decreased significantly. The number of biopsies classified based on the tumor size, BI-RADS category, and pathologic findings all increased over time, except for BI-RADS categories 1 or 2 and category 3 (odds ratio [OR] = 0.951 per year, 95% confidence interval [CI]: 0.902, 1.002 and odds ratio = 0.979, 95% CI: 0.970, 0.988, respectively). Both the unadjusted and adjusted total malignancy rates and the DCIS rate increased significantly over time. BI-RADS categories 4a, 4b, and 4c showed a significant increasing trend in the total malignancy rate and DCIS rate.

Conclusion

The malignancy rate in the results of US-guided 14-gauge CNB for breast lesions increased as the total number of biopsies increased from 2005 to 2016. This trend persisted after adjusting for the BI-RADS category.

Strap muscle invasion in differentiated thyroid cancer does not impact disease-specific survival: a population-based study

The American Joint Committee on Cancer (AJCC) 8^th TNM staging system of differentiated thyroid cancer defines gross strap muscle invasion as T3b stage. However, the impact of strap muscle invasion on disease-specific survival (DSS) remains controversial. To elucidate the survival impact of strap muscle invasion of any degree in thyroid cancers, the Surveillance, Epidemiology, and End Results (SEER) database (1973–2018) was queried for thyroid cancer only patients on July 2019 (n = 19,914). The Cox proportional hazard analysis with multivariable adjustment revealed that strap muscle invasion was not a significant factor for DSS in tumors equal to or smaller than 40 mm (hazard ratio (HR) = 1.620 [confidence interval (CI) 0.917 – 2.860]; p = 0.097). The competing risk analysis with multivariable adjustment showed that strap muscle invasion did not significantly impact DSS regardless of tumor size or cause of death (cancer-caused death (Subdistribution HR (SDHR) = 1.567 [CI 0.984 – 2.495]; p = 0.059); deaths to other causes (SDHR = 1.155 [CI 0.842 – 1.585]; p = 0.370). A “modified” staging schema discarding strap muscle invasion as a T stage criterion showed better 10-year DSS distinction between T stages. The modified staging schema may better reflect cancer-caused death risk and may prevent potential overstaging.

Guideline Implementation on Fine-Needle Aspiration for Thyroid Nodules: Focusing on Micronodules

OBJECTIVE

We investigated patients who were referred to our institution after fine-needle aspiration (FNA) was performed at outside clinics to evaluate how many nodules satisfied the FNA indications of the Korean Thyroid Imaging Reporting and Data System (K-TIRADS) and compare that to the number of thyroid nodules that satisfy the FNA indications of the American College of Radiology (ACR)-TIRADS and American Thyroid Association (ATA) guidelines.

METHODS

Between January 2018 and December 2018, 2,628 patients were included in our study. The included patients were those referred for thyroid surgery after having a suspicious thyroid nodule. We retrospectively applied the three guidelines to each thyroid nodule and determined whether each nodule satisfied the FNA indications. We compared the proportion of nodules satisfying the FNA indications of each guideline using a generalized linear model and generalized estimating equation.

RESULTS

The median size of the 2,628 thyroid nodules was 0.9 cm (range, 0.2 to 9.5 cm). We found that FNA was not indicated for 54.1%, 47.7%, and 19.1% of nodules and 87.3%, 99.0%, and 97.8% among them were micronodules (<1 cm) according to the ACR-TIRADS, ATA guideline, and K-TIRADS, respectively. The proportion of micronodules which satisfied the FNA indications was significantly higher for the K-TIRADS (65.1%) compared to the ACR TIRADS (12.1%) and ATA guideline (12.1%) (P<.001).

CONCLUSION

Among patients referred for thyroid surgery to our institutions, about 35% of the micronodules underwent FNA despite not being appropriate for indications by the K-TIRADS. Systematic training for physicians as well as modifications to increase the sensitivity of the guideline may be needed to reduce the overdiagnosis of thyroid cancers, especially for micronodules.

Survival Rates of Breast Cancer Patients Aged 40 to 49 Years according to Detection Modality in Korea: Screening Ultrasound versus Mammography

Objective

The aim of this study was to compare the survival rates of Korean females aged 40 to 49 years with breast cancer detected by supplemental screening ultrasound (US) or screening mammography alone.

Materials and Methods

This single-institution retrospective study included 240 patients with breast cancer (mean age, 45.1 ± 2.8 years) detected by US or mammography who had undergone breast surgery between 2003 and 2008. Medical records were reviewed for clinicopathologic characteristics and detection methods. Disease-free survival (DFS) and overall survival (OS) were compared between patients with breast cancer in the US and mammography groups using the log-rank test. Multivariable cox regression analysis was used to identify independent variables associated with DFS and OS.

Results

Among the 240 cases of breast cancer, 43 were detected by supplemental screening US and 197 by screening mammography (mean follow-up: 7.4 years, 93.3% with dense breasts). There were 19 recurrences and 16 deaths, all occurring in the mammography group. While the US group did not differ from the mammography group in tumor stage, the patients in this group were more likely to undergo breast-conserving surgery and radiation therapy than the mammography group. The US group also showed better DFS (p = 0.016); however, OS did not differ between the two groups (p = 0.058). In the multivariable analysis, the US group showed a lower risk of recurrence (hazard ratio, 0.097; 95% confidence interval, 0.001–0.705) compared to the mammography group.

Conclusion

Our study found that Korean females aged 40–49 years with US-detected breast cancer showed better DFS than those with mammography-detected breast cancer. However, there were no statistically significant differences in OS.

Ultrasonography surveillance in papillary thyroid carcinoma patients after total thyroidectomy according to dynamic risk stratification

PURPOSE

To investigate the role of neck US surveillance in patients with papillary thyroid carcinoma (PTC) after total thyroidectomy according to dynamic risk stratification (DRS) based on response to initial therapy.

METHODS

This retrospective study included 812 patients with PTC who underwent total thyroidectomy with prophylactic central neck dissection from January 2003 through February 2007. The relative risk of recurrence/persistence according to DRS was evaluated with the multivariable Cox regression proportional hazard model.

RESULTS

There were 132 men and 680 women. The mean age at surgery was 45.2 years. Postoperative US was used for DRS. According to DRS, 676 patients had excellent response, 78 indeterminate response, 40 biochemical incomplete response, and 18 structural incomplete response to initial therapy. Neck US was performed during follow-up and detected locoregional recurrences in 21 patients (2.6%): 12 with excellent response, 2 with biochemical incomplete response, and 7 with structural incomplete response according to DRS. Only 1 patient (0.1%) with excellent response had a locoregional recurrence that exceeded 8 mm in its shortest diameter, which is the size cut-off for diagnostic US fine-needle aspiration in suspicious lymph nodes. This patient did not develop biochemical abnormalities during follow-up.

CONCLUSIONS

Postoperative neck US surveillance after total thyroidectomy with prophylactic central neck dissection is not essential in PTC patients who show excellent response to initial therapy. Future studies are needed to verify the role of US surveillance in patients who receive variable degrees of treatments.

Diagnostic performances and unnecessary US-FNA rates of various TIRADS after application of equal size thresholds

We compared the diagnostic performances and unnecessary FNA rates of several guidelines and modified versions using the size threshold of the ACR TIRADS. Our Institutional Review Board approved this retrospective study and waived the requirement for informed consent and all methods were performed in accordance with the Declaration of Helsinki. A total of 1,384 thyroid nodules in 1,301 patients with definitive cytopathologic findings were included. US categories were assigned according to each guideline. We applied the size threshold suggested by the ACR TIRADS for FNA to the Kwak, ATA and EU guidelines and defined these modified guidelines as the modified Kwak (mKwak), modified ATA (mATA) and modified EU (mEU) guidelines. Diagnostic performances and unnecessary FNA rates of all guidelines were evaluated. Of 1,384 thyroid nodules, 291 (21%) were malignant. Among the original guidelines, the ACR TIRADS had the highest specificity, accuracy, LR and AUC (62.2%, 66%, 2.128 and 0.713). The mKwak, mATA and mEU guidelines had higher specificity, accuracy, LR and AUC (P < 0.001 for all), and fewer unnecessary FNAs, compared with their original guidelines. Among all original and modified guidelines, the mKwak guideline had the highest specificity, accuracy, LR and AUC (64%, 68.6%, 2.389 and 0.75). The unnecessary FNA rate was the lowest with the mKwak guideline (61.1%). The highest sensitivity was observed with the ATA guideline (98.6%). After incorporating the size threshold of the ACR TIRADS to other TIRADS, all guidelines showed higher diagnostic accuracy and lower unnecessary FNA rates than their original versions. The mKwak guideline showed the best diagnostic performances.

First Experience in Korea of Stereotactic Partial Breast Irradiation for Low-Risk Early-Stage Breast Cancer

Purpose: Accelerated partial breast irradiation (A-PBI) in Korean women has been considered impracticable, owing to small breast volume and lack of high-precision radiotherapy experience. We present the first experience of stereotactic-PBI (S-PBI) with CyberKnife M6 to investigate feasibility of use and early toxicities in Korean women with early breast cancers.

Materials and Methods: A total of 104 breasts receiving S-PBI at our institution between September 2017 and October 2018 were reviewed. Patients were selected based on the American Society for Radiation Oncology (ASTRO), American Brachytherapy Society, American Society of Breast Surgeons, and Groupe Européen de Curiethérapie-European Society for Therapeutic Radiology and Oncology guidelines. A dose of 30 Gy in 5 fractions (NCT01162200) was used. Gold fiducials were routinely inserted near the tumor bed for tracking. Constraints regarding organs-at-risk followed the NSABP-B39/RTOG 0413 protocol.

Results: Median follow-up was for 13 months. Patients were categorized as “suitable” (71.2%) or “cautionary” (28.8%) according to 2017 the ASTRO guidelines. No tracking failure of inserted gold fiducials occurred. Median planning target volume (PTV) and PTV-to-whole breast volume ratio was 73.6 mL (interquartile range, 58.8–103.9 mL) and 17.0% (13.3–19.1%), respectively. Median PTV V_95%, PTV D_max, and ipsilateral breast V_50% were 97.8% (96.2–98.8%), 105.3% (104.2–106.4%), and 35.5% (28.3–39.8%), respectively. No immediate post-S-PBI toxicity ≥ grade 2 was reported, except grade 2 induration in three breasts. All patients remain disease-free to date.

Conclusion: The first use of S-PBI in Korean women was feasible and safe for selected early breast cancer. Based on these results, we have initiated a prospective study (NCT03568981) to test S-PBI in whole-breast irradiation for low-risk early breast cancer.

Factors Predicting Breast Cancer Development in Women During Surveillance After Surgery for Atypical Ductal Hyperplasia of the Breast: Analysis of Clinical, Radiologic, and Histopathologic Features

BACKGROUND

This study aimed to evaluate the risk of breast cancer development for women under surveillance after surgery for atypical ductal hyperplasia (ADH), as well as the clinical and pathologic factors associated with breast cancer development.

METHODS

From November 2003 to December 2014, the study included 205 women (mean age, 47.1 ± 11.2 years; range 18-73 years) with a pathologic diagnosis of ADH at surgical excision who had preoperative mammography and ultrasonography (US) images and pathology slides available for review. The patients were classified into three groups according to the detection method as follows: negative group (with ADH occult on imaging), mammography group (with ADH detected on mammography), and US group (with ADH detected on US only). Clinical, radiologic, and histopathologic factors associated with breast cancer development after ADH surgery were evaluated.

RESULTS

Breast cancer developed in 15 patients (7.3%) during surveillance after ADH surgery (follow-up period, 63.9 ± 40.8 months). Palpable lesions had significantly higher rates of breast cancer development after ADH surgery (26.7% vs 6.8%; P = 0.045). Breast cancer development after ADH surgery did not differ according to the detection method (P = 0.654). Palpability was significantly associated with breast cancer development during surveillance after ADH surgery (hazard ratio, 3.579; 95% confidence interval 1.048-12.220; P = 0.042).

CONCLUSION

The breast cancer development rate for women under surveillance after ADH surgery was 7.3%. Palpability at the time of ADH diagnosis was significantly associated with breast cancer development.

MRI Radiomic Features: Association with Disease-Free Survival in Patients with Triple-Negative Breast Cancer

Radiomic features hold potential to improve prediction of disease-free survival (DFS) in triple-negative breast cancer (TNBC) and may show better performance if developed from TNBC patients. We aimed to develop a radiomics score based on MRI features to estimate DFS in patients with TNBC. A total of 228 TNBC patients who underwent preoperative MRI and surgery between April 2012 and December 2016 were included. Patients were temporally divided into the training (n = 169) and validation (n = 59) set. Radiomic features of the tumor were extracted from T2-weighted and contrast-enhanced T1- weighted MRI. Then a radiomics score was constructed with the least absolute shrinkage and selection operator regression in the training set. Univariate and multivariate Cox proportional hazards models were used to determine what associations the radiomics score and clinicopathologic variables had with DFS. A combined clinicopathologic-radiomic (CCR) model was constructed based on multivariate Cox analysis. The incremental values of the radiomics score were evaluated by using the integrated area under the receiver operating characteristic curve (iAUC) and bootstrapping (n = 1000). The radiomics score, which consisted of 5 selected MRI features, was significantly associated with worse DFS in both the training and validation sets (p = 0.002, p = 0.033, respectively). In both the training and validation set, the radiomics score showed comparable performance with the clinicopathologic model. The CCR model demonstrated better performance than the clinicopathologic model in the training set (iAUC, 0.844; difference in iAUC, p < 0.001) and validation set (iAUC, 0.765, difference in iAUC, p < 0.001). In conclusion, MRI-based radiomic features can improve the prediction of DFS when integrated with clinicopathologic data in patients with TNBC.

Pattern-based vs. score-based guidelines using ultrasound features have different strengths in risk stratification of thyroid nodules

OBJECTIVE

To evaluate and compare the diagnostic performances between recently published pattern-based and score-based TIRADS according to the experience level of the performer.

METHODS

From July 2013 to January 2019, 8657 thyroid nodules in 8364 patients that had been cytopathologically diagnosed as benign or malignant were included (mean size, 22.0 mm ± 12.1). Thyroid nodules were categorized into US-based final assessment categories and US-FNA indications of five recently published TIRADS. Radiologists performing the US examinations were divided into the experienced vs. inexperienced group. Diagnostic performances and unnecessary biopsy rates were calculated and compared between the five TIRADS, also the experienced vs. inexperienced group.

RESULTS

Of the 8657 thyroid nodules, 6706 (77.5%) were benign and 1951 (22.5%) were malignant. Diagnostic performances for US-based final assessment categories showed higher sensitivity and NPV for EU-TIRADS (92.7% and 96.5%), while Kwak-TIRADS had higher specificity, PPV, accuracy, and AUC (89.6%, 68.0%, 86.5%, and 0.878; all p < 0.05, respectively). Diagnostic performances for US-FNA indications showed higher sensitivity and NPV for KTA/KSThR TIRADS (98.5% and 97.0%), while Kwak-TIRADS had higher specificity, PPV, accuracy, and AUC (70.3%, 46.6%, 74.5%, and 0.797; all p < 0.05, respectively). Unnecessary biopsy rates were the lowest in Kwak-TIRADS for both US categories and US-FNA indications (32.0% and 53.4%, p < 0.001). Similar trends were seen in both the experienced and inexperienced group.

CONCLUSION

The currently published score-based guidelines for thyroid nodules have significantly higher specificity, PPV, accuracy, and AUC and lower unnecessary biopsy rates, whereas pattern-based guidelines have higher sensitivity and NPV, regardless of the level of experience of the performer.

KEY POINTS

• For US-based final assessment categories, EU-TIRADS had higher sensitivity and NPV, while Kwak-TIRADS had higher specificity, PPV, accuracy, and AUC. • For US-FNA indications, KTA/KSThR TIRADS had higher sensitivity and NPV, while Kwak-TIRADS had higher specificity, PPV, accuracy, and AUC. • Similar trends were seen in diagnostic performances for both experienced and inexperienced groups.

Radiomics in predicting mutation status for thyroid cancer: A preliminary study using radiomics features for predicting BRAFV600E mutations in papillary thyroid carcinoma

Purpose

To evaluate whether if ultrasonography (US)-based radiomics enables prediction of the presence of BRAF^V600E mutations among patients diagnosed as papillary thyroid carcninoma (PTC).

Methods

From December 2015 to May 2017, 527 patients who had been treated surgically for PTC were included (training: 387, validation: 140). All patients had BRAF^V600E mutation analysis performed on surgical specimen. Feature extraction was performed using preoperative US images of the 527 patients (mean size of PTC: 16.4mm±7.9, range, 10–85 mm). A Radiomics Score was generated by using the least absolute shrinkage and selection operator (LASSO) regression model. Univariable/multivariable logistic regression analysis was performed to evaluate the factors including Radiomics Score in predicting BRAF^V600E mutation. Subgroup analysis including conventional PTC <20-mm (n = 389) was performed (training: 280, validation: 109).

Results

Of the 527 patients diagnosed with PTC, 428 (81.2%) were positive and 99 (18.8%) were negative for BRAF^V600E mutation. In both total 527 cancers and 389 conventional PTC<20-mm, Radiomics Score was the single factor showing significant association to the presence of BRAF^V600E mutation on multivariable analysis (all P<0.05). C-statistics for the validation set in the total cancers and the conventional PTCs<20-mm were lower than that of the training set: 0.629 (95% CI: 0.516–0.742) to 0.718 (95% CI: 0.650–0.786), and 0.567 (95% CI: 0.434–0.699) to 0.729 (95% CI: 0.632–0.826), respectively.

Conclusion

Radiomics features extracted from US has limited value as a non-invasive biomarker for predicting the presence of BRAF^V600E mutation status of PTC regardless of size.

Core-Needle Biopsy Does Not Show Superior Diagnostic Performance to Fine-Needle Aspiration for Diagnosing Thyroid Nodules

PURPOSE

To compare the diagnostic performances of fine-needle aspiration (FNA) and core-needle biopsy (CNB) for thyroid nodules according to nodule size.

MATERIALS AND METHODS

This retrospective study included 320 thyroid nodules from 320 patients who underwent both FNA and CNB at outside clinics and proceeded with surgery in our institution between July 2012 and May 2019. According to nodule size, the diagnostic performances of FNA and CNB were calculated using various combinations of test-negatives and test-positives defined by the Bethesda categories and were compared using the generalized estimated equation and the Delong method.

RESULTS

There were 279 malignant nodules in 279 patients and 41 benign nodules in 41 patients. The diagnostic performance of FNA was mostly not different from CNB regardless of nodule size, except for negative predictive value, which was better for FNA than CNB when applying Criteria 1 and 2. When applying Criteria 3, the specificity and positive predictive value of FNA were superior to CNB regardless of size. When applying Criteria 4, diagnostic performance did not differ between FNA and CNB regardless of size. After applying Criteria 5, diagnostic performance did not differ between FNA and CNB in nodules ≥2 cm. However, in nodules ≥1 cm and all nodules, the sensitivity, accuracy, and negative predictive value of CNB were better than those of FNA.

CONCLUSION

CNB did not show superior diagnostic performance to FNA for diagnosing thyroid nodules.

Follow-Up Strategies for Thyroid Nodules with Benign Cytology on Ultrasound-Guided Fine Needle Aspiration: Malignancy Rates of Management Guidelines Using Ultrasound Before and After the Era of the Bethesda System

Background: To evaluate differences in malignancy rates and consequent follow-up strategies for cytologically benign thyroid nodules before and after the introduction of the Bethesda system according to the risk stratification categories of four thyroid management guidelines. Methods: This retrospective study was approved by our institutional review board. In this study, 1716 thyroid nodules with initially benign cytologic diagnosis at ultrasound-guided fine needle aspiration (US-FNA) in 1695 patients were included: 1187 nodules from the pre-Bethesda period and 529 nodules from the post-Bethesda period. Based on US features, the thyroid nodules were categorized into the final assessment categories of the 2015 American Thyroid Association (ATA), the 2016 American Association of Clinical Endocrinologists, American College of Endocrinology, and Associazione Medici Endocrinologi, the American College of Radiology Thyroid Imaging Reporting and Data System (ACR-TIRADS), and the European Thyroid Association guidelines for Ultrasound Malignancy Risk Stratification systems. Estimated malignancy rates before and after propensity score matching according to follow-up intervals were obtained. Results: Of the 1716 thyroid nodules benign on initial US-FNA, the malignancy rate was 3.2% (38 of 1187) in the pre-Bethesda period and 2.6% (14 of 529) in the post-Bethesda period (p = 0.641). The 2015 ATA high suspicion pattern and the ACR-TIRADS category 5 had high estimated malignancy rates of >5% in the post-Bethesda period (6.52 and 8.57, respectively). Positive findings that indicated US-FNA in the ACR-TIRADS had estimated malignancy rates of 5.26 and 5.67, respectively, while the remaining guidelines had estimated malignancy rates of <5% in both periods. Conclusions: Immediate diagnostic intervention after benign cytologic diagnosis may not be necessary regardless of the cytologic criteria applied, but it can be considered for the highly suspicious categories in the 2015 ATA or the ACR-TIRADS for benign cytologic diagnosis of the Bethesda system.

Diffusional kurtosis imaging for differentiation of additional suspicious lesions on preoperative breast MRI of patients with known breast cancer

PURPOSE

To investigate the potential of diffusional kurtosis imaging (DKI) and conventional diffusion-weighted imaging (DWI) in the evaluation of additional suspicious lesions at preoperative breast magnetic resonance imaging (MRI) in patients with breast cancer.

MATERIALS AND METHODS

Fifty-three additional suspicious lesions in 45 patients with breast cancer, which were detected on preoperative breast MRI, were examined with a 3-T MR system. DKI and DWI data were obtained using a spin-echo single-shot echo-planar imaging sequence with b-values of 0, 50, 600, 1000, and 3000 s/mm². Histogram parameters (mean, standard deviation, minimum, maximum, 10th, 25th, 50th, 75th, 90th percentiles, kurtosis, skewness and entropy) of ADC from DWI and diffusivity (D), kurtosis (K) from DKI were calculated after postprocessing. Parameters were compared between benign vs. ductal carcinoma in situ (DCIS) vs. invasive breast lesions and diagnostic performances were evaluated by receiver operating characteristic (ROC) analysis. Correlation between the mean values of D and K was analyzed according to lesion type.

RESULTS

Multiple histogram parameters of D (mean, 25th, 50th percentile, 75th percentile, and entropy) differed between benign and invasive breast lesions (all P < 0.005), but none differed between benign vs. DCIS. D-90th percentile differed between DCIS vs. invasive cancer (P = 0.040). K-10th percentile differed between benign vs. DCIS (P = 0.015). ADC-75th percentile differed between benign vs. invasive cancer and ADC-75th percentile, ADC-90th percentile differed between DCIS vs. invasive cancer, respectively (all P < 0.005). ROC curve analysis showed high specificity for discrimination between benign and invasive cancer. D-mean and K-mean showed strong correlation in benign (r_s = -0.813) and invasive lesions (r_s = -0.853), but no significant correlation in DCIS.

CONCLUSION

DKI may aid in the differentiation of additional suspicious lesions at preoperative breast MRI. Both ADC and DKI may have lower potential in differentiating DCIS from benign lesions.

Automated artificial intelligence quantification of fibroglandular tissue on breast MRI

e12071

Background: The objective of this study is to examine if a convolutional neural network can be utilized to automate breast fibroglandular tissue segmentation, a risk factor for breast cancer, on MRIs. Methods: This institutional review board approved study assessed retrospectively acquired MRI T1 pre-contrast image data for 238 patients. Ground truth parameters were derived through manual segmentation. A hybrid 3D/2D U-Net architecture was developed for fibroglandular tissue segmentation. The network was trained with T1 pre-contrast MRI data and their corresponding ground-truth labels. The analysis was started with image pre-processing. Each MRI volume was re-sampled and normalized using z-scores. Convolution operations reduced 3D volumes into a 2D slice in the contracting arm of the U-Net architecture. Results: A 5-fold cross validation was performed and the Dice similarity coefficient was used to assess the accuracy of fibroglandular tissue segmentation. Cross-validation results showed that the automated hybrid CNN approach resulted in a Dice similarity coefficient of 0.848 and a Pearson correlation of 0.961 in comparison to the ground-truth for fibroglandular breast tissue segmentation, which demonstrates high accuracy. Conclusions: The results demonstrate significant application of deep learning in accurately automating segmentation of breast fibroglandular tissue.

Automated breast cancer lesion detection on breast MRI using artificial intelligence

e14612

Background: Over 40,000 women in the US will die from breast cancer. Early detection of cancer is crucial and is a potential avenue to improve survival. The objective of this research study is to develop a convoluted neural network (CNN), a subset of artificial intelligence, in order to enhance computerized detection of breast lesions on MRIs. Methods: This is an institutional review board approved retrospective study with post contrast MRI data from 238 patients. Breast tumor segmentation was automated with a hybrid 3D/2D CNN designed adapted from U-net, a popular neural network architecture in biomedical image analysis. T1 post-contrast MRI volumes were used to train the network. The data set was separated into training (80%) and validation (20%) sets. Re-sampling and normalization using z-scores were applied to each volume before training. Contracting and expanding arms of the model consist of successive convolutions followed by batch normalization and ReLU operations. Ground truth was established through manual segmentation and previously conducted readings of the images used to train our network. Results: A 5-fold cross validation was performed for analysis. The Dice similarity coefficient was used to assess segmentation accuracy. The hybrid 3D/2D U-Net architecture yielded a Dice score of 0.753 and a Pearson correlation of 0.548 for the breast tumor segmentation. Conclusions: These results demonstrated the feasibility for artificial intelligence applications in accurately identifying the presence of lesions on breast MRI images.

Comparison of breast tissue markers for tumor localization in breast cancer patients undergoing neoadjuvant chemotherapy

PURPOSE

The purpose of this study was to compare the visibility of breast tissue markers in cases of breast cancer on ultrasonography (US) after neoadjuvant chemotherapy (NAC) and to analyze whether the type of marker affected the choice of localization method after NAC.

METHODS

We included 153 tissue markers inserted within breast cancers that showed pathologically complete response (pCR) after NAC from January 2012 to April 2017. One of three types of markers (a surgical clip, Cormark, or UltraClip) was inserted. Medical records and imaging findings were retrospectively reviewed. We compared the visibility of the different types of tissue markers on US after NAC, and also compared the imaging modalities used in the preoperative localization. The chi-square test, Fisher exact test, and multiple logistic regression were used for analysis.

RESULTS

Of the 153 tissue markers, 56 were surgical clips, 61 Cormark, and 36 UltraClip. After NAC, residual lesions were not seen on US in 42 cases (27.5%). In multivariate analysis, the visibility of the surgical clips and Cormark markers was better than that of the UltraClip markers (odds ratio [OR], 5.467; 95% confidence interal [CI], 1.717 to 17.410; P=0.004 and OR, 3.045; 95% CI, 1.074 to 8.628; P=0.036, respectively). Among the 131 cases where localization targeting the marker was required, the proportion of US-guided localizations was significantly higher when a surgical clip was used than when an UltraClip marker was used (OR, 5.566; 95% CI, 1.610 to 19.246; P=0.007) in the multivariate analysis.

CONCLUSION

The type of breast tissue marker affected its visibility on US in cases with pCR after NAC, which in turn affected the localization methodology.

Role of elastography for downgrading BI-RADS category 4a breast lesions according to risk factors

BACKGROUND

Elastography has been introduced as an additional diagnostic tool to ultrasonography (US) which helps clinicians decide whether or not to perform biopsy on US-detected lesions.

PURPOSE

To evaluate the role of strain elastography in downgrading Breast Imaging Reporting and Data System (BI-RADS) category 4a breast lesions according to personal risk factors for breast cancer in asymptomatic women.

MATERIAL AND METHODS

Strain elastography features of a total of 255 asymptomatic category 4a lesions were classified as soft and not soft (intermediate and hard). Malignancy was confirmed by surgery or biopsy, and benignity was confirmed by surgery or biopsy with no change on US for at least six months. Malignancy rates of lesions with soft and not soft elastography were calculated according to the presence of risk factors.

RESULTS

Of 255 lesions, 25 (9.8%) were malignant and 230 (90.2%) were benign. Of 195 lesions in average-risk women, the malignancy rate of lesions with soft elastography was 1.5% (1/68), which was significantly lower than the 14.2% (18/127) of lesions with not soft elastography ( P = 0.004). Of 60 lesions in increased-risk women, the malignancy rate of lesions with soft elastography was 15.0% (3/20), which was not significantly different from the 7.5% (3/40) of lesions with not soft elastography ( P = 0.390).

CONCLUSION

In average-risk women, category 4a lesions with soft elastography could be followed up with US because of a low malignancy rate of 1.5%.

Comparison of Clinical and Pathologic Characteristics of Ductal Carcinoma in Situ Detected on Mammography versus Ultrasound Only in Asymptomatic Patients

The main purpose of this study was to retropectively compare the clinical and pathologic characteristics of ductal carcinoma in situ (DCIS) detected on mammography and ultrasound (US) in asymptomatic patients. From February 2014 to September 2016, 236 asymptomatic patients with primary pure DCIS and dense breasts were included. The patients were classified into two groups. The mammography group (n = 165) included patients with DCIS detected on mammography, and the US group (n = 71) included patients with DCIS detected on US only. Clinicopathologic characteristics were compared between the two groups. Subgroup analyses were performed with a cutoff age of 50 y and a cutoff tumor size of 20 mm. In 236 patients, younger age, smaller tumor size, low nuclear grade, no comedo necrosis and progesterone receptor positivity were observed more in the US group (p < 0.05). HER2 and Ki67 positivity was observed more frequently in the mammography group (p < 0.05). Similar results were obtained in 168 patients with DCIS <20 mm and in patients ≥50 y. In patients <50 y, smaller tumor size, low nuclear grade and no comedo necrosis were observed significantly more often in the US group. DCIS in the US group significantly more often manifested low nuclear grade, no comedo necrosis and hormone receptor positivity, whereas HER2 and Ki67 positivity was observed significantly more often in the mammography group.

Associations between Bethesda categories and tumor characteristics of conventional papillary thyroid carcinoma

PURPOSE

The aim of this study was to investigate the associations of Bethesda categories III, V, and VI with the clinical and pathological features of thyroid nodules surgically confirmed as conventional papillary thyroid carcinomas (PTCs).

METHODS

We analyzed 1,990 consecutive patients diagnosed with conventional PTC at surgery with preoperative Bethesda categories III, V, or VI. We determined the odds ratio (ORs) of the clinical and pathological variables associated with categories III and V, using category VI as the reference.

RESULTS

Category III and V PTCs had a smaller pathological tumor size (OR, 0.934 and OR, 0.969, respectively) and less frequently had central lymph node metastasis (OR, 0.487 and OR, 0.780, respectively) than category VI PTCs. Category III PTCs less frequently showed suspicious ultrasonographic features (OR, 0.296) than category VI PTCs, and category V PTCs less frequently had gross extrathyroidal extension, with borderline significance (OR, 0.643; P=0.059).

CONCLUSION

Conventional PTCs with a preoperative Bethesda category of III or V may less frequently exhibit poor prognostic factors than those with malignant cytology.

Application of Various Additional Imaging Techniques for Thyroid Ultrasound: Direct Comparison of Combined Various Elastography and Doppler Parameters to Gray-Scale Ultrasound in Differential Diagnosis of Thyroid Nodules

This study was conducted to evaluate the role in the differential diagnosis of thyroid nodules of various elastographic and Doppler parameters when added to gray-scale ultrasonography (US). One-hundred seventy-one thyroid nodules (63 malignant, 108 benign) in 169 patients were included. Elastography (strain and shear wave elastography) and Doppler (power Doppler, superb microvascular imaging and microflow imaging) images of the same thyroid nodule were obtained using a single US machine. The diagnostic performance parameters of gray-scale US with and without elastography and those of Doppler US were calculated and compared. The specificity, positive predictive value and accuracy of gray-scale US were significantly higher than those of US combined with elastographic parameters (all p values < 0.05). The area under the receiver operating characteristic curve for gray-scale US was 0.877, significantly higher than that for US combined with elastography patterns, shear wave elastography ratio (all p values < 0.05) and Doppler parameters. Adding additional imaging modalities such as elastography and Doppler does not improve the diagnostic performance of gray-scale US in differentiating thyroid nodules.

Can Biannual Ultrasound Surveillance Detect Smaller Second Cancers or Detect Cancers Earlier in Patients with Breast Cancer History?

The aim of the work described here was to evaluate whether surveillance with biannual ultrasound (US) plus annual mammography (biannual group) for women with a history of breast cancer surgery results in earlier detection or in the detection of smaller second cancers than annual US plus mammography (annual group). Additionally, we compared the prevalence of distant metastases or palpable second cancers between the biannual and annual groups. The institutional review board of our institution approved this retrospective study, and patient consent was waived. Between January 2011 and December 2012, we retrospectively reviewed the clinical and imaging follow-up of 3023 patients with mammographic and US surveillance after breast cancer surgery to assess second cancers detected by local surveillance (locoregional recurrence, contralateral breast cancer or distant metastasis). The biannual and annual groups were divided with respect to the mean surveillance interval and compared with respect to clinicopathologic findings. Multivariable logistic regression with propensity score methods was used to examine the effect of the type of surveillance on outcomes. As for the size of the second cancer, no difference was seen between the biannual and annual groups (12.8 ± 6.6 mm vs. 14.1 ± 7.1 mm, p = 0.461); neither was there a significant difference between the groups in the presence of symptoms at the time of diagnosis of the second cancer (17.0% [8/47] vs. 10% [2/20], p = 0.711). Regardless of detection by local surveillance, the prevalence of distant metastases did not differ between the two groups (1.1% [27/2370] vs. 1.0% [7/653], p = 0.88) on univariate or multivariate analysis. The results of our retrospective study indicate that second cancers detected by biannual US surveillance in patients with a history of breast cancer surgery are not smaller and do not occur earlier than those detected by annual US surveillance. However, a randomized controlled study is required to verify these results before they can be generalized to clinical practice.

Intrinsic Subtypes of Breast Cancers Initially Assessed as Probably Benign or of Low Suspicion on Ultrasonography Differ According to Tumor Size

OBJECTIVES

To investigate whether the intrinsic subtypes of breast cancers initially assessed as American College of Radiology Breast Imaging and Reporting System (BI-RADS) category 3 or 4a differ according to tumor size.

METHODS

A total of 444 breast cancers in 439 patients initially assessed as BI-RADS 3 or 4a through ultrasound-guided core needle biopsy were included. Tumors were classified by the size criterion of 10 mm or smaller or larger than 10 mm and categorized as the luminal type (estrogen receptor [ER] positive and/or progesterone receptor [PR] positive and human epidermal growth factor receptor 2 [HER2] negative), HER2 type (HER2 positive regardless of ER or PR status), or triple-negative type (ER negative, PR negative, and HER2 negative). The relationships between tumor size and breast cancer intrinsic subtypes were analyzed.

RESULTS

A total of 247 (55.6%) cancers were 10 mm or smaller, and 197 (44.4%) were larger than 10 mm. The luminal type was more frequently present in tumors of 10 mm or smaller (185 of 247 [74.9%]) than those larger than 10 mm (112 of 197 [56.9%]; P < .001). No significant difference was noted in the frequency of the HER2 type in tumors of 10 and mm smaller and those larger than 10 mm (27 of 247 [10.9%] versus 36 of 197 [18.3%]; P = .084). The triple-negative type was more frequently present in tumors larger than 10 mm than those of 10 mm or smaller (49 of /197 [24.9%] versus 35 of 247 [14.2%]; P = .012).

CONCLUSIONS

Breast cancers assessed as BI-RADS category 3 or 4a had differing intrinsic subtypes according to tumor size, as the luminal type was more frequently present in tumors of 10 mm or smaller than those larger than 10 mm, whereas the triple-negative type was more frequently present in tumors larger than 10 mm.