Improving mammography interpretation for both novice and experienced readers: a comparative study of two commercial artificial intelligence software.

OBJECTIVES: To evaluate the improvement of mammography interpretation for novice and experienced radiologists assisted by two commercial AI software. METHODS: We compared the performance of two AI software (AI-1 and AI-2) in two experienced and two novice readers for 200 mammographic examinations (80 cancer cases). Two reading sessions were conducted within 4 weeks. The readers rated the likelihood of malignancy (range, 1-7) and the percentage probability of malignancy (range, 0-100%), with and without AI assistance. Differences in AUROC, sensitivity, and specificity were analyzed. RESULTS: Mean AUROC increased in both novice (0.86 to 0.90 with AI-1 [p = 0.005]; 0.91 with AI-2 [p < 0.001]) and experienced readers (0.87 to 0.92 with AI-1 [p < 0.001]; 0.90 with AI-2 [p = 0.004]). Sensitivities increased from 81.3 to 88.8% with AI-1 (p = 0.027) and to 91.3% with AI-2 (p = 0.005) in novice readers, and from 81.9 to 90.6% with AI-1 (p = 0.001) and to 87.5% with AI-2 (p = 0.016) in experienced readers. Specificity did not decrease significantly in both novice (p > 0.999, both) and experienced readers (p > 0.999 with AI-1 and 0.282 with AI-2). There was no significant difference in the performance change depending on the type of AI software (p > 0.999). CONCLUSION: Commercial AI software improved the diagnostic performance of both novice and experienced readers. The type of AI software used did not significantly impact performance changes. Further validation with a larger number of cases and readers is needed. CLINICAL RELEVANCE STATEMENT: Commercial AI software effectively aided mammography interpretation irrespective of the experience level of human readers. KEY POINTS: • Mammography interpretation remains challenging and is subject to a wide range of interobserver variability. • In this multi-reader study, two commercial AI software improved the sensitivity of mammography interpretation by both novice and experienced readers. The type of AI software used did not significantly impact performance changes. • Commercial AI software may effectively support mammography interpretation irrespective of the experience level of human readers.

Ultrasound malignancy risk stratification of thyroid nodules based on the degree of hypoechogenicity and echotexture.

OBJECTIVES: The ultrasound (US) lexicon of nodule echogenicity and echotexture is one of the major differences among various risk stratification systems of thyroid nodules. This study aimed to stratify the US malignancy risk of thyroid nodules based on their degree of hypoechogenicity and echotexture. MATERIAL AND METHODS: This retrospective study included a total of 2255 consecutive thyroid nodules (≥ 1 cm) with final diagnoses (malignancy rate, 13%) from 2011 to 2016. Thyroid nodules were stratified according to the US degree of hypoechogenicity (mild, moderate, or marked hypoechogenicity) and echotexture (homogeneous vs. heterogeneous). The calculated malignancy risk was compared between each category. RESULTS: There was no significant difference of malignancy risk between the homogeneous markedly hypoechoic and moderately hypoechoic nodules (p ≥ .18). However, the malignancy risks of markedly and moderately hypoechoic nodules were significantly higher than those of mildly hypoechoic nodules (p < .001). Heterogeneous predominantly hypoechoic thyroid nodules showed a significantly higher malignancy risk than predominantly iso- or hyperechoic thyroid nodules (p < .001). There were no significant differences of malignancy risk between heterogeneous predominantly hypoechoic and homogeneous hypoechoic nodules according to the degree of hypoechogenicity (p ≥ .12) and between heterogeneous predominantly iso- or hyperechoic nodules and homogeneous iso- or hyperechoic thyroid nodules (p = .36). CONCLUSIONS: The malignancy risk of nodule hypoechogenicity is stratified as mild vs. moderate to marked hypoechogenicity, and the malignancy risk of nodules with heterogeneous echotexture is stratified by the predominant echogenicity of the nodules. KEY POINTS: • Thyroid nodule echogenicity is categorized as marked, moderate, or mild hypoechogenicity and iso- or hyperechogenicity with the reference standard of adjacent thyroid tissue and anterior neck muscles. • The malignancy risk of thyroid nodule echogenicity is stratified as iso- or hyperechoic vs. mild vs. moderate or marked hypoechogenicity. • The malignancy risk of nodules with heterogeneous echotexture is stratified by the predominant echogenicity.

Similarities and Differences Between Thyroid Imaging Reporting and Data Systems.

OBJECTIVE. The purposes of this study were to identify the similarities and differences among the Korean Thyroid Imaging Reporting and Data System (TIRADS), American College of Radiology (ACR) TI-RADS, and European TIRADS and to compare the diagnostic performance of sonographic fine-needle aspiration (FNA) criteria for detecting malignant thyroid nodules. MATERIALS AND METHODS. This study included 2274 consecutively evaluated thyroid nodules 1 cm or larger in 1836 patients with final diagnoses from January 2011 to December 2016. US features of the nodules were retrospectively reviewed and classified according to three published guidelines from international societies. We compared the US lexicons, categories, and diagnostic performance of the FNA criteria for malignant nodules among the three reporting systems. RESULTS. Of the 2274 thyroid nodules, 1974 (86.8%) were benign and 300 (13.2%) were malignant; 93.7% of all malignancies were papillary carcinoma. Most of the US lexicons were similar among the three systems except for the definition of echogenicity of a nodule of mixed echogenicity in European TIRADS. Although the categories had strong correlations (r = 0.777-0.877, all p < 0.001) among the three systems, significant differences were observed in categories 5, 4, and 3 of nodules (all p < 0.001) and in the sensitivity, specificity, and rate of unnecessary FNA of the FNA criteria for malignancy (all p < 0.001). CONCLUSION. Although the three systems had similarities in most US lexicons, significant differences were observed in the classified categories and diagnostic performance of the FNA criteria for malignancy.

Malignancy risk of thyroid nodules with nonshadowing echogenic foci.

PURPOSE: This study was conducted to determine the malignancy risk and diagnostic value of various types of nonshadowing echogenic foci (NEF) in the risk stratification of thyroid nodules. METHODS: A total of 1,018 consecutive thyroid nodules (≥1 cm) with final diagnoses were included. The presence of NEF was determined and types of NEF were classified according to the presence of a comet tail artifact (CTA), location, and size through a prospective evaluation. The associations with malignancy, malignancy risk, and diagnostic value of various types of NEF were assessed. RESULTS: Intrasolid punctate NEF without CTA was the only type of NEF that was an independent predictor of malignancy (P<0.001). The malignancy risk of intrasolid punctate NEF without CTA was substantially higher in solid hypoechoic nodules than in isoechoic or nonsolid nodules (71.3% vs. 9.2%, P<0.001). In solid hypoechoic nodules, slightly increased sensitivity (70.8% vs. 67.9%) for malignancy and a similar malignancy risk (71.4% vs. 71.3%) were observed for intrasolid punctate NEF (with or without CTA) and intrasolid punctate NEF without CTA, respectively. NEF with CTA at the margin of the cystic component was not associated with malignancy or benignity in nonsolid nodules (P>0.05). CONCLUSION: Intrasolid punctate NEF without CTA was the only independent predictor of malignancy. However, solid hypoechoic nodules with intrasolid punctate NEF should be classified as high-suspicion nodules regardless of coexisting CTA. Other types of NEF had no added value for detecting malignancy compared to intrasolid punctate NEF without CTA.

Malignancy Risk Stratification of Thyroid Nodules with Macrocalcification and Rim Calcification Based on Ultrasound Patterns.

OBJECTIVE: To determine the association of macrocalcification and rim calcification with malignancy and to stratify the malignancy risk of thyroid nodules with macrocalcification and rim calcification based on ultrasound (US) patterns. MATERIALS AND METHODS: The study included a total of 3603 consecutive nodules (≥ 1 cm) with final diagnoses. The associations of macrocalcification and rim calcification with malignancy and malignancy risk of the nodules were assessed overall and in subgroups based on the US patterns of the nodules. The malignancy risk of the thyroid nodules was categorized as high (> 50%), intermediate (upper-intermediate: > 30%, ≤ 50%; lower-intermediate: > 10%, ≤ 30%), and low (≤ 10%). RESULTS: Macrocalcification was independently associated with malignancy in all nodules and solid hypoechoic (SH) nodules (p < 0.001). Rim calcification was not associated with malignancy in all nodules (p = 0.802); however, it was independently associated with malignancy in partially cystic or isoechoic and hyperechoic (PCIH) nodules (p = 0.010). The malignancy risks of nodules with macrocalcification were classified as upper-intermediate and high in SH nodules, and as low and lower-intermediate in PCIH nodules based on suspicious US features. The malignancy risks of nodules with rim calcification were stratified as low and lower-intermediate based on suspicious US features. CONCLUSION: Macrocalcification increased the malignancy risk in all and SH nodules with or without suspicious US features, with low to high malignancy risks depending on the US patterns. Rim calcification increased the malignancy risk in PCIH nodules, with low and lower-intermediate malignancy risks based on suspicious US features. However, the role of rim calcification in risk stratification of thyroid nodules remains uncertain.

Diagnostic performance of the modified Korean Thyroid Imaging Reporting and Data System for thyroid malignancy according to nodule size: a comparison with five society guidelines.

PURPOSE: The aim of this study was to evaluate the diagnostic performance of the modified Korean Thyroid Imaging Reporting and Data System (K-TIRADS) compared with five society risk stratification systems (RSSs) according to nodule size. METHODS: In total, 3,826 consecutive thyroid nodules (≥1 cm) with final diagnoses in 3,088 patients were classified according to five RSSs. The K-TIRADS was modified by raising the biopsy size threshold for low-suspicion nodules and subcategorizing intermediate-suspicion nodules. We assessed the performance of the RSSs as triage tests and their diagnostic accuracy according to nodule size (with a threshold of 2 cm). RESULTS: Of all nodules, 3,277 (85.7%) were benign and 549 (14.3%) were malignant. In small thyroid nodules (≤2 cm), the American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) had the highest reduction rate of unnecessary biopsies (76.3%) and the lowest sensitivity (76.1%). The modified K-TIRADS had the second highest reduction rate of unnecessary biopsies (67.6%) and sensitivity (86.6%). The modified K-TIRADS and ACR TI-RADS had the highest diagnostic odds ratios (P=0.165) and the highest areas under the curve (P=0.315). In large nodules (>2 cm), the sensitivity of the ACR TI-RADS for malignancy was significantly lower (88.8%) than the sensitivities of the modified K-TIRADS and other RSSs, which were very high (98.7%-99.3%) (P<0.001). CONCLUSION: The modified K-TIRADS allows a large proportion of unnecessary biopsies to be avoided, while maintaining high sensitivity and diagnostic accuracy for small malignant tumors and very high sensitivity for large malignant tumors.

CT features of thyroid nodules with isolated macrocalcifications detected by ultrasonography.

PURPOSE: A thyroid nodule with an isolated macrocalcification is visualized as a calcified nodule with complete posterior shadowing on ultrasonography (US). This study aimed to determine the computed tomography (CT) features of isolated macrocalcifications detected using US. METHODS: This study included 20 patients who had thyroid nodules with isolated macroalcifications and underwent neck CT or chest CT. The patients were enrolled from a sample of 82 patients with isolated macrocalcifications detected by US drawn from 7,142 consecutive patients who underwent thyroid biopsy at two institutions. We evaluated the CT features of nodules with isolated macrocalcifications and categorized them as central or rim calcifications. We assessed the nodule size and the frequency of nondiagnostic fine-needle aspiration (FNA) results and malignant tumors according to the CT features of isolated macrocalcifications. RESULTS: CT scans showed central calcifications in 18 (90.0%) and rim calcifications in two (10.0%) of the 20 nodules with isolated macrocalcifications. Among the 18 nodules with central isolated macrocalcifications, complete compact calcification was found in six nodules and partial coarse calcification in 12 nodules. In 18 nodules with central isolated macrocalcifications, the nondiagnostic FNA rate and frequency of malignant tumors were not significantly different between complete and partial central calcifications (P=0.620 and P=0.999, respectively). Malignant tumors were only found in nodules with central isolated macrocalcifications. CONCLUSION: The majority of nodules with isolated macrocalcifications showed central calcifications on CT. Thyroid nodules with isolated macrocalcifications detected by US should not be classified as having a type of rim or peripheral calcification.

Thyroid Nodules with Isolated Macrocalcifications: Malignancy Risk of Isolated Macrocalcifications and Postoperative Risk Stratification of Malignant Tumors Manifesting as Isolated Macrocalcifications.

OBJECTIVE: To determine the malignancy risk of isolated macrocalcifications (a calcified nodule with complete posterior acoustic shadowing) detected on ultrasonography (US) and to evaluate the postoperative American Thyroid Association (ATA) risk stratification of malignant tumors manifesting as isolated macrocalcifications. MATERIALS AND METHODS: A total of 3852 thyroid nodules (≥ 1 cm) of 3061 consecutive patients who had undergone biopsy between January 2011 and June 2018 were included in this study. We assessed the prevalence, malignancy rate, and size distribution of isolated macrocalcifications and evaluated the histopathologic features and postoperative ATA risk stratification of malignant tumors manifesting as isolated macrocalcifications. RESULTS: Isolated macrocalcifications were found in 38 (1.2%) of the 3061 patients. Final diagnosis was established in 30 (78.9%) nodules; seven malignant tumors were diagnosed as papillary thyroid carcinomas (PTCs). The malignancy rate of the isolated macrocalcifications was 23.3% in the 30 nodules with final diagnoses and 18.4% in all nodules. Among the six surgically-treated malignant tumors, five (83.3%) had an extrathyroidal extension (ETE) (minor ETE 1, gross ETE 4), and two (33.3%) had macroscopic lymph node metastasis. Four (66.7%) malignant tumors were categorized as high-risk tumors, one as an intermediate-risk tumor, and one as a low-risk tumor using the ATA risk stratification. Histopathologically, out of the six malignant tumors, ossifications were noted in four (66.7%) and predominant calcifications in two (33.3%). CONCLUSION: The US pattern of isolated macrocalcifications (≥ 1 cm) showed an intermediate malignancy risk (at least 18.4%). All malignant tumors were PTCs, and most showed an aggressive behavior and a high or intermediate postoperative ATA risk.