Collaborative Enhancement of Consistency and Accuracy in US Diagnosis of Thyroid Nodules Using Large Language Models.

Background Large language models (LLMs) hold substantial promise for medical imaging interpretation. However, there is a lack of studies on their feasibility in handling reasoning questions associated with medical diagnosis. Purpose To investigate the viability of leveraging three publicly available LLMs to enhance consistency and diagnostic accuracy in medical imaging based on standardized reporting, with pathology as the reference standard. Materials and Methods US images of thyroid nodules with pathologic results were retrospectively collected from a tertiary referral hospital between July 2022 and December 2022 and used to evaluate malignancy diagnoses generated by three LLMs-OpenAI's ChatGPT 3.5, ChatGPT 4.0, and Google's Bard. Inter- and intra-LLM agreement of diagnosis were evaluated. Then, diagnostic performance, including accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC), was evaluated and compared for the LLMs and three interactive approaches: human reader combined with LLMs, image-to-text model combined with LLMs, and an end-to-end convolutional neural network model. Results A total of 1161 US images of thyroid nodules (498 benign, 663 malignant) from 725 patients (mean age, 42.2 years ± 14.1 [SD]; 516 women) were evaluated. ChatGPT 4.0 and Bard displayed substantial to almost perfect intra-LLM agreement (κ range, 0.65-0.86 [95% CI: 0.64, 0.86]), while ChatGPT 3.5 showed fair to substantial agreement (κ range, 0.36-0.68 [95% CI: 0.36, 0.68]). ChatGPT 4.0 had an accuracy of 78%-86% (95% CI: 76%, 88%) and sensitivity of 86%-95% (95% CI: 83%, 96%), compared with 74%-86% (95% CI: 71%, 88%) and 74%-91% (95% CI: 71%, 93%), respectively, for Bard. Moreover, with ChatGPT 4.0, the image-to-text-LLM strategy exhibited an AUC (0.83 [95% CI: 0.80, 0.85]) and accuracy (84% [95% CI: 82%, 86%]) comparable to those of the human-LLM interaction strategy with two senior readers and one junior reader and exceeding those of the human-LLM interaction strategy with one junior reader. Conclusion LLMs, particularly integrated with image-to-text approaches, show potential in enhancing diagnostic medical imaging. ChatGPT 4.0 was optimal for consistency and diagnostic accuracy when compared with Bard and ChatGPT 3.5. © RSNA, 2024 Supplemental material is available for this article.

US-based Sequential Algorithm Integrating an AI Model for Advanced Liver Fibrosis Screening.

Background Noninvasive tests can be used to screen patients with chronic liver disease for advanced liver fibrosis; however, the use of single tests may not be adequate. Purpose To construct sequential clinical algorithms that include a US deep learning (DL) model and compare their ability to predict advanced liver fibrosis with that of other noninvasive tests. Materials and Methods This retrospective study included adult patients with a history of chronic liver disease or unexplained abnormal liver function test results who underwent B-mode US of the liver between January 2014 and September 2022 at three health care facilities. A US-based DL network (FIB-Net) was trained on US images to predict whether the shear-wave elastography (SWE) value was 8.7 kPa or higher, indicative of advanced fibrosis. In the internal and external test sets, a two-step algorithm (Two-step#1) using the Fibrosis-4 Index (FIB-4) followed by FIB-Net and a three-step algorithm (Three-step#1) using FIB-4 followed by FIB-Net and SWE were used to simulate screening scenarios where liver stiffness measurements were not or were available, respectively. Measures of diagnostic accuracy were calculated using liver biopsy as the reference standard and compared between FIB-4, SWE, FIB-Net, and European Association for the Study of the Liver guidelines (ie, FIB-4 followed by SWE), along with sequential algorithms. Results The training, validation, and test data sets included 3067 (median age, 42 years [IQR, 33-53 years]; 2083 male), 1599 (median age, 41 years [IQR, 33-51 years]; 1124 male), and 1228 (median age, 44 years [IQR, 33-55 years]; 741 male) patients, respectively. FIB-Net obtained a noninferior specificity with a margin of 5% (P < .001) compared with SWE (80% vs 82%). The Two-step#1 algorithm showed higher specificity and positive predictive value (PPV) than FIB-4 (specificity, 79% vs 57%; PPV, 44% vs 32%) while reducing unnecessary referrals by 42%. The Three-step#1 algorithm had higher specificity and PPV compared with European Association for the Study of the Liver guidelines (specificity, 94% vs 88%; PPV, 73% vs 64%) while reducing unnecessary referrals by 35%. Conclusion A sequential algorithm combining FIB-4 and a US DL model showed higher diagnostic accuracy and improved referral management for all-cause advanced liver fibrosis compared with FIB-4 or the DL model alone. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Ghosh in this issue.

Shear-wave elastography combined with contrast-enhanced ultrasound algorithm for noninvasive characterization of focal liver lesions.

PURPOSE: To establish shear-wave elastography (SWE) combined with contrast-enhanced ultrasound (CEUS) algorithm (SCCA) and improve the diagnostic performance in differentiating focal liver lesions (FLLs). MATERIAL AND METHODS: We retrospectively selected patients with FLLs between January 2018 and December 2019 at the First Affiliated Hospital of Sun Yat-sen University. Histopathology was used as a standard criterion except for hemangiomas and focal nodular hyperplasia. CEUS with SonoVue (Bracco Imaging) and SCCA combining CEUS and maximum value of elastography with < 20 kPa and > 90 kPa thresholds were used for the diagnosis of FLLs. The diagnostic performance of CEUS and SCCA was calculated and compared. RESULTS: A total of 171 FLLs were included, with 124 malignant FLLs and 47 benign FLLs. The area under curve (AUC), sensitivity, and specificity in detecting malignant FLLs were 0.83, 91.94%, and 74.47% for CEUS, respectively, and 0.89, 91.94%, and 85.11% for SCCA, respectively. The AUC of SCCA was significantly higher than that of CEUS (P = 0.019). Decision curves indicated that SCCA provided greater clinical benefits. The SCCA provided significantly improved prediction of clinical outcomes, with a net reclassification improvement index of 10.64% (P = 0.018) and integrated discrimination improvement of 0.106 (P = 0.019). For subgroup analysis, we divided the FLLs into a chronic-liver-disease group (n = 88 FLLs) and a normal-liver group (n = 83 FLLs) according to the liver background. In the chronic-liver-disease group, there were no differences between the CEUS-based and SCCA diagnoses. In the normal-liver group, the AUC of SCCA and CEUS in the characterization of FLLs were 0.89 and 0.83, respectively (P = 0.018). CONCLUSION: SCCA is a feasible tool for differentiating FLLs in patients with normal liver backgrounds. Further investigations are necessary to validate the universality of this algorithm.

Reproducibility of radiomics features from ultrasound images: influence of image acquisition and processing.

OBJECTIVES: To systematically assess the reproducibility of radiomics features from ultrasound (US) images during image acquisition and processing. MATERIALS AND METHODS: A standardized phantom was scanned to obtain US images. Reproducibility of radiomics features from US images, also known as ultrasomics features, was explored via (a) intra-US machine: changing the US acquisition parameters including gain, focus, and frequency; (b) inter-US machine: comparing three different scanners; (c) changing segmentation locations; and (d) inter-platform: comparing features extracted by the Ultrasomics and PyRadiomics algorithm platforms. Reproducible ultrasomics features were selected based on coefficients of variation. RESULTS: A total of 108 US images from three scanners were obtained; 5253 ultrasomics features including seven categories of features were extracted and evaluated for each US image. From intra-US machine analysis, 37.0-38.8% of features showed good reproducibility. From inter-US machine analysis, 42.8% (2248/5253) of features exhibited good reproducibility. From segmentation location analysis, 55.7-57.6% of features showed good reproducibility. No significant difference in the normalized feature ranges was found between the 100 features extracted by the Ultrasomics and PyRadiomics platforms with the same algorithm (p = 0.563). A total of 1452 (27.6%) ultrasomics features were reproducible whenever intra-/inter-US machine or segmentation location were changed, most of which were wavelet and shearlet features. CONCLUSIONS: Different acquisition parameters, US scanners, segmentation locations, and feature extraction platforms affected the reproducibility of ultrasomics features. Wavelet and shearlet features showed the best reproducibility across all procedures. KEY POINTS: • Different acquisition parameters, US scanners, segmentation locations, and feature extraction platforms affected the reproducibility of ultrasomics features. • A total of 1452 (27.6%) ultrasomics features were reproducible whenever intra-/inter-US machine or segmentation location were changed. • Wavelet and shearlet features showed the best reproducibility across all procedures.

Performance and comparison of artificial intelligence and human experts in the detection and classification of colonic polyps.

OBJECTIVE: The main aim of this study was to analyze the performance of different artificial intelligence (AI) models in endoscopic colonic polyp detection and classification and compare them with doctors with different experience. METHODS: We searched the studies on Colonoscopy, Colonic Polyps, Artificial Intelligence, Machine Learning, and Deep Learning published before May 2020 in PubMed, EMBASE, Cochrane, and the citation index of the conference proceedings. The quality of studies was assessed using the QUADAS-2 table of diagnostic test quality evaluation criteria. The random-effects model was calculated using Meta-DISC 1.4 and RevMan 5.3. RESULTS: A total of 16 studies were included for meta-analysis. Only one study (1/16) presented externally validated results. The area under the curve (AUC) of AI group, expert group and non-expert group for detection and classification of colonic polyps were 0.940, 0.918, and 0.871, respectively. AI group had slightly lower pooled specificity than the expert group (79% vs. 86%, P < 0.05), but the pooled sensitivity was higher than the expert group (88% vs. 80%, P < 0.05). While the non-experts had less pooled specificity in polyp recognition than the experts (81% vs. 86%, P < 0.05), and higher pooled sensitivity than the experts (85% vs. 80%, P < 0.05). CONCLUSION: The performance of AI in polyp detection and classification is similar to that of human experts, with high sensitivity and moderate specificity. Different tasks may have an impact on the performance of deep learning models and human experts, especially in terms of sensitivity and specificity.

Identification of the origin of tumor in vein: comparison between CEUS LI-RADS v2017 and v2016 for patients at high risk.

OBJECTIVES: To compare the diagnostic performance of the Contrast-Enhanced Ultrasound (CEUS) Liver Imaging Report and Data System (LI-RADS) v2016 and v2017 in identifying the origin of tumor in vein (TIV). METHODS: From April 2014 to December 2018, focal liver lesions (FLLs) accompanied by TIV formation in patients at high risk for hepatocellular carcinoma (HCC) were enrolled. Histologic evaluation or composite imaging reference standard were served as the reference standard. Each case was categorized according to the CEUS LI-RADS v2016 and v2017, respectively. Diagnostic performance of CEUS LI-RADS v2016 and v2017 in identifying the originated tumor of TIV was validated via sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value. RESULTS: A total of 273 FLLs with TIV were analyzed finally, including 266 HCCs and 7 non-HCCs. In v2016, when adopting all TIV as LR-5V, the accuracy and PPV in identifying the originated tumor were both 97.4%. In v2017, when assigning TIV according to contiguous FLLs CEUS LI-RADS category, the accuracy and PPV were 61.9% and 99.4% in subclass of LR-5 as the diagnostic criteria of HCC, and 64.1% and 99.4% in subclass of LR-4/5 as the criteria of HCC diagnosis. There were significant differences in diagnostic accuracy between CEUS LI-RADS v2016 and v2017 in identifying the originated tumor of TIV (p < 0.001). CONCLUSIONS: CEUS LI-RADS v2016 could be better than v2017 in identifying the originated tumor of TIV.

Contrast-Enhanced Ultrasound-Based Nomogram: A Potential Predictor of Individually Postoperative Early Recurrence for Patients With Combined Hepatocellular-Cholangiocarcinoma.

PURPOSES: To evaluate the postsurgical prognostic implication of contrast-enhanced ultrasound (CEUS) for combined hepatocellular-cholangiocarcinoma (CHC). To build a CEUS-based early recurrence prediction classifier for CHC, in comparison with tumor-node-metastasis (TNM) staging. METHODS: The CEUS features and clinicopathological findings of each case were analyzed, and the Liver Imaging Reporting and Data System categories were assigned. The recurrence-free survival associated factors were evaluated by Cox proportional hazard model. Incorporating the independent factors, nomograms were built to estimate the possibilities of 3-month, 6-month, and 1-year recurrence and whose prognostic value was determined by time-dependent receiver operating characteristics, calibration curves, and hazard layering efficiency validation, comparing with TNM staging system. RESULTS: In the multivariable analysis, the levels of carbohydrate antigen 19-9, prothrombin time and total bilirubin, and tumor shape, the Liver Imaging Reporting and Data System category were independent factors for recurrence-free survival. The LR-M category showed longer recurrence-free survival than did the LR-4/5 category. The 3-month, 6-month, and 1-year area under the curves of the CEUS-clinical nomogram, clinical nomogram, and TNM staging system were 0.518, 0.552, and 0.843 versus 0.354, 0.240, and 0.624 (P = .048, .049, and .471) vs. 0.562, 0.545, and 0.843 (P = .630, .564, and .007), respectively. The calibration curves of the CEUS-clinical model at different prediction time pionts were all close to the ideal line. The CEUS-clinical model effectively stratified patients into groups of high and low risk of recurrence in both training and validation set, while the TNM staging system only works on the training set. CONCLUSIONS: Our CEUS-clinical nomogram is a reliable early recurrence prediction tool for hepatocellular-cholangiocarcinoma and helps postoperative risk stratification.

Using new criteria to improve the differentiation between HCC and non-HCC malignancies: clinical practice and discussion in CEUS LI-RADS 2017.

PURPOSE: Using contrast-enhanced ultrasound (CEUS) to evaluate the diagnostic performance of liver imaging reporting and data system (LI-RADS) version 2017 and to explore potential ways to improve the efficacy. METHODS: A total of 315 nodules were classified as LR-1 to LR-5, LR-M, and LR-TIV. New criteria were applied by adjusting the early washout onset (< 45 s) and the time of marked washout (within 3 min). Two subgroups of the LR-M nodules were recategorized as LR-5, respectively. The diagnostic performance was evaluated by calculating the accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). RESULTS: By adjusting early washout onset to < 45 s, the LR-5 as a standard for diagnosing HCC had an improved sensitivity (74.1% vs. 56.1%, P < 0.001) without significant change in PPV (93.3% vs. 96.1%, P = 0.267), but the specificity was decreased (48.3% vs. 78.5%, P = 0.018). The LR-M as a standard for the diagnosis of non-HCC malignancies had an increase in specificity (89.2% vs. 66.2%, P < 0.001) but a decrease in sensitivity (31.5% vs. 68.4%, P = 0.023). After reclassification according to the time of marked washout, the sensitivity of the LR-5 increased (80% vs. 56.1%, P < 0.001) without a change in PPV (94.9% vs. 96.1%, P = 0.626) and specificity (80% vs. 78.5%, P = 0.879). For reclassified LR-M nodules, the specificity increased (87.5% versus 66.2%, P < 0.001) with a non-significant decrease in sensitivity (47.3% vs. 68.4%, P = 0.189). CONCLUSIONS: The CEUS LI-RADS showed good confidence in diagnosing HCC while tended to misdiagnose HCC as non-HCC malignancies. Adjusting the marked washout time within 3 min would reduce the possibility of this misdiagnosis.

Pathological considerations of CEUS LI-RADS: correlation with fibrosis stage and tumour histological grade.

OBJECTIVES: To evaluate the influence of pathological factors, such as fibrosis stage and histological grade, on the Liver Imaging Reporting and Data System (LI-RADS) v2017 category of contrast-enhanced ultrasonography (CEUS) in patients with high risk of hepatocellular carcinoma (HCC). MATERIALS AND METHODS: Between June 2015 and December 2016, 441 consecutive patients at high risk of HCC with 460 pathologically proven HCCs were enrolled in this retrospective study. All patients underwent a CEUS examination. The major features (arterial phase hyperenhancement, late and mild washout) were assessed, and LI-RADS categories were assigned according to CEUS LI-RADS v2017. CEUS LI-RADS categories and major features were compared in different histological grades and fibrosis stages. RESULTS: The CEUS LR-5 category was more frequently assigned in the low-grade group (151/280) than in the high-grade group (66/159) (p = 0.013), whereas the LR-TIV category was more frequently assigned in the high-grade group (36/159) than in the low-grade group (40/280) (p = 0.035). CEUS LI-RADS category was not significantly different among different fibrosis stages (p ≥ 0.05). Arterial phase hyperenhancement (APHE) and the hepatic fibrosis stage showed a significant correlation in HCCs ≥ 2 cm and the low-grade group (p = 0.027 and p = 0.003, respectively). No major features of CEUS LI-RADS showed statistically significant differences between the low- and high-grade groups (p ≥ 0.05). CONCLUSION: Hepatic fibrosis stage can influence APHE but showed no impact on the CEUS LI-RADS classification, whereas the histological grade of HCC influenced the LR-5 and LR-TIV categories. KEY POINTS: • Histological grade influenced CEUS LR-5 and LR-TIV category (p = 0.013 and p = 0.035 respectively). Low-grade HCCs occurred more frequently in LR-5 category whereas high-grade HCCs occurred more frequently in LR-TIV category. • Fibrosis stage shows significant influence on APHE on HCCs of the size ≥ 2 cm and low-grade group (p = 0.027 and p = 0.003, respectively). • Hepatic fibrosis stage and HCC histological grade exhibited limited impact on CEUS LI-RADS. CEUS LI-RADS may be feasible for diagnosing HCC in patients regardless of histological grade and fibrosis stage.

Inter-reader agreement of CEUS LI-RADS among radiologists with different levels of experience.

OBJECTIVES: To investigate the inter-reader agreement of contrast-enhanced ultrasound (CEUS) of Liver Imaging Reporting and Data System version 2017 (LI-RADS v2017) categories among radiologists with different levels of experience. MATERIALS AND METHODS: From January 2014 to December 2014, a total of 326 patients at high risk of hepatocellular carcinoma (HCC) who underwent CEUS were included in this retrospective study. All lesions were classified according to LI-RADS v2017 by six radiologists with different levels of experiences: two residents, two fellows, and two specialists. Kappa coefficient was used to assess consistency of LI-RADS categories and major features among radiologists with different levels of experience. The diagnostic performance of HCC was described by accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC). RESULTS: Inter-reader agreement among radiologists of different experience levels was substantial agreement for arterial phase hyperenhancement, washout appearance, and early or late washout. Inter-reader agreement for LI-RADS categories was moderate to substantial. When LR-5 was used as criteria to determinate HCC, the AUC of LI-RADS for HCC was 0.67 for residents, 0.72 for fellows, and 0.78 for specialist radiologists. When compared between residents and specialists, accuracy, sensitivity, and AUC were significantly different (all p < 0.05). However, there were no significant differences in specificity, PPV, and NPV between the two groups. CONCLUSION: CEUS LI-RADS showed good diagnostic consistency among radiologists with different levels of experience, and consistency increased with experience levels. KEY POINTS: • The inter-reader agreement for LI-RADS categories was moderate to substantial agreement (κ, 0.60-0.80). • When compared between residents and specialists, accuracy, sensitivity, and AUC showed significantly different (all p < 0.05). However, there were no significant differences for specificity, PPV, and NPV between these two groups. • Among the radiologists with more than 1 year of experience, there was no significant difference in the diagnostic performance of HCC, suggesting that CEUS LI-RADS is a good standardized categorization system for high-risk patients.

Artificial intelligence assists identifying malignant versus benign liver lesions using contrast-enhanced ultrasound.

BACKGROUND AND AIM: This study aims to construct a strategy that uses assistance from artificial intelligence (AI) to assist radiologists in the identification of malignant versus benign focal liver lesions (FLLs) using contrast-enhanced ultrasound (CEUS). METHODS: A training set (patients = 363) and a testing set (patients = 211) were collected from our institute. On four-phase CEUS images in the training set, a composite deep learning architecture was trained and tuned for differentiating malignant and benign FLLs. In the test dataset, AI performance was evaluated by comparison with radiologists with varied levels of experience. Based on the comparison, an AI assistance strategy was constructed, and its usefulness in reducing CEUS interobserver heterogeneity was further tested. RESULTS: In the test set, to identify malignant versus benign FLLs, AI achieved an area under the curve of 0.934 (95% CI 0.890-0.978) with an accuracy of 91.0%. Comparing with radiologists reviewing videos along with complementary patient information, AI outperformed residents (82.9-84.4%, P = 0.038) and matched the performance of experts (87.2-88.2%, P = 0.438). Due to the higher positive predictive value (PPV) (AI: 95.6% vs residents: 88.6-89.7%, P = 0.056), an AI strategy was defined to improve the malignant diagnosis. With the assistance of AI, radiologists exhibited a sensitivity improvement of 97.0-99.4% (P < 0.05) and an accuracy of 91.0-92.9% (P = 0.008-0.189), which was comparable with that of the experts (P = 0.904). CONCLUSIONS: The CEUS-based AI strategy improved the performance of residents and reduced CEUS's interobserver heterogeneity in the differentiation of benign and malignant FLLs.

Preliminary investigation of the diagnostic value of shear wave elastography in evaluating the testicular spermatogenic function in patients with azoospermia.

To assess the diagnostic value of shear wave elastography (SWE) for evaluating the histological spermatogenic function of azoospermic males, 91 patients with azoospermia who underwent standardised greyscale ultrasound and SWE examinations followed by testicular biopsy were retrospectively recruited. Spermatogenic function was classified by biopsy as normal testicular spermatogenesis (n = 61), hypospermatogenesis (n = 18), spermatogenesis arrest (n = 6) and Sertoli cell-only syndrome (n = 6). Significant differences in testicular size and SWE values were observed between these 4 groups (p < .01). The mean SWE value had good discrimination power (AUC = 0.79) with a cut-off value of 1.55 KPa, a sensitivity of 0.58, specificity of 0.85, positive predictive value (PPV) of 0.36 and negative predictive value (NPV) of 0.93. Testicular volume had an AUC of 0.75. With a cut-off value of 8.41 ml, the testicular volume had a sensitivity of 0.58, specificity of 0.92, PPV of 0.54 and NPV of 0.93. The mean SWE value and testicular volume efficiently discriminated patients with normal spermatogenesis and hypospermatogenesis from patients with Sertoli cell-only syndrome and spermatogenesis arrest.

Testicular quantitative ultrasound: A noninvasive monitoring method for evaluating spermatogenic function in busulfan-induced testicular injury mouse models.

Busulfan-induced testicular injury mouse models are commonly used for experiments on spermatogonial stem cell transplantation, treatments for azoospermia due to spermatogenic failure and preserving male fertility after chemotherapy. Here, we investigated the value of testicular quantitative ultrasound for evaluating spermatogenic function in this model. In this study, testicular ultrasound was performed on mice from day 0 to 126 after busulfan treatment (n = 48), and quantitative data, including the testicular volume, mean pixel intensity and pixel uniformity, were analysed. The results revealed that from day 0 to 36, the testicular volume was positively associated with the testicle-to-body weight ratio (r = .92). On day 63, the pixel uniformity, which remained stable from day 0 to 36, declined significantly compared with that on day 36 (p < .01). On day 126, when the whole progression of spermatogenesis could be observed in most tubules, the mean pixel intensity also returned to normal (p > .05). In conclusion, testicular quantitative ultrasound could be used as a noninvasive and accurate monitoring method for evaluating spermatogenic function in busulfan-induced testicular injury mouse models.

CT-based radiomics scores predict response to neoadjuvant chemotherapy and survival in patients with gastric cancer.

BACKGROUND: Neoadjuvant chemotherapy is a promising treatment option for potential resectable gastric cancer, but patients' responses vary. We aimed to develop and validate a radiomics score (rad_score) to predict treatment response to neoadjuvant chemotherapy and to investigate its efficacy in survival stratification. METHODS: A total of 106 patients with neoadjuvant chemotherapy before gastrectomy were included (training cohort: n = 74; validation cohort: n = 32). Radiomics features were extracted from the pre-treatment portal venous-phase CT. After feature reduction, a rad_score was established by Randomised Tree algorithm. A rad_clinical_score was constructed by integrating the rad_score with clinical variables, so was a clinical score by clinical variables only. The three scores were validated regarding their discrimination and clinical usefulness. The patients were stratified into two groups according to the score thresholds (updated with post-operative clinical variables), and their survivals were compared. RESULTS: In the validation cohort, the rad_score demonstrated a good predicting performance in treatment response to the neoadjuvant chemotherapy (AUC [95% CI] =0.82 [0.67, 0.98]), which was better than the clinical score (based on pre-operative clinical variables) without significant difference (0.62 [0.42, 0.83], P = 0.09). The rad_clinical_score could not further improve the performance of the rad_score (0.70 [0.51, 0.88], P = 0.16). Based on the thresholds of these scores, the high-score groups all achieved better survivals than the low-score groups in the whole cohort (all P < 0.001). CONCLUSION: The rad_score that we developed was effective in predicting treatment response to neoadjuvant chemotherapy and in stratifying patients with gastric cancer into different survival groups. Our proposed strategy is useful for individualised treatment planning.

Differential diagnosis between hepatic alveolar echinococcosis and intrahepatic cholangiocarcinoma with conventional ultrasound and contrast-enhanced ultrasound.

BACKGROUND: Misclassifications of hepatic alveolar echinococcosis (HAE) as intrahepatic cholangiocarcinoma (ICC) may lead to inappropriate treatment strategies. The aim of this study was to explore the differential diagnosis with conventional ultrasound and contrast-enhanced ultrasound (CEUS). METHODS: Sixty HAE lesions with 60 propensity score-matched ICC lesions were retrospectively collected. The 120 lesions were randomly divided into a training set (n = 80) and a testing set (n = 40). In the training set, the most useful independent conventional ultrasound and CEUS features was selected for differentiating between HAE and ICC. Then, a simplified US scoring system for diagnosing HAE was constructed based on selected features with weighted coefficients. The constructed US score for HAE was validated in both the training set and the testing set, and diagnostic performance was evaluated. RESULTS: Compared with ICC lesions, HAE lesions were mostly located in the right lobe and had mixed echogenicity, a pseudocystic appearance and foci calcifications on conventional ultrasound. On CEUS, HAE lesions showed more regular rim-like enhancement than ICC lesions and had late washout with a long enhancement duration. The simplified US score consisted of echogenicity, pseudocystic/calcification, bile duct dilatation, enhancement pattern, enhancement duration, and marked washout. In the testing set, the sensitivity, specificity, LR+, LR- and the area under the ROC curve for the score to differentiate HAE from ICC were 80.0, 81.3%, 4.27, 0.25 and 0.905, respectively. CONCLUSIONS: The US score based on typical features from both conventional ultrasound and CEUS could accurately differentiate HAE from ICC.

Malignancy risk stratification and FNA recommendations for thyroid nodules: A comparison of ACR TI-RADS, AACE/ACE/AME and ATA guidelines.

OBJECTIVE: To compare diagnostic performance and malignancy risk stratification among guidelines set forth by the American Thyroid Association (ATA) in 2015, the American Association of Clinical Endocrinologists (AACE), the American College of Endocrinology (ACE) and the Association Medici Endocrinologi (AME) in 2016, and the American College of Radiology (ACR) in 2017. METHODS: The retrospective study was approved by the hospital ethics committee, and the informed consent requirement was waived. From October 2015 to March 2016, a total of 230 patients with 230 consecutive thyroid nodules were enrolled in this study. Each nodule was classified by one junior and one senior radiologist separately according to ACR TI-RADS, AACE/ACE/AME and ATA guidelines. The malignancy diagnostic performance and the number of FNA recommendations were pairwise compared among three guidelines using chi-square tests. RESULTS: Of the 230 thyroid nodules, 137 were malignant, and 93 were benign. However, 19.6% of the nodules (45 of 230) did not match any pattern using the ATA guidelines but with a high risk of malignancy (68.9%). The ACR TI-RADS derived the highest diagnostic performance, from both junior radiologist (AUC 0.815) and senior radiologist (AUC 0.864). The ACR guidelines also showed the greatest level of sensitivity (junior: 86.1%, senior: 94.9%), compared with AACE/ACE/AME and ATA guidelines. The number of thyroid nodules recommended to fine-needle aspiration (FNA) was the lowest (37.8%, 40.4%) by ACR TI-RADS, and meanwhile, the malignant detection rate within these nodules was highest (64.4%, 68.8%). CONCLUSIONS: The ACR guidelines present a higher level of diagnostic indicators and may offer a meaningful reduction in FNA recommendations with a higher malignancy detection rate.

CT-based radiomics for preoperative prediction of early recurrent hepatocellular carcinoma: technical reproducibility of acquisition and scanners.

PURPOSE: To test the technical reproducibility of acquisition and scanners of CT image-based radiomics model for early recurrent hepatocellular carcinoma (HCC). METHODS: We included primary HCC patient undergone curative therapies, using early recurrence as endpoint. Four datasets were constructed: 109 images from hospital #1 for training (set 1: 1-mm image slice thickness), 47 images from hospital #1 for internal validation (sets 2 and 3: 1-mm and 10-mm image slice thicknesses, respectively), and 47 images from hospital #2 for external validation (set 4: vastly different from training dataset). A radiomics model was constructed. Radiomics technical reproducibility was measured by overfitting and calibration deviation in external validation dataset. The influence of slice thickness on reproducibility was evaluated in two internal validation datasets. RESULTS: Compared with set 1, the model in set 2 indicated favorable prediction efficiency (the area under the curve 0.79 vs. 0.80, P = 0.47) and good calibration (unreliability statistic U: P = 0.33). However, in set 4, significant overfitting (0.63 vs. 0.80, P < 0.01) and calibration deviation (U: P < 0.01) were observed. Similar poor performance was also observed in set 3 (0.56 vs. 0.80, P = 0.02; U: P < 0.01). CONCLUSIONS: CT-based radiomics has poor reproducibility between centers. Image heterogeneity, such as slice thickness, can be a significant influencing factor.

Multiparametric ultrasomics of significant liver fibrosis: A machine learning-based analysis.

OBJECTIVE: To assess significant liver fibrosis by multiparametric ultrasomics data using machine learning. MATERIALS AND METHODS: This prospective study consisted of 144 patients with chronic hepatitis B. Ultrasomics-high-throughput quantitative data from ultrasound imaging of liver fibrosis-were generated using conventional radiomics, original radiofrequency (ORF) and contrast-enhanced micro-flow (CEMF) features. Three categories of features were explored using pairwise correlation and hierarchical clustering. Features were selected using diagnostic tests for fibrosis, activity and steatosis stage, with the histopathological results as the reference. The fibrosis staging performance of ultrasomics models with combinations of the selected features was evaluated with machine-learning algorithms by calculating the area under the receiver-operator characteristic curve (AUC). RESULTS: ORF and CEMF features had better predictive power than conventional radiomics for liver fibrosis stage (both p < 0.01). CEMF features exhibited the highest diagnostic value for activity stage (both p < 0.05), and ORF had the best diagnostic value for steatosis stage (both p < 0.01). The machine-learning classifiers of adaptive boosting, random forest and support vector machine were found to be optimal algorithms with better (all mean AUCs = 0.85) and more stable performance (coefficient of variation = 0.01-0.02) for fibrosis staging than decision tree, logistic regression and neural network (mean AUC = 0.61-0.72, CV = 0.07-0.08). The multiparametric ultrasomics model achieved much better performance (mean AUC values of 0.78-0.85) than the features from a single modality in discriminating significant fibrosis (≥ F2). CONCLUSION: Machine-learning-based analysis of multiparametric ultrasomics can help improve the discrimination of significant fibrosis compared with mono or dual modalities. KEY POINTS: • Multiparametric ultrasomics has achieved much better performance in the discrimination of significant fibrosis (≥ F2) than the single modality of conventional radiomics, original radiofrequency and contrast-enhanced micro-flow. • Adaptive boosting, random forest and support vector machine are the optimal algorithms for machine learning.

Ultrasound-based radiomics score: a potential biomarker for the prediction of microvascular invasion in hepatocellular carcinoma.

PURPOSE: To develop an ultrasound (US)-based radiomics score for preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC). METHODS: Between January 1, 2012, and October 31, 2017, a total of 482 HCC patients who underwent contrast-enhanced ultrasound (CEUS) were retrospectively reviewed. The study population was divided into a training cohort (n = 341) and a validation cohort (n = 141) based on a cutoff time of January 1, 2016. Radiomics features were extracted from the grayscale US images of HCC. After features selection, a radiomics score was developed from the training cohort. The incremental value of the radiomics score to the clinic-pathological factors for MVI prediction was assessed in the validation cohort with respect to discrimination, calibration, and clinical usefulness. RESULTS: The US-based radiomics score consisted of six selected features. Multivariate logistic regression analysis showed that the radiomics score, alpha-fetoprotein (AFP), and tumor size were independent predictors of MVI. The radiomics nomogram (based on the three factors) showed better performance for MVI detection (area under the curve [AUC] 0.731[0.647, 0.815] than the clinical nomogram (based on AFP and tumor size) (0.634 [0.543, 0.724]) (p = 0.015). Both nomograms showed good calibration. Decision curve analysis demonstrated that in terms of clinical usefulness, the radiomics nomogram outperformed the clinical nomogram. CONCLUSION: The US-based radiomics score was an independent predictor of MVI in HCC. Combining the radiomics score with clinical factors improved the prediction efficacy. KEY POINTS: • Radiomics can be applied in US images. • US-based radiomics score was an independent predictor of MVI. • Radiomics nomogram incorporated with the radiomics score showed good performance for MVI prediction.

Comparison between M-score and LR-M in the reporting system of contrast-enhanced ultrasound LI-RADS.

OBJECTIVE: To develop a contrast-enhanced ultrasound (CEUS) M-score and compare it with LR-M in CEUS Liver Imaging Reporting and Data System (LI-RADS). METHODS: We retrospectively enrolled 105 consecutive high-risk patients with hepatocellular carcinoma (HCC) and 105 with intrahepatic cholangiocarcinoma (ICC). The subjects were selected by propensity score matching between November 2003 and December 2017. A CEUS M-score for predicting ICC was constructed based on specific CEUS features by the least absolute shrinkage and selection operator regularised regression. M-score was used to develop a modified CEUS LI-RADS. The diagnostic performance of the modified CEUS LI-RADS using M-score for diagnosing HCC and ICC was compared with American College of Radiology (ACR) CEUS LI-RADS using LR-M. RESULTS: The most useful features for ICC were as follows: poorly circumscribed (69.52%), rim enhancement (63.81%), early washout (92.38%), intratumoural vein (56.19%), obscure boundary of intratumoural non-enhanced area (57.14%), and marked washout (59.05%, all p < 0.001). For predicting ICC, the M-score had a higher specificity (88.57% vs. 63.81%) with lower sensitivity (89.52% vs. 95.24%) compared with LR-M. For diagnosing HCC, the sensitivity of modified LI-RADS (80.95%) was much higher than that of ACR LI-RADS (57.14%), but the specificity was lower (90.48% vs. 96.19%). The area under the curve (AUC) of modified LI-RADS (0.857) was much higher than that of ACR LI-RADS (0.767, p = 0.0001). The modified positive predictive value (PPV) of ACR LI-RADS and modified LI-RADS were 99.42% and 98.99%, respectively. CONCLUSIONS: The modified LI-RADS with M-score had higher sensitivity for diagnosing HCC and higher specificity for diagnosing ICC than ACR LI-RADS. KEY POINTS: • For predicting ICC, the M-score had a higher specificity (88.57% vs. 63.81%) with lower sensitivity (89.52% vs. 95.24%) compared with LR-M. • A CEUS M-score for predicting ICC consisted of more detailed CEUS features (poorly circumscribed, rim enhancement, early washout, intratumoural vein, obscure boundary of intratumoural non-enhanced area, and marked washout) was constructed. • For diagnosing HCC, the sensitivity of modified LI-RADS (80.95%) was much higher than that of ACR LI-RADS (57.14%), but the specificity was lower (90.48% vs. 96.19%). The modified positive predictive value (PPV) of ACR LI-RADS and modified LI-RADS were 99.42% and 98.99%, respectively.