A deep learning fusion network trained with clinical and high-frequency ultrasound images in the multi-classification of skin diseases in comparison with dermatologists: a prospective and multicenter study.

Background: Clinical appearance and high-frequency ultrasound (HFUS) are indispensable for diagnosing skin diseases by providing internal and external information. However, their complex combination brings challenges for primary care physicians and dermatologists. Thus, we developed a deep multimodal fusion network (DMFN) model combining analysis of clinical close-up and HFUS images for binary and multiclass classification in skin diseases. Methods: Between Jan 10, 2017, and Dec 31, 2020, the DMFN model was trained and validated using 1269 close-ups and 11,852 HFUS images from 1351 skin lesions. The monomodal convolutional neural network (CNN) model was trained and validated with the same close-up images for comparison. Subsequently, we did a prospective and multicenter study in China. Both CNN models were tested prospectively on 422 cases from 4 hospitals and compared with the results from human raters (general practitioners, general dermatologists, and dermatologists specialized in HFUS). The performance of binary classification (benign vs. malignant) and multiclass classification (the specific diagnoses of 17 types of skin diseases) measured by the area under the receiver operating characteristic curve (AUC) were evaluated. This study is registered with www.chictr.org.cn (ChiCTR2300074765). Findings: The performance of the DMFN model (AUC, 0.876) was superior to that of the monomodal CNN model (AUC, 0.697) in the binary classification (P = 0.0063), which was also better than that of the general practitioner (AUC, 0.651, P = 0.0025) and general dermatologists (AUC, 0.838; P = 0.0038). By integrating close-up and HFUS images, the DMFN model attained an almost identical performance in comparison to dermatologists (AUC, 0.876 vs. AUC, 0.891; P = 0.0080). For the multiclass classification, the DMFN model (AUC, 0.707) exhibited superior prediction performance compared with general dermatologists (AUC, 0.514; P = 0.0043) and dermatologists specialized in HFUS (AUC, 0.640; P = 0.0083), respectively. Compared to dermatologists specialized in HFUS, the DMFN model showed better or comparable performance in diagnosing 9 of the 17 skin diseases. Interpretation: The DMFN model combining analysis of clinical close-up and HFUS images exhibited satisfactory performance in the binary and multiclass classification compared with the dermatologists. It may be a valuable tool for general dermatologists and primary care providers. Funding: This work was supported in part by the National Natural Science Foundation of China and the Clinical research project of Shanghai Skin Disease Hospital.

A multi-parameter intrahepatic cholangiocarcinoma scoring system based on modified contrast-enhanced ultrasound LI-RADS M criteria for differentiating intrahepatic cholangiocarcinoma from hepatocellular carcinoma.

PURPOSE: To develop a multi-parameter intrahepatic cholangiocarcinoma (ICC) scoring system and compare its diagnostic performance with contrast-enhanced ultrasound (CEUS) liver imaging reporting and data system M (LR-M) criteria for differentiating ICC from hepatocellular carcinoma (HCC). METHODS: This retrospective study enrolled 62 high-risk patients with ICCs and 62 high-risk patients with matched HCCs between January 2022 and December 2022 from two institutions. The CEUS LR-M criteria was modified by adjusting the early wash-out onset (within 45 s) and the marked wash-out (within 3 min). Then, a multi-parameter ICC scoring system was established based on clinical features, B-mode ultrasound features, and modified LR-M criteria. RESULT: We found that elevated CA 19-9 (OR=12.647), lesion boundary (OR=11.601), peripheral rim-like arterial phase hyperenhancement (OR=23.654), early wash-out onset (OR=7.211), and marked wash-out (OR=19.605) were positive predictors of ICC, whereas elevated alpha-fetoprotein (OR=0.078) was a negative predictor. Based on these findings, an ICC scoring system was established. Compared with the modified LR-M and LR-M criteria, the ICC scoring system showed the highest area under the curve (0.911 vs. 0.831 and 0.750, both p<0.05) and specificity (0.935 vs. 0.774 and 0.565, both p<0.05). Moreover, the numbers of HCCs categorized as LR-M decreased from 27 (43.5%) to 14 (22.6%) and 4 (6.5%) using the modified LR-M criteria and ICC scoring system, respectively. CONCLUSION: The modified LR-M criteria-based multi-parameter ICC scoring system had the highest specificity for diagnosing ICC and reduced the number of HCC cases diagnosed as LR-M category.

Quantitative analysis of pre-treatment dynamic contrast-enhanced ultrasound for assessing the response of colorectal liver metastases to chemotherapy plus targeted therapy: a dual-institutional study.

OBJECTIVES: To investigate the clinical value of pre-treatment quantitative contrast-enhanced ultrasound (CEUS) in assessing the response of colorectal liver metastases (CRLM) to chemotherapy plus targeted therapy. METHODS: This study retrospectively enrolled 50 CRLM patients from the Zhongshan Hospital, Fudan University as the training cohort and 14 patients from Shanghai Tenth People's Hospital as the testing cohort. Patients underwent the CEUS examination before receiving chemotherapy (CAPOX, FOLFOX, FOLFIRI, or FOLFOXIRI) plus targeted therapy (Bevacizumab or Cetuximab). The therapy response was determined according to Response Evaluation Criteria in Solid Tumors version 1.1 based on pre-treatment CT and 3-month follow-up CT after therapy. Dynamic analysis was performed by VueBox® software. Time-intensity curves with quantitative perfusion parameters were obtained. In the training cohort, univariable and multivariable logistic regression analyses were used to develop the predictive model of therapy response. The predictive performance of the developed model was validated in the testing cohort. RESULTS: After the logistic regression analyses, the peak enhancement (PE) (odds ratio = 1.640; 95% confidence intervals [CI] 1.022-2.633) and time to peak (TTP) (odds ratio = 0.495; 95% CI 0.246-0.996) were determined as independent predictive factors. PE and TTP generated from VueBox® were not affected by ultrasound instruments and contrast agent dosage in therapy response evaluation (P > 0.05). The logistic regression model achieved satisfactory prediction performance (area under the curve: 0.923 in the training cohort and 0.854 in the testing cohort). CONCLUSION: CEUS with dynamic quantitative perfusion analysis, which presents high consistency, has potential practical value in predicting the response of CRLM to chemotherapy plus targeted therapy.

Response Evaluation Using Contrast-Enhanced Ultrasound for Unresectable Advanced Hepatocellular Carcinoma Treated With Tyrosine Kinase Inhibitors Plus Anti-PD-1 Antibody Therapy.

OBJECTIVE: The aim of the work described here was to evaluate the role of contrast-enhanced ultrasound (CEUS) in response evaluation for unresectable advanced hepatocellular carcinoma (HCC) treated with tyrosine kinase inhibitors (TKIs) plus anti-programmed cell death protein-1 (PD-1) antibody therapy. METHODS: A prospective cohort of consecutive patients with HCC who received combined TKI/anti-PD-1 antibody treatment for unresectable HCC between January 2022 and October 2022 was included in this study. The patients underwent unenhanced ultrasound (US) and CEUS examinations before treatment and at follow-up. Changes in the largest diameters of the target tumor on unenhanced US and the largest diameters of the enhancing target tumors on CEUS were evaluated. Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 with unenhanced US and magnetic resonance imaging/computed tomography (MRI/CT) and modified RECIST (mRECIST) with CEUS and CEMRI/CT were used to assess treatment response. RESULTS: A total of 24 HCC patients (23 men and 1 woman; mean age: 56.5 ± 8.5 y; Barcelona Clinic Liver Cancer stage C, 62.5%; 29 intrahepatic target tumors) were studied. Calculations of degree of necrosis in the target tumors revealed no significant differences between CEUS and CEMRI/CT (44.5 ± 36.2% vs. 45.3 ± 36.8%, p = 0.862). As for the differentiation of responders from non-responders, the agreement between RECIST version 1.1 of unenhanced US and mRECIST-CEUS was poor (κ coefficient = 0.233). Meanwhile, there was a high degree of concordance between mRECIST-CEUS and mRECIST-CEMRI/CT (κ coefficient = 0.812). CONCLUSION: CEUS proved to be superior to baseline US and is comparable to CEMRI/CT in defining treatment outcome for combined TKI/anti-PD-1 antibody therapy.

Risk stratification of gallbladder masses by machine learning-based ultrasound radiomics models: a prospective and multi-institutional study.

OBJECTIVE: This study aimed to evaluate the diagnostic performance of machine learning (ML)-based ultrasound (US) radiomics models for risk stratification of gallbladder (GB) masses. METHODS: We prospectively examined 640 pathologically confirmed GB masses obtained from 640 patients between August 2019 and October 2022 at four institutions. Radiomics features were extracted from grayscale US images and germane features were selected. Subsequently, 11 ML algorithms were separately used with the selected features to construct optimum US radiomics models for risk stratification of the GB masses. Furthermore, we compared the diagnostic performance of these models with the conventional US and contrast-enhanced US (CEUS) models. RESULTS: The optimal XGBoost-based US radiomics model for discriminating neoplastic from non-neoplastic GB lesions showed higher diagnostic performance in terms of areas under the curves (AUCs) than the conventional US model (0.822-0.853 vs. 0.642-0.706, p < 0.05) and potentially decreased unnecessary cholecystectomy rate in a speculative comparison with performing cholecystectomy for lesions sized over 10 mm (2.7-13.8% vs. 53.6-64.9%, p < 0.05) in the validation and test sets. The AUCs of the XGBoost-based US radiomics model for discriminating carcinomas from benign GB lesions were higher than the conventional US model (0.904-0.979 vs. 0.706-0.766, p < 0.05). The XGBoost-US radiomics model performed better than the CEUS model in discriminating GB carcinomas (AUC: 0.995 vs. 0.902, p = 0.011). CONCLUSIONS: The proposed ML-based US radiomics models possess the potential capacity for risk stratification of GB masses and may reduce the unnecessary cholecystectomy rate and use of CEUS. CLINICAL RELEVANCE STATEMENT: The machine learning-based ultrasound radiomics models have potential for risk stratification of gallbladder masses and may potentially reduce unnecessary cholecystectomies. KEY POINTS: • The XGBoost-based US radiomics models are useful for the risk stratification of GB masses. • The XGBoost-based US radiomics model is superior to the conventional US model for discriminating neoplastic from non-neoplastic GB lesions and may potentially decrease unnecessary cholecystectomy rate for lesions sized over 10 mm in comparison with the current consensus guideline. • The XGBoost-based US radiomics model could overmatch CEUS model in discriminating GB carcinomas from benign GB lesions.

Three-dimensional convolutional neural network model to identify clinically significant prostate cancer in transrectal ultrasound videos: a prospective, multi-institutional, diagnostic study.

Background: Identifying patients with clinically significant prostate cancer (csPCa) before biopsy helps reduce unnecessary biopsies and improve patient prognosis. The diagnostic performance of traditional transrectal ultrasound (TRUS) for csPCa is relatively limited. This study was aimed to develop a high-performance convolutional neural network (CNN) model (P-Net) based on a TRUS video of the entire prostate and investigate its efficacy in identifying csPCa. Methods: Between January 2021 and December 2022, this study prospectively evaluated 832 patients from four centres who underwent prostate biopsy and/or radical prostatectomy. All patients had a standardised TRUS video of the whole prostate. A two-dimensional CNN (2D P-Net) and three-dimensional CNN (3D P-Net) were constructed using the training cohort (559 patients) and tested on the internal validation cohort (140 patients) as well as on the external validation cohort (133 patients). The performance of 2D P-Net and 3D P-Net in predicting csPCa was assessed in terms of the area under the receiver operating characteristic curve (AUC), biopsy rate, and unnecessary biopsy rate, and compared with the TRUS 5-point Likert score system as well as multiparametric magnetic resonance imaging (mp-MRI) prostate imaging reporting and data system (PI-RADS) v2.1. Decision curve analyses (DCAs) were used to determine the net benefits associated with their use. The study is registered at https://www.chictr.org.cn with the unique identifier ChiCTR2200064545. Findings: The diagnostic performance of 3D P-Net (AUC: 0.85-0.89) was superior to TRUS 5-point Likert score system (AUC: 0.71-0.78, P = 0.003-0.040), and similar to mp-MRI PI-RADS v2.1 score system interpreted by experienced radiologists (AUC: 0.83-0.86, P = 0.460-0.732) and 2D P-Net (AUC: 0.79-0.86, P = 0.066-0.678) in the internal and external validation cohorts. The biopsy rate decreased from 40.3% (TRUS 5-point Likert score system) and 47.6% (mp-MRI PI-RADS v2.1 score system) to 35.5% (2D P-Net) and 34.0% (3D P-Net). The unnecessary biopsy rate decreased from 38.1% (TRUS 5-point Likert score system) and 35.2% (mp-MRI PI-RADS v2.1 score system) to 32.0% (2D P-Net) and 25.8% (3D P-Net). 3D P-Net yielded the highest net benefit according to the DCAs. Interpretation: 3D P-Net based on a prostate grayscale TRUS video achieved satisfactory performance in identifying csPCa and potentially reducing unnecessary biopsies. More studies to determine how AI models better integrate into routine practice and randomized controlled trials to show the values of these models in real clinical applications are warranted. Funding: The National Natural Science Foundation of China (Grants 82202174 and 82202153), the Science and Technology Commission of Shanghai Municipality (Grants 18441905500 and 19DZ2251100), Shanghai Municipal Health Commission (Grants 2019LJ21 and SHSLCZDZK03502), Shanghai Science and Technology Innovation Action Plan (21Y11911200), and Fundamental Research Funds for the Central Universities (ZD-11-202151), Scientific Research and Development Fund of Zhongshan Hospital of Fudan University (Grant 2022ZSQD07).

Transrectal ultrasound examination of prostate cancer guided by fusion imaging of multiparametric MRI and TRUS: avoiding unnecessary mpMRI-guided targeted biopsy.

The purpose of this study was to explore transrectal ultrasound (TRUS) findings of prostate cancer (PCa) guided by multiparametric magnetic resonance imaging (mpMRI) and to improve the Prostate Imaging Reporting and Data System (PI-RADS) system for avoiding unnecessary mpMRI-guided targeted biopsy (TB). From January 2018 to October 2019, fusion mpMRI and TRUS-guided biopsies were performed in 162 consecutive patients. The study included 188 suspicious lesions on mpMRI in 156 patients, all of whom underwent mpMRI-TRUS fusion imaging-guided TB and 12-core transperineal systematic biopsy (SB). Univariate analyses were performed to investigate the relationship between TRUS features and PCa. Then, logistic regression analysis with generalized estimating equations was performed to determine the independent predictors of PCa and obtain the fitted probability of PCa. The detection rates of PCa based on TB alone, SB alone, and combined SB and TB were 55.9% (105 of 188), 52.6% (82 of 156), and 62.8% (98 of 156), respectively. The significant predictors of PCa on TRUS were hypoechogenicity (odds ratio [OR]: 9.595, P = 0.002), taller-than-wide shape (OR: 3.539, P = 0.022), asymmetric vascular structures (OR: 3.728, P = 0.031), close proximity to capsule (OR: 3.473, P = 0.040), and irregular margins (OR: 3.843, P = 0.041). We propose subgrouping PI-RADS score 3 into categories 3a, 3b, 3c, and 3d based on different numbers of TRUS predictors, as the creation of PI-RADS 3a (no suspicious ultrasound features) could avoid 16.7% of mpMRI-guided TBs. Risk stratification of PCa with mpMRI-TRUS fusion imaging-directed ultrasound features could avoid unnecessary mpMRI-TBs.

Added value of shear-wave elastography in the prediction of extracapsular extension and seminal vesicle invasion before radical prostatectomy.

The purpose of this study was to analyze the value of transrectal shear-wave elastography (SWE) in combination with multivariable tools for predicting adverse pathological features before radical prostatectomy (RP). Preoperative clinicopathological variables, multiparametric magnetic resonance imaging (mp-MRI) manifestations, and the maximum elastic value of the prostate (Emax) on SWE were retrospectively collected. The accuracy of SWE for predicting adverse pathological features was evaluated based on postoperative pathology, and parameters with statistical significance were selected. The diagnostic performance of various models, including preoperative clinicopathological variables (model 1), preoperative clinicopathological variables + mp-MRI (model 2), and preoperative clinicopathological variables + mp-MRI + SWE (model 3), was evaluated with area under the receiver operator characteristic curve (AUC) analysis. Emax was significantly higher in prostate cancer with extracapsular extension (ECE) or seminal vesicle invasion (SVI) with both P < 0.001. The optimal cutoff Emax values for ECE and SVI were 60.45 kPa and 81.55 kPa, respectively. Inclusion of mp-MRI and SWE improved discrimination by clinical models for ECE (model 2 vs model 1, P = 0.031; model 3 vs model 1, P = 0.002; model 3 vs model 2, P = 0.018) and SVI (model 2 vs model 1, P = 0.147; model 3 vs model 1, P = 0.037; model 3 vs model 2, P = 0.134). SWE is valuable for identifying patients at high risk of adverse pathology.

Synchronous tele-ultrasonography is helpful for a naive operator to perform high-quality thyroid ultrasound examinations.

PURPOSE: This study investigated the value of synchronous tele-ultrasonography (TUS) for naive operators in thyroid ultrasonography (US) examinations. METHODS: Ninety-seven patients were included in this prospective, parallel-controlled trial. Thyroid scanning and diagnosis were completed by resident A independently, resident B with guidance from a US expert through synchronous TUS, and an on-site US expert. The on-site expert's findings constituted the reference standard. Two other off-site US experts analyzed all data in a blind manner. Inter-operator consistency between the two residents and the on-site US expert for thyroid size measurements, nodule measurements, nodule features, American College of Radiology (ACR) Thyroid Imaging Reporting and Data System (TI-RADS) categories, and image quality was compared. Two questionnaires were completed to evaluate the clinical benefit. RESULTS: Resident B detected more nodules consistent with the on-site expert than resident A did (89.4% vs. 56.5%, P<0.001). Resident B achieved excellent consistency with the on-site expert in terms of ACR TI-RADS categories, nodule composition, shape, echogenic foci, and vascularity (all intra-class correlation coefficients [ICCs] >0.75), while resident A achieved lower consistency in ACR TI-RADS categories, composition, echogenicity, margin, echogenic foci, and vascularity (all ICCs 0.40-0.75). Residents A and B had excellent consistency in target nodule measurements (all ICCs >0.75). Resident B achieved better performance than resident A for gray values, time gain compensation, depth, color Doppler adjustment, and the visibility of key information (all P<0.05). Furthermore, 61.9% (60/97) of patients accepted synchronous TUS, and 59.8% (58/97) patients were willing to pay for it. CONCLUSION: Synchronous TUS can help inexperienced residents achieve comparable thyroid diagnostic capability to a US expert.

Improving the quality of breast ultrasound examination performed by inexperienced ultrasound doctors with synchronous tele-ultrasound: a prospective, parallel controlled trial.

PURPOSE: This prospective study explored the value of synchronous tele-ultrasound (US) to aid doctors inexperienced in US with breast US examinations. METHODS: In total, 99 patients were enrolled. Two trainee doctors who were inexperienced in US (trainee A [TA] and trainee B [TB]) and one doctor who was an expert in US completed the US examinations sequentially. TA completed the US examinations independently, while TB was instructed by the expert using synchronous tele-US. Subsequently, the expert performed on-site US examinations in person. Separately, they selected the most clinically significant nodule as the target nodule. Consistency with the expert and image quality were compared between TA and TB to evaluate tele-US. Furthermore, TB and the patients evaluated tele-US through questionnaires. RESULTS: TB demonstrated higher consistency with the expert in terms of target nodule selection than TA (93.3% vs. 63.3%, P<0.001). TB achieved good inter-observer agreement (ICC, >0.75) with the expert on five US features (5/9, 55.6%), while TA only did so for one (1/9, 11.1%) (P=0.046). TB's image quality was higher than TA's in gray value, time gain compensation, depth, color Doppler adjustment, and the visibility of key information (P=0.018, P<0.001, P<0.001, P=0.033, and P=0.006, respectively). The comprehensive assessment score was higher for TB than for TA (3.96±0.82 vs. 3.09±0.87, P<0.001). Tele-US was helpful in 69.7% of US examinations and had a training effect in 68.0%. Furthermore, 63.6% of patients accepted tele-US and 60.6% were willing to pay. CONCLUSION: Tele-US can help doctors inexperienced in US to perform breast US examinations.