Background: Accurate diagnosis of unexplained cervical lymphadenopathy (CLA) using medical images heavily relies on the experience of radiologists, which is even worse for CLA patients in underdeveloped countries and regions, because of lack of expertise and reliable medical history. This study aimed to develop a deep learning (DL) radiomics model based on B-mode and color Doppler ultrasound images for assisting radiologists to improve their diagnoses of the etiology of unexplained CLA.Methods: Patients with unexplained CLA who received ultrasound examinations from three hospitals located in underdeveloped areas of China were retrospectively enrolled. They were all pathologically confirmed with reactive hyperplasia, tuberculous lymphadenitis, lymphoma, or metastatic carcinoma. By mimicking the diagnosis logic of radiologists, three DL sub-models were developed to achieve the primary diagnosis of benign and malignant, the secondary diagnosis of reactive hyperplasia and tuberculous lymphadenitis in benign candidates, and of lymphoma and metastatic carcinoma in malignant candidates, respectively. Then, a CLA hierarchical diagnostic model (CLA-HDM) integrating all sub-models was proposed to classify the specific etiology of each unexplained CLA. The assistant effectiveness of CLA-HDM was assessed by comparing six radiologists between without and with using the DL-based classification and heatmap guidance.Results: A total of 763 patients with unexplained CLA were enrolled and were split into the training cohort (n=395), internal testing cohort (n=171), and external testing cohorts 1 (n=105) and 2 (n=92). The CLA-HDM for diagnosing four common etiologies of unexplained CLA achieved AUCs of 0.873 (95% CI: 0.838–0.908), 0.837 (95% CI: 0.789–0.889), and 0.840 (95% CI: 0.789–0.898) in the three testing cohorts, respectively, which was systematically more accurate than all the participating radiologists. With its assistance, the accuracy, sensitivity, and specificity of six radiologists with different levels of experience were generally improved, reducing the false-negative rate of 2.2–10% and the false-positive rate of 0.7–3.1%.Conclusions: Multi-cohort testing demonstrated our DL model integrating dual-modality ultrasound images achieved accurate diagnosis of unexplained CLA. With its assistance, the gap between radiologists with different levels of experience was narrowed, which is potentially of great significance for benefiting CLA patients in underdeveloped countries and regions worldwide.
Ultrasound (US) imaging is part of conventional medical imaging in clinical practice that is low-cost, non-ionizing, portable and capable of real-time image acquisition and display. However, in certain cases, US has limited sensitivity and specificity in differentiating between malignant and benign lesions. Ultrasound-based radiomics, as a new branch of radiomics, can provide additional features such as heterogeneity of lesions that are invisible to the naked eye, alone or in combination with demographic, histological, genomic or proteomic data, thereby improving the accuracy of US in diagnosis of disease. This article provides an introduction to ultrasound-based radiomics, covering its workflow, the application of machine learning, and current research status. Current limitations of radiomics, such as consistency of image acquisition, parameter variations, and difficulty in calibrating quantitative methods in ultrasound, will also be covered.
The immunosuppressive tumor microenvironment (TME) significantly inhibits the effective anti-tumor immune response, greatly affecting the efficacy of immunotherapy. Most tumor-associated macrophages (TAMs) belong to the M2 phenotype, which contributes significantly to the immunosuppressive effects in non-small cell lung cancer (NSCLC) TME. The interaction between signal regulatory protein α (SIRPα) expressed on macrophages and CD47, a transmembrane protein overexpressed on cancer cells, activates the "eat-me-not" signaling pathway, inhibiting phagocytosis. In this study, a folic acid (FA)-modified ultrasound responsive gene/drugs delivery system, named FA@ PFP @ Fe3O4 @LNB-SIRPα siRNA (FA-PFNB-SIRPα siRNA), was developed using 1,2-dioleoacyl-3-trimethylammonium-propane (DOTAP), FA-1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N- [amino (polyethylene glycol)2000] (DSPE-PEG2000-FA), cholesterol, and perfluoropentane (PFP), for the delivery of siRNA encoding SIRPα mRNA and immune adjuvant Fe3O4 nanoparticles. Under ultrasound conditions, the nanobubbles effectively transfected macrophages, inhibiting SIRPα mRNA and protein expression, promoting the phagocytosis of TAMs, and synergistically reversing M2 polarization. This system promotes the infiltration of T cells, enhances the proliferation and activation of cytotoxic T cells, and inhibits the infiltration of immunosuppressive cells in tumor tissues. Administration of FA-PFNB-SIRPα siRNA combined with ultrasound significantly inhibits NSCLC progression. The study highlights the potential of ultrasound nanotechnology-enabled delivery of SIRPα siRNA and Fe3O4 as an effective strategy for macrophage-based immunotherapy to reshape the immunosuppressive TME for cancer therapy.
Purpose: Early and accurate preoperative diagnosis of central lymph node metastasis (CLNM) is crucial to improve surgical management of patients with clinical lymph node-negative papillary thyroid carcinoma (PTC). Towards improving diagnosis of CLNM, we assessed the value of combining preoperative clinical characteristics, conventional ultrasound, and contrast-enhanced ultrasound (CEUS) in preoperative prediction of CLNM of different sized PTCs. Patients and Methods: Patients were divided according to tumor size: a PTC group (> 10 mm) and a papillary thyroid microcarcinoma (PTMC) group (≤ 10 mm). We retrospectively analyzed the clinical and ultrasonographic features of 120 PTC patients and 165 PTMC patients. Multivariate logistic regression analysis was used to screen independent risk factors and establish prediction models. Receiver operating characteristic curves were used to determine the best cut-off values for continuous variables and assess the performance of prediction models. Results: Independent risk predictors of CLNM for the PTC group were extrathyroidal extension in CEUS (OR=7.923), tumor size > 14 mm (OR=5.491), and multifocality (OR=3.235). For the PTMC group, the independent risk factors were the distance from the thyroid capsule =0 mm (OR=4.629), male (OR=3.315), tumor size > 5 mm (OR=3.304), and microcalcification (OR=2.560). The predictive model of combined method had better performance in predicting CLNM of PTC compared with models based on CEUS and conventional ultrasound alone (area under the curve: 0.832 vs 0.739, P=0.0011; 0.832 vs 0.678, P=0.0012). For PTMC, comparing with CEUS, the combined method and conventional ultrasound performed better than CEUS alone in predicting CLNM (area under the curve: 0.783 vs 0.636, P=0.0016; 0.738 vs 0.636, P=0.0196). Conclusion: The predictive models of combined method obtained from significant preoperative clinical and ultrasonographic features can potentially improve the preoperative diagnosis and individual treatment of CLNM in patients with PTC and PTMC. CEUS may be helpful in predicting CLNM of PTC, but CEUS would be ineffective in predicting CLNM of PTMC.