ABSTRACT
Objective.Automated detection and segmentation of breast masses in ultrasound images are critical for breast cancer diagnosis, but remain challenging due to limited image quality and complex breast tissues. This study aims to develop a deep learning-based method that enables accurate breast mass detection and segmentation in ultrasound images.Approach.A novel convolutional neural network-based framework that combines the You Only Look Once (YOLO) v5 network and the Global-Local (GOLO) strategy was developed. First, YOLOv5 was applied to locate the mass regions of interest (ROIs). Second, a Global Local-Connected Multi-Scale Selection (GOLO-CMSS) network was developed to segment the masses. The GOLO-CMSS operated on both the entire images globally and mass ROIs locally, and then integrated the two branches for a final segmentation output. Particularly, in global branch, CMSS applied Multi-Scale Selection (MSS) modules to automatically adjust the receptive fields, and Multi-Input (MLI) modules to enable fusion of shallow and deep features at different resolutions. The USTC dataset containing 28 477 breast ultrasound images was collected for training and test. The proposed method was also tested on three public datasets, UDIAT, BUSI and TUH. The segmentation performance of GOLO-CMSS was compared with other networks and three experienced radiologists.Main results.YOLOv5 outperformed other detection models with average precisions of 99.41%, 95.15%, 93.69% and 96.42% on the USTC, UDIAT, BUSI and TUH datasets, respectively. The proposed GOLO-CMSS showed superior segmentation performance over other state-of-the-art networks, with Dice similarity coefficients (DSCs) of 93.19%, 88.56%, 87.58% and 90.37% on the USTC, UDIAT, BUSI and TUH datasets, respectively. The mean DSC between GOLO-CMSS and each radiologist was significantly better than that between radiologists (p< 0.001).Significance.Our proposed method can accurately detect and segment breast masses with a decent performance comparable to radiologists, highlighting its great potential for clinical implementation in breast ultrasound examination.
Subject(s)
Breast Neoplasms , Deep Learning , Image Processing, Computer-Assisted , Humans , Breast Neoplasms/diagnostic imaging , Image Processing, Computer-Assisted/methods , Ultrasonography/methods , Female , Ultrasonography, Mammary/methods , Neural Networks, ComputerABSTRACT
Aim: Annual T1 stage papillary thyroid carcinoma (PTC) incidence rates continue to rise, yet the optimal treatment for this cancer type remains controversial. Central lymph node metastasis (CLNM) is a critical determinant in the context of treatment decision-making. While several prior studies have evaluated patients with clinica l T1a(cT1a) stage PTC, there have been fewer analyses of clinical T1b(cT1b) disease to date. The present study was thus formulated to explore predictors of CLNM in patients with cT1a and cT1b stage PTC. Methods: A retrospective analysis of data including clinicopathological characteristics and BRAFV600E mutation status was conducted for 452 PTC patients undergoing surgical treatment. Logistic univariate and multivariate analyses were performed to identify risk factors associated with CLNM in particular patients' characteristics and the accuracy of the established logistic regression models was evaluated using the R software platform. Results: Respective CLNM incidence rates in cT1a and cT1b disease were 39.39% and 67.21%. Factors associated with a higher risk of CLNM among PTC(cT1a) patients included male sex, young age, tumor size, contact with capsule, and multifocality as determined through comparisons of the area under the curve for logistic regression models. Whereas male sex and age were associated with CLNM risk in PTC(cT1b) patients in univariate and multivariate analyses, age was the only risk factor associated with CLNM incidence among women with PTC(cT1b). Conclusion: Predictors of CLNM differ between PTC patients with cT1a and cT1b stage disease, and a comprehensive assessment of these risk factors should thus be conducted when designing individualized treatment regimens for PTC patients.
Subject(s)
Carcinoma, Papillary , Thyroid Neoplasms , Carcinoma, Papillary/pathology , Female , Humans , Lymphatic Metastasis , Male , Retrospective Studies , Risk Factors , Thyroid Cancer, Papillary/pathology , Thyroid Cancer, Papillary/surgery , Thyroid Neoplasms/epidemiology , Thyroid Neoplasms/genetics , Thyroid Neoplasms/surgeryABSTRACT
BACKGROUND: Percutaneous radiofrequency ablation (RFA) is an effective treatment for unresectable hepatocellular carcinoma (HCC) and a minimally invasive alternative to hepatectomy for treating tumour recurrence. RFA is often performed using contrast-enhanced computed tomography (CECT) and/or ultrasonography. In recent years, angiographic systems with flat panel image detectors and advanced image reconstruction algorithms have broadened the clinical applications of cone-beam computed tomography (CBCT), including RFA. Several studies have shown the effectiveness of using CBCT for immediate treatment assessments and follow-ups. AIM: To assess the treatment response to RFA for HCC using CBCT. METHODS: Forty-eight patients (44 men; aged 37-89 years) with solitary HCC [median size: 3.2 (1.2-6.6) cm] underwent RFA and were followed for 25.6 (median; 13.5-35.2) mo. Image fusion of CBCT and pre-operative CECT or magnetic resonance imaging (MRI) was used for tumour segmentation and needle path and ablation zone planning. Real-time image guidance was provided by overlaying the three-dimensional image of the tumour and needle path on the fluoroscopy image. Treatment response was categorized as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD). Disease progression, death, time to progression (TTP), and overall survival (OS) were recorded. Kaplan-Meier and Cox regression analyses were performed. RESULTS: Initial post-RFA CECT/MRI showed 38 cases of CR (79.2%), 10 of PR (20.8%), 0 of SD, and 0 of PD, which strongly correlated with the planning estimation (42 CR, 87.5%; 6 PR, 12.5%; 0 SD; and 0 PD; accuracy: 91.7%, P < 0.01). Ten (20.8%) patients died, and disease progression occurred in 31 (35.4%, median TTP: 12.8 mo) patients, resulting in 12-, 24-, and 35-mo OS rates of 100%, 81.2%, and 72.2%, respectively, and progression-free survival (PFS) rates of 54.2%, 37.1%, and 37.1%, respectively. The median dose-area product of the procedures was 79.05 Gy*cm2 (range 40.95-146.24 Gy*cm2), and the median effective dose was 10.27 mSv (range 5.32-19.01 mSv). Tumour size < 2 cm (P = 0.008) was a significant factor for OS, while age (P = 0.001), tumour size < 2 cm (P < 0.001), tumour stage (P = 0.010), and initial treatment response (P = 0.003) were significant factors for PFS. CONCLUSION: Reliable RFA treatment planning and satisfactory outcomes can be achieved with CBCT.
ABSTRACT
RATIONALE AND OBJECTIVES: The objective of this study was to investigate the impact of a dual-phase cone-beam computed tomography (DP-CBCT)-based navigation imaging during transarterial chemoembolization (TACE) of hepatocellular carcinoma (HCC) in a perspective randomized study. MATERIALS AND METHODS: Forty-two patients with HCC (39 men, 57 ± 9 years, 13 first-time TACE) underwent TACE using three-dimensional image guidance with automatic detection of tumor-feeding vessels computed from DP-CBCT (early and delayed arterial phases). Forty-nine other patients with HCC (44 men, 55 ± 12 years, 14 first-time TACE) were treated conventionally using digital subtraction angiography (DSA). Tumor detectability in DP-CBCT was compared to DSA and preoperative CT or magnetic resonance (MR) imaging. Tumor-feeding vessel visibility was rated (good, fair, and poor) intraoperatively by the operators. The superselective embolization success rate, the number of DSA acquisitions, fluoroscopy time, and patient radiation dose were collected and compared using paired t test and the Mann-Whitney U test. RESULTS: Tumor detection of DP-CBCT was superior to DSA (100% vs 83%, P = .001) and comparable to CT-MR (96%, P = .456). Tumor and feeder visibilities were significantly enhanced by DP-CBCT (P < .001). Compared to using DSA, more superselective embolization was achieved (60% vs 49%) with less DSA acquisitions (n = 2.6 ± 0.8 vs n = 3.4 ± 0.7, P < .001) and shorter fluoroscopy time (4.1 ± 2.6 vs 7.1 ± 4.2 minutes, P < .001) with a slight increase in patient radiation exposure, that is, air kerma (median: 0.33, first to third quartiles: 0.24-0.48 vs 0.30, 0.24-0.44 Gy; P = .519) and dose-area product (134, 92-181 vs 97, 75-140 Gyâ cm2, P = .048). CONCLUSIONS: DP-CBCT and navigation imaging improve tumor detectability and superselective embolization in TACE.
Subject(s)
Carcinoma, Hepatocellular/diagnostic imaging , Carcinoma, Hepatocellular/therapy , Chemoembolization, Therapeutic/methods , Cone-Beam Computed Tomography/methods , Liver Neoplasms/diagnostic imaging , Liver Neoplasms/therapy , Adult , Aged , Angiography, Digital Subtraction , Blood Vessels/diagnostic imaging , Carcinoma, Hepatocellular/blood supply , Female , Fluoroscopy , Humans , Imaging, Three-Dimensional/methods , Liver Neoplasms/blood supply , Magnetic Resonance Imaging , Male , Middle Aged , Prospective Studies , Radiation Exposure , Radiology, Interventional/methods , Time FactorsABSTRACT
Operator radiation and the radiation protection efficacy of a ceiling-suspended lead screen were assessed during coronary angiography (CA) in a catheterization laboratory. An anthropomorphic phantom was placed under the X-ray beam to simulate patient attenuation in eight CA projections. Using real-time dosimeters, radiation dose rates were measured on models mimicking a primary operator (PO) and an assistant. Subsequently, a ceiling-suspended lead screen was placed in three commonly used positions to compare the radiation protection efficacy. The radiation exposure to the PO was 2.3 to 227.9 (mean: 67.2 ± 49.0) µSv/min, with the left anterior oblique (LAO) 45°/cranial 25° and cranial 25° projections causing the highest and the lowest dose rates, respectively. The assistant experienced significantly less radiation overall (mean: 20.1 ± 19.6 µSv/min, P < 0.003), with the right anterior oblique (RAO) 30° and cranial 25° projections resulting in the highest and lowest exposure levels, respectively. Combined with table-side shielding, the ceiling-suspended lead screen reduced the radiation to the PO by 76.8%, 81.9% and 93.5% when placed close to the patient phantom, at the left side and close to the PO, respectively, and reduced the radiation to the assistant by 70.3%, 76.7% and 90.0%, respectively. When placed close to the PO, a ceiling-suspended lead screen provides substantial radiation protection during CA.
ABSTRACT
RATIONALE AND OBJECTIVES: To assess patient radiation dose reduction and the image quality of a new X-ray imaging technology during repetitive transarterial chemoembolization (TACE) for hepatocellular carcinoma (HCC). METHODS: Fifty HCC patients (36 men; 57 ± 11 years) undergoing repetitive TACE were first randomly assigned to receive a TACE treatment on a reference X-ray system or a low-dose system with advanced real-time image processing. The alternate system was used for a repeated TACE (treatment interval, 0.5-6 months). Fluoroscopy time, number of digital subtraction angiography (DSA), air kerma (AK), and dose area product (DAP) were compared between the two systems and between the two repetitive TACE. Three interventional radiologists independently rated the image quality in blinded offline readings. RESULTS: Fluoroscopy time (8.7 ± 5.9 minutes vs. 8.7 ± 7.9 minutes, P = .981), numbers of DSA runs (6 ± 4 vs. 6 ± 4, P = .735), and exposure images (173 ± 86 vs. 168 ± 91, P = .916) were equivalent between the two systems. No statistical difference in X-ray usage was found between repeated treatments. Compared to the reference system, the technology significantly reduced AK and DAP by 48.6% (0.17 ± 0.13 Gy vs. 0.41 ± 0.36 Gy, P < .0001) and 50.3% (77.3 ± 55.2 Gy cm(2) vs. 195.0 ± 155.5 Gy cm(2), P < .0001), respectively. Image quality was rated comparable between the new system and the reference, with average scores of 3.9 ± 0.3 versus 4.4 ± 0.3 in fluoroscopy and 4.5 ± 0.2 versus 4.3 ± 0.3 in DSA. CONCLUSIONS: Patient radiation exposure can be substantially reduced by a factor of approximately two with the novel X-ray imaging technology while maintaining image quality.
Subject(s)
Angiography, Digital Subtraction/methods , Carcinoma, Hepatocellular/diagnostic imaging , Carcinoma, Hepatocellular/therapy , Embolization, Therapeutic , Fluoroscopy/methods , Liver Neoplasms/diagnostic imaging , Liver Neoplasms/therapy , Radiation Dosage , Radiographic Image Enhancement , Female , Humans , Male , Middle Aged , Prospective StudiesABSTRACT
OBJECTIVE: The purpose of this study was threefold: to estimate the organ doses and effective doses (EDs) for seven neurovascular imaging protocols, to study the effect of beam collimation on ED, and to derive protocol-specific dose-area product (DAP)-to-ED conversion factors. MATERIALS AND METHODS: A cone-beam CT system was used to measure the organ doses for seven neurovascular imaging protocols. Two datasets were obtained: seven protocols without beam collimation (FOV, entire head) and four with beam collimation (FOV, from the base to the top of the skull). Measurements were performed on an adult male anthropomorphic phantom with 20 metal oxide semiconductor field-effect transistor (MOSFET) detectors placed in selected organs. The DAP values were recorded from the console. The EDs of five protocols were also estimated using Monte Carlo simulations software. The ED values were computed by multiplying measured organ doses to corresponding International Commission on Radiological Protection tissue-weighting factors. RESULTS: Without collimation, the EDs ranged from 0.16 to 1.6 mSv, and the DAP-to-ED conversion factors ranged from 0.035 to 0.076 mSv/Gy·cm(2). For the four protocols investigated with beam collimation, the ED was reduced by a factor of approximately 2, and the DAP-to-ED conversion factors were reduced by approximately 30%. For the five protocols also estimated with the Monte Carlo method, the estimated EDs were in agreement (< 20% deviation) with those determined by the MOSFET method. CONCLUSION: We have estimated ED for standard adult neuroimaging protocols in a 3D rotational angiography system. Our results provide a simple means of ED estimation using DAP console readings.
Subject(s)
Angiography/methods , Clinical Protocols , Cone-Beam Computed Tomography/standards , Neuroimaging/methods , Phantoms, Imaging , Radiation Dosage , Adult , Humans , Male , Monte Carlo MethodABSTRACT
To assess patient radiation during catheter ablation procedures and operator differences. From 84 patients (51 males, age 63 ± 10 years) undergoing complex catheter ablation by three experienced operators we collected: body mass index (BMI), procedure type and time, fluoroscopy time, dose area product (DAP), air kerma and X-ray system setting (cine, collimation and angiographic imaging angle). A new factor, fluoroscopy DAP-fluoroscopy time ratio, was introduced to compare operator differences. The results show the average procedure time was 179 (± 57) minutes (min), fluoroscopy time was 31 (± 21) min, DAP was 26.4 (± 19.6) Gy.cm(2) and air kerma was 0.26 (± 0.19) Gy. Procedure types were: pulmonary vein isolation (PVI) (52 %), redo PVI (11 %), pulmonary vein ablation catheter (PVAC) (14 %), ventricular tachycardia (VT) (8 %) and others (15 %). Inter-operator difference was observed in fluoroscopy and cine usage. Fluoroscopy DAP-time ratios showed a similar level of patient radiation dose rate by operator A and B (correlation: 0.89), and a significantly higher dose rate by operator C (correlation: 0.20, p<0.001; 0.26, p<0.01, to operator A and B). In conclusion, operators should be aware of patient radiation exposure levels and the influencing factors. Inter- and intra-operator differences can be measured and bench marked for improvement in X-ray efficiency and patient radiation reduction.