A5027756101- Medical Imaging Techniques and Applications
- Radiation Dose and Imaging
- Advanced X-ray and CT Imaging
- Head and Neck Surgical Oncology
- Sinusitis and nasal conditions
- Retinal Imaging and Analysis
- Machine Learning and Data Classification
- Imbalanced Data Classification Techniques
- Anomaly Detection Techniques and Applications
- Acute Ischemic Stroke Management
- Dental Radiography and Imaging
- Medical Image Segmentation Techniques
Seoul National University
2021
Seoul National University Bundang Hospital
2021
Accurate image interpretation of Waters’ and Caldwell view radiographs used for sinusitis screening is challenging. Therefore, we developed a deep learning algorithm diagnosing frontal, ethmoid, maxillary on both views. The datasets were selected the training validation set (n = 1403, sinusitis% 34.3%) test 132, 29.5%) by temporal separation. can simultaneously detect classify each paranasal sinus using views without manual cropping. Single- multi-view models compared. Our proposed...
Blood vessel segmentation (BVS) of 3D medical imaging such as computed tomography and magnetic resonance angiography (MRA) is an essential task in the clinical field. Automation BVS using deep supervised learning being researched, U-Net-based approaches, which are considered standard for image segmentation, proposed a lot. However, inherent characteristics blood vessels, e.g., they complex narrow, well resolution sensitivity modalities increases difficulty BVS. We propose novel model named...
The performance of deep learning algorithm (DLA) to that radiologists was compared in detecting low contrast objects CT phantom images under various imaging conditions. For training, 10,000 were created using American College Radiology as the background. In half images, 3–20 mm size and 5–30 HU difference generated random locations. Binary responses used ground truth. testing, 640 Catphan® used, which had either 5 or 9 with 10 difference. Twelve evaluated presence on a five-point scale....
The average accuracy is one of major evaluation metrics for classification systems, while the deviation another important performance metric used to evaluate various deep neural networks. In this paper, we present a new ensemble-like fast network, Harmony, that can reduce among categories without degrading overall accuracy. Harmony consists three sub-models, namely, Target model, Complementary and Conductor model. an object classified by using either model or conventional network general...