• Journal of radiological science and technology
• /
• v.47 no.1
• /
• pp.29-37
• /
• 2024

This study aims to evaluate the performance of the U-Net based learning model that may vary depending on the histogram equalization algorithm. The subject of the experiment were 17 radiology students of this college, and 1,727 data sets in which the region of interest was set in the thyroid after acquiring ultrasound image data were used. The training set consisted of 1,383 images, the validation set consisted of 172 and the test data set consisted of 172. The equalization algorithm was divided into Histogram Equalization(HE) and Contrast Limited Adaptive Histogram Equalization(CLAHE), and according to the clip limit, it was divided into CLAHE8-1, CLAHE8-2. CLAHE8-3. Deep Learning was learned through size control, histogram equalization, Z-score normalization, and data augmentation. As a result of the experiment, the Attention U-Net showed the highest performance from CLAHE8-2 to 0.8355, and the U-Net and BSU-Net showed the highest performance from CLAHE8-3 to 0.8303 and 0.8277. In the case of mIoU, the Attention U-Net was 0.7175 in CLAHE8-2, the U-Net was 0.7098 and the BSU-Net was 0.7060 in CLAHE8-3. This study attempted to confirm the effects of U-Net, Attention U-Net, and BSU-Net models when histogram equalization is performed on ultrasound images. The increase in Clip Limit can be expected to increase the ROI match with the prediction mask by clarifying the boundaries, which affects the improvement of the contrast of the thyroid area in deep learning model learning, and consequently affects the performance improvement.

• Proceedings of the KOSOMBE Conference
• /
• v.1994 no.12
• /
• pp.39-42
• /
• 1994

An automatic multi-threshold algorithm for segmentation of 2D ultrasound images based on average filtering and the characteristics of speckle noise in 2D ultrasound image is proposed. To do this, we investigate the histogram of difference between $7{\times}7$ averaging histogram and $3{\times}3$ averaging histogram. And, we find zero crossing points in the positive portion of the differenced histogram and select middle points of the zero crossing points. We assign these selected points to characteristic points. The thresholds are the center of two characteristic points. Then we segment 2D ultrasound image by using these thresholds and extract edges from applying edge operator to optimal segmented image. Experimental results show that the segmented regions are devided accurately around the homogeneous region.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.8
• /
• pp.27-31
• /
• 2013

In this paper, we propose a new quantization method using fuzzy C-means clustering(FCM) for lumbar ultrasound image recognition. Unlike usual histogram based quantization, our method first classifies regions into 10 clusters and sorts them by the central value of each cluster. Those clusters are represented with different colors. This method is efficient to handle lumbar ultrasound image since in this part of human body, the brightness values are distributed to doubly skewed histogram in general thus the usual histogram based quantization is not strong to extract different areas. Experiment conducted with 15 real lumbar images verified the efficacy of proposed method.

• International journal of advanced smart convergence
• /
• v.11 no.2
• /
• pp.185-193
• /
• 2022

There have been various studies in Korea to develop a %IMF(Intramuscular Fat Percentage) estimation method suitable for Hanwoo. Recently, a %IMF estimation method using a convolutional neural network (CNN), a kind of deep learning method among artificial intelligence methods, has been studied. In this study, we performed a performance comparison when various preprocessing methods were applied to the %IMF estimation of ultrasound images using CNN as mentioned above. The preprocessing methods used in this study are normalization, histogram equalization, edge enhancement, and a method combining normalization and edge enhancement. When estimating the %IMF of Hanwoo by the conventional method that did not apply preprocessing in the experiment, the accuracy was 98.2%. The other hand, we found that the accuracy improved to 99.5% when using preprocessing with histogram equalization alone or combined regularization and edge enhancement.

• Journal of Veterinary Clinics
• /
• v.24 no.4
• /
• pp.568-572
• /
• 2007

This study was carried out to show at the fractal analysis complements the practical disadvantage of gray level histogram which is designed to measure the quantitative classification of echo patterns in ultrasonographic image of parenchymal organs such as spleen and kidney and it is a practical method of measurement for quantitative classification. By using ultrasonographs, kidney and spleen of 21 healthy Beagles were fixed under different gain settings to be scanned for echo patterns and results were analyzed with body gray level histogram and fractal analysis. Then it was compared based on the statistical data obtained. Although there was a proportionate increase in histogram along with gain settings, there were consistencies in the fractal dimension. In terms of quantitative analysis in ultrasonographic images, fractal analysis is concluded to complement the practical disadvantage of gray level histogram.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2018.05a
• /
• pp.79-81
• /
• 2018

혈관내 초음파(Intravascular Ultrasound, IVUS)는 혈관 내벽의 단면을 보여주는 검사 방법으로 관상 동맥 내의 내강, 죽상 경화반, 그리고 혈관벽의 변화에 관한 직접적이고 구체적인 정보를 제공한다. 본 논문에서는 IVUS 영상에서 내막과 외막을 추출하고 각 막의 지름을 자동적으로 추출하는 방법을 제안한다. 제안된 방법은 IVUS 영상에 Histogram Equalization 기법을 적용하여 명암 대비를 강조한 후에 퍼지 이진화 기법과 평균 이진화 기법을 각각 적용하여 내막과 외막을 추출하기 위해 이진화한다. 이진화된 내막과 외막의 각 영역 중에서 혈관내 초음파 영상 중심에서 가장 큰 영역의 정보를 이용하여 라벨링 기법을 적용하여 내막과 외막 영역을 추출하고 각 막의 지름을 계산한다. 제안된 방법을 IVUS 영상을 대상으로 실험한 결과, 내막과 외막의 지름이 비교적 정확히 추출되는 것을 실험을 통하여 확인하였다.

• Journal of the Korean Society of Radiology
• /
• v.10 no.4
• /
• pp.247-253
• /
• 2016

Ultrasound devices diagnose many disease, which is widely used, can not be standardized quantitative evaluated in order to evaluate sonography image of quality. Therefore, in this papers, aims to get correct image in order to accurate diagnosis by figuring out the appropriate parameter based on each target by measuring distortion which results in the analyzation of the sensitivity of SNR and the histogram of signal by manipulating parameter of 8 mm target in ATS-539 multipurpose phantom. Equipment using Acuson sequoia 512, convex probe and utilizes multi-objective phantom. experiment method is that first you put the phantom on the flat and acquire 85 sheets of image, changing frequency(2,3,4 MHz, harmonic 3, 4, 4.5 MHz), Focus(2, 4, 6 unit), and Dynamic Range(58, 68, 78, 88, 98) for a 8 mm structure. through the Image J program. The sensitivity angle of 8mm target through Image J program is gauged by each separate target SNR and the distorted angle subtract and measure Histogram of background from Histogram of signal and take top 40% from the given result value above. According to parameter variation we found out proper parameter by acquiring SNR of sensitivity and distortion data for aspect of transition. The more this findings have Focus, the lower distortion value and at 4 MHz frequency this result have high SNR and low distortion value. Dynamic Range got an appropriate image on 88 and 98. It is considered on the basis of the experimental data, the probability of disease diagnosis will get higher.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.6
• /
• pp.11-17
• /
• 2014

While safe and convenient, ultrasound imaging analysis is often criticized by its subjective decision making nature by field experts in analyzing musculoskeletal system. In this paper, we propose a new automatic method to extract muscle area using ART2 neural network based quantization. A series of image processing algorithms such as histogram smoothing and End-in search stretching are applied in pre-processing phase to remove noises effectively. Muscle areas are extracted by considering various morphological features and corresponding analysis. In experiment, our ART2 based Quantization is verified as more effective than other general quantization methods.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.47 no.5
• /
• pp.24-32
• /
• 2010

In this study, we have proposed a pre-processing algorithm newly developed for improving the tracking efficiency of the optical flow method. The developed pre-processing algorithm consists of a median filter, binarization, morphology, canny edge, contour detecting and an approximation method. In order to evaluate whether the optical flow tracking capacity increases, this study applied the pre-processing algorithm to the Lucas-Kanade(LK) optical flow algorithm, and comparatively analyzed its images and tracking results with those of optical flow without the pre-processing algorithm and with the existing pre-processing algorithm(composed of median filter and histogram equalization). As a result, it was observed that the tracking performance derived from the LK optical flow algorithm with the pre-processing algorithm, shows better tracking accuracy, compared to the one without the pre-processing algorithm and the one with the existing pre-processing algorithm. It seems to have resulted by successful segmentation for characteristic areas and subdivision into inner and outer contour lines.

• International journal of advanced smart convergence
• /
• v.8 no.1
• /
• pp.24-34
• /
• 2019

In this research, a practical deep learning framework to differentiate the lesions and nodules in breast acquired with ultrasound imaging has been proposed. 7408 ultrasound breast images of 5151 patient cases were collected. All cases were biopsy proven and lesions were semi-automatically segmented. To compensate for the shift caused in the segmentation, the boundaries of each lesion were drawn using Fully Convolutional Networks(FCN) segmentation method based on the radiologist's specified point. The data set consists of 4254 benign and 3154 malignant lesions. In 7408 ultrasound breast images, the number of training images is 6579, and the number of test images is 829. The margin between the boundary of each lesion and the boundary of the image itself varied for training image augmentation. The training images were augmented by varying the margin between the boundary of each lesion and the boundary of the image itself. The images were processed through histogram equalization, image cropping, and margin augmentation. The networks trained on the data with augmentation and the data without augmentation all had AUC over 0.95. The network exhibited about 90% accuracy, 0.86 sensitivity and 0.95 specificity. Although the proposed framework still requires to point to the location of the target ROI with the help of radiologists, the result of the suggested framework showed promising results. It supports human radiologist to give successful performance and helps to create a fluent diagnostic workflow that meets the fundamental purpose of CADx.