• Korean Journal of Radiology
• /
• v.23 no.10
• /
• pp.949-958
• /
• 2022

Objective: To investigate the feasibility of using a deep learning-based analysis of auscultation data to predict significant stenosis of arteriovenous fistulas (AVF) in patients undergoing hemodialysis requiring percutaneous transluminal angioplasty (PTA). Materials and Methods: Forty patients (24 male and 16 female; median age, 62.5 years) with dysfunctional native AVF were prospectively recruited. Digital sounds from the AVF shunt were recorded using a wireless electronic stethoscope before (pre-PTA) and after PTA (post-PTA), and the audio files were subsequently converted to mel spectrograms, which were used to construct various deep convolutional neural network (DCNN) models (DenseNet201, EfficientNetB5, and ResNet50). The performance of these models for diagnosing ≥ 50% AVF stenosis was assessed and compared. The ground truth for the presence of ≥ 50% AVF stenosis was obtained using digital subtraction angiography. Gradient-weighted class activation mapping (Grad-CAM) was used to produce visual explanations for DCNN model decisions. Results: Eighty audio files were obtained from the 40 recruited patients and pooled for the study. Mel spectrograms of "pre-PTA" shunt sounds showed patterns corresponding to abnormal high-pitched bruits with systolic accentuation observed in patients with stenotic AVF. The ResNet50 and EfficientNetB5 models yielded an area under the receiver operating characteristic curve of 0.99 and 0.98, respectively, at optimized epochs for predicting ≥ 50% AVF stenosis. However, Grad-CAM heatmaps revealed that only ResNet50 highlighted areas relevant to AVF stenosis in the mel spectrogram. Conclusion: Mel spectrogram-based DCNN models, particularly ResNet50, successfully predicted the presence of significant AVF stenosis requiring PTA in this feasibility study and may potentially be used in AVF surveillance.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.29 no.7 s.144
• /
• pp.1018-1026
• /
• 2005

Recent development of 3D scanner and software is regarded as a promising method of acquiring replicas from human body indirectly. It would be very helpful if we could predict the characteristics of 2D pattern from the simple parameters related to 3D shape for ordinary user. Therefore, in this study, investigation of 2D pattern of waist/abdominal area from the 3D geometrical model was conducted for the pattern development of waist nipper. To create body models and develop the surface of them, one ortho commonly used CAD/CAM program, IDEAS(UGS-plm solutions, USA) was used. As for the size of the models, the width, thickness, and circumference ranges of adult women's torso reported in National Anthropometric Survey of Korea (1997) were used as a standard model. Seven size variations were made by changing the width of the waist only, from 19 cm to 40 cm. Therefore, simulated body models include not only the normal body but also obese body who has wider waist and abdomen width than hip width. As results, it was found that the curvature of the unfolded 2D pattern around the abdominal area decreases as the waist width increases. As the width of the waist increases more and more, so that the comparative ratios around the torso becomes in abnormal ranges, there appears inflection points and the direction of curvature was changed. 2D Patterns obtained in this research were quantified by curvature, length of the curve and angle of deflection in the reference frame box for the convenience of the actual pattern making process. It was also possible to find that the shape of patterns of abnormal body resulted in a quite interesting change in the curves of 2D pattern, which could be applied to the custom made waist nipper for obese women.

• Journal of Computational Design and Engineering
• /
• v.2 no.4
• /
• pp.218-232
• /
• 2015

Piecewise linear (G01-based) tool paths generated by CAM systems lack $G_1$ and $G_2$ continuity. The discontinuity causes vibration and unnecessary hesitation during machining. To ensure efficient high-speed machining, a method to improve the continuity of the tool paths is required, such as B-spline fitting that approximates G01 paths with B-spline curves. Conventional B-spline fitting approaches cannot be directly used for tool path B-spline fitting, because they have shortages such as numerical instability, lack of chord error constraint, and lack of assurance of a usable result. Progressive and Iterative Approximation for Least Squares (LSPIA) is an efficient method for data fitting that solves the numerical instability problem. However, it does not consider chord errors and needs more work to ensure ironclad results for commercial applications. In this paper, we use LSPIA method incorporating Energy term (ELSPIA) to avoid the numerical instability, and lower chord errors by using stretching energy term. We implement several algorithm improvements, including (1) an improved technique for initial control point determination over Dominant Point Method, (2) an algorithm that updates foot point parameters as needed, (3) analysis of the degrees of freedom of control points to insert new control points only when needed, (4) chord error refinement using a similar ELSPIA method with the above enhancements. The proposed approach can generate a shape-preserving B-spline curve. Experiments with data analysis and machining tests are presented for verification of quality and efficiency. Comparisons with other known solutions are included to evaluate the worthiness of the proposed solution.

• Journal of the Korean Society for Precision Engineering
• /
• v.30 no.4
• /
• pp.397-402
• /
• 2013

In this study, we carried out the each lines of section, using GC (green silicon carbide) whetstone, the SCM415 material which separated by after and before heat treatments process, in 3+2 axis machining centers for integrated grinding after cutting end mill works, the spindle speed 8000 rpm and feed rate 150 mm/min. For the analysis of the centerline average roughness (Ra), we measured by 10 steps stages. Using Finite element analysis, we found the result of the load analysis effect of the assembly parts, when applied the 11 kg's load on both side of the ATC (Automatic tool change) arm. The result is as follows. For the centerline average roughness (Ra) in the non-heat treatment work pieces, are appeared the most favorable in the tenth section are $0.510{\mu}m$, that were shown in the near the straight line section which is the smallest deformation of curve. In addition, the bad surface roughness appears on the path is to long by changing angle, the more inclined depth of cut, because the chip discharging is not smoothly.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.7 no.3
• /
• pp.99-103
• /
• 1998

The curved surface machined by plate end mill causes a excess non-cutting volume, in these cases ball end mill is used for the curved surfaces. This study is aimed to obtain the optimum cutting conditions of various cutting speed, table speed, tool diameter, radius of curvature roughness on the conditions of various cutting speed, tool diameter, radius of curvature when machining the curved surface using the ball end mill. After designing curve rates, obtaining NC data by CAD/CAM system through CC-Cartesian method and transferred the data through DNC system, we machined the specimens by the CNC machining center, The surface roughness of specimens was measured by surface roughness tester and CNC 3D coordinate measuring machine. The cutting condition were the same as follow velocity; 15, 20, 25 30m/min, feed rate;40, 60, 80, 100m/min and radius of curvature; 30,40,50,60mm, tool diameters; ø8, ø12, ø16, ø 20mm. Analizing the working results, we can acquire the optimum cutting condition of curved specimen at the cutting velocity of 20~25m/min and the feed rate of 80mm/min. As the same cutting condition the best surface roughness was showed at ø16mm of the tool diameter. But the tool diameter was smaller than ø8mm. we could improve for the surface roughness by controlling the cusp.

• Journal of the Korean Society of Radiology
• /
• v.18 no.3
• /
• pp.187-201
• /
• 2024

Osteoporosis is a major health issue globally, often remaining undetected until a fracture occurs. To facilitate early detection, deep learning (DL) models were developed to classify osteoporosis using abdominal computed tomography (CT) scans. This study was conducted using retrospectively collected data from 3,012 contrast-enhanced abdominal CT scans. The DL models developed in this study were constructed for using image data, demographic/clinical information, and multi-modality data, respectively. Patients were categorized into the normal, osteopenia, and osteoporosis groups based on their T-scores, obtained from dual-energy X-ray absorptiometry, into normal, osteopenia, and osteoporosis groups. The models showed high accuracy and effectiveness, with the combined data model performing the best, achieving an area under the receiver operating characteristic curve of 0.94 and an accuracy of 0.80. The image-based model also performed well, while the demographic data model had lower accuracy and effectiveness. In addition, the DL model was interpreted by gradient-weighted class activation mapping (Grad-CAM) to highlight clinically relevant features in the images, revealing the femoral neck as a common site for fractures. The study shows that DL can accurately identify osteoporosis stages from clinical data, indicating the potential of abdominal CT scans in early osteoporosis detection and reducing fracture risks with prompt treatment.

• Journal of Preventive Medicine and Public Health
• /
• v.13 no.1
• /
• pp.27-34
• /
• 1980

Purpose of this study is to find out proper means of estimating the urinary mercury excretion in the normal individuals. Whole void volume was collected every 2 hours beginning from 6 o'clock in the morning until 6 o'clock next morning. Mercury excretion in each urine specimen was measured by NIOSH recommended dithizone colorimetric method (Method No.: P & CAM 145). Urinary concentration of mercury was adjusted by two means: specific gravity of 1.024 and a gram of creatinine excretion per liter of urine comparing the data with the unadjusted ones. Mercury excretion in 24-hour urine specimen was calculated by adding the amounts measured with the hourly collected specimens of each individual. Statistical analysis of the urinary mercury excretion revealed the following results: 1. Frequency distribution curve of mercury excreted in urine of hourly specimens was best fitted to power function expressed in the form of $y=ax^b$. Adjustment of the urinary mercury concentration by creatinine excretion was shown to be superior($y=1674x^{-1.52},\;r^2=0.95$) over nonadjustment($y=2702x^{-1.57},\;r^2=0.92$) and adjustment by specific gravity of 1.024($y=4535x^{-1.66},\;r^2=0.93$). 2. Both log-transformed mercury excretion in hourly voided specimens and mercury excretion itself in 24 hour specimens showed the normal distributions. 3. The frequency distribution of mercury adjusting the urinary concentration of mercury by creatinine excretion was best fitted to a theoretical normal distribution with the sample means and standard deviation than those unadjusted or adjusted with specific gravity of 1.024. 4. Average urinary mercury excretions in 24-hour urine specimen in an individual were as follows: a) Unadjusted mercury excretion mean and standard deviation : $$18.6{\pm}13.68{\mu}gHg/l$$. median : $$16.0\;{\mu}gHg/l$$. range : $$0.0-55.10\;{\mu}gHg/l$$. b) Adjusted with specific gravity mean : $$20.7{\pm}11.76\;{\mu}gHg/l{\times}\frac{0.024}{S.G-1.000}$$ median : $$20.7\;{\mu}gHg/l{\times}\frac{0.024}{S.G-1.000}$$ range : $$0.0-52.9\;{\mu}gHg/l{\times}\frac{0.024}{S.G-1.000}$$ c) Adjusted with creatinine excretion mean and standard deviation : $$10.5{\pm}6.98\;{\mu}gHg/g$$ creatinine/l median : $$9.4\;{\mu}gHg/g$$ creatinine/l range : $$0.0-26.7\;{\mu}gHg/g$$ creatinine/l 5. No statistically significant differences were found between means calculated from 24-hour urine specimens and those from hourly specimens transformed into logarithmic values. (P<0.05).