KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 7 of 7

Some of the metrics are blocked by your
consent settings
A Cascade of Encoder–Decoder with Atrous Convolution and Ensemble Deep Convolutional Neural Networks for Tuberculosis Detection
(2025-07-01)
Maneerat, Noppadol
;
Narkthewan, Athasart
;
Hamamoto, Kazuhiko
Tuberculosis (TB) is the most serious worldwide infectious disease and the leading cause of death among people with HIV. Early diagnosis and prompt treatment can cut off the rising number of TB deaths, and analysis of chest X-rays is a cost-effective method. We describe a deep learning-based cascade algorithm for detecting TB in chest X-rays. Firstly, the lung regions were segregated from other anatomical structures by an encoder–decoder with an atrous separable convolution network—DeepLabv3+ with an XceptionNet backbone, DLabv3+X, and then cropped by a bounding box. Using the cropped lung images, we trained several pre-trained Deep Convolutional Neural Networks (DCNNs) on the images with hyperparameters optimized by a Bayesian algorithm. Different combinations of trained DCNNs were compared, and the combination with the maximum accuracy was retained as the winning combination. The ensemble classifier was designed to predict the presence of TB by fusing DCNNs from the winning combination via weighted averaging. Our lung segmentation was evaluated on three publicly available datasets: it provided better Intercept over Union (IoU) values: 95.1% for Montgomery County (MC), 92.8% for Shenzhen (SZ), and 96.1% for JSRT datasets. For TB prediction, our ensemble classifier produced a better accuracy of 92.7% for the MC dataset and obtained a comparable accuracy of 95.5% for the SZ dataset. Finally, occlusion sensitivity and gradient-weighted class activation maps (Grad-CAM) were generated to indicate the most influential regions for the prediction of TB and to localize TB manifestations.
Some of the metrics are blocked by your
consent settings
Apply PLC to Control 3-Axis Machines
(2025-01-01)
Cheypoca, Thepjit
;
Thaiyai, In
;
Maneerat, Noppadol
;
Narkthewan, Athasart
This report presents a method of using PLC (Programable Logic Control) to control the work instead of the 3-Axis machine position controller. PLC is popularly used in various types of control systems. It has a lot of control capabilities, is efficient, stable, durable, easy to use, and cheap. It is the use of commands to control the work of all 3 axes, 2 axes, and 1 axle. The commands used for moving in both straight lines and curves, and there are other devices that need to be controlled together with the machine that is being controlled.
Some of the metrics are blocked by your
consent settings
Glioma Brain Tumor Classification using Transfer Learning
(2024-01-01)
Pilaoon, Pongsak
;
Narkthewan, Athasart
;
Wadlom, Noppanat
;
Varakulsiripunth, Ruttikorn
;
Hamamoto, Kazuhiko
In this research the glioma brain tumor binary classification using transfer learning was introduced. The MRI image from 2 datasets comprised with REMBRANDT and BraTS2021 with increasing number of MRI images was proposed to prevent overfitting problem. MRI images were converted to JPEG format and heavily imbalanced with normal brain image is minority class. Morphological operation was used to remove skeletons and artifacts from brain region. We have introduced Contrast Limited Adaptive Histogram Equalization to preprocess and enhance contrast before classify using various CNNs. To handle imbalanced dataset problem, we proposed image augmentation to increase the number of images and obtain balanced dataset. The various CNNs transfer learning was implemented to classify glioma brain tumor. Finally, the best classifier is InceptionV3 with balanced dataset that obtained accuracy 99.19%, sensitivity 98.83%, and specificity 100% respectively, better than our past research work.
Some of the metrics are blocked by your
consent settings
Detection of hard exudate for diabetic retinopathy using unsupervised classification method
(2020-07-01)
Maneerat, Noppadol
;
Thongpasri, Teerapon
;
Narkthewan, Athasart
;
Kimpan, Chom
Diabetic retinopathy (DR) causes retinal disorders such as blood vessel blockage, the leaks of blood, and the proteins in water bleeding into the tissues of retina. All of the symptoms lead to the destruction of retina resulting in reduced visibility or finally lose vision. Therefore, this study presents an image processing method to extract hard exudates in the retinal image, which is a serious symptom of diabetic retinopathy using an unsupervised classification method. The proposed hard exudates extraction method composes of 3 steps. Firstly, the optic disc similar to hard exudate is eliminated from the retinal image. Subsequently, the green channel of the RGB color model is selected for data analysis because it represents all hard exudates better than the red and blue channels. The features of hard exudates in the retinal image are then extracted by various methods such as dilation, erosion, entropy analysis, and standard deviation analysis and it also appeared in many dimensions. Finally, the proposed method uses k-mean, which is an unsupervised classification technique for hard exudates clustering. The determination of hard exudates from the retinal image is achieved using two datasets (DIARETDB0 and DIARETDB1). These datasets are usually used for algorithm efficiency analysis to retinal image evaluation. The results show that the maximum specificity is approximate 97%. It indicates that the proposed method can be applied for the automatic detection of diabetic retinopathy.
Some of the metrics are blocked by your
consent settings
Automatic localization of optic disc in fundus image using iterative background removal
(2019-07-01)
Narkthewan, Athasart
;
Maneerat, Noppadol
Diabetic retinopathy (DR) is leading to cause of blood clot stimulating the formation of abnormal new blood vessels. hemorrhage and protein secretion from blood vessels to retinal tissue. The retinal damage causes the loss of sight thus the localization of Optic Disc is necessary for analysis of the abnormal retinal image. However, the data located at the Optic Disc image are similar to Hard Exudates in the retinal image. Therefore, the aim of the present study is to detect Optic Disc using the technique of background removal. The principle of image processing is also applied for the retinal image. The image data from Fundus camera are recorded in RGB color model separated into 3 channels: Red, green and blue channel. The data of all channels are evaluated by the preprocessing algorithm to make a more clear appearance of Optic Disc. Subsequently, the preprocessing image is analyzed by iterative background removal using entropy evaluation of the image data. Finally., the variance analysis of data intensity is performed in both the horizontal and vertical axis to determine the localization and size of Optic Disc. The result shows that the accuracy of localization and size of Optic Disc is approximately 98%. It indicates that this method is useful for retinal image analysis of diabetic retinopathy.
Some of the metrics are blocked by your
consent settings
Retina blood vessel detection for diabetic retinopathy diagnosis
(2019-03-28)
Narkthewan, Athasart
;
Maneerat, Noppadol
Diabetes affects the microangiopathy in the retina which causes to retinal disorders such as blood vessel blockage then the abnormal blood vessel is occurred. The microvascular leakage will decrease or loss of sight. The aim of this research is to find the retinal blood vessel detection method for diagnosis of diabetic retinopathy. This study was carried out using the principle of image processing to analyze the retina image. The green channel was used for data processing. Consequently, several image processing techniques were applied to the green channel data as image enhancement, scaling, morphological operator and filter to extract the features of the retinal blood vessel in the retina image. The retinal blood vessel was extracted and displayed on the screen for diagnosis. The efficiency of algorithm for the retinal blood vessel detection was presented in this study. All different twenty retinal images from the DRIVE database were tested for blood vessel extraction. The error detection data was compared with the ground truth image. The results show that the maximum specificity and accuracy were 99.66% and 96.80%, respectively. It indicated that the proposed method could detect the blood vessel from retina image.
Some of the metrics are blocked by your
consent settings
Silk's Texture Defect Detection by Wavelet Transform
(2003-10-16)
Narkthewan, Athasart
;
Chitsakul, Kitiphol
At present, Thai textile industries still mostly employ an expert team as inspector in product quality control. This is a time consuming process and could provoke the human related errors, This research presents a novel technique to automatic detect and classify some deflections of texture of silks on a microcomputer. Based on the wavelet transforms, silk's images, size 256 × 256 pixels digitally collected were analyzed in three resolutions in which some kinds of deflection such as losing of weft, warp or holds appeared. The statistical method such as analysis of variance, ANOVA has been used to classify such deflects on images. The performances of the technique were verified by using simulated texture images and real scanned silk's images with some artificial deflects.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results