KMITL

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

Browse

Search Results

Now showing 1 - 10 of 35

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
Brain Tumor Classification using Pretrained Deep Convolutional Neural Network
(2023-01-01)
Pilaoon, Pongsak
;
Maneerat, Noppadol
;
Nakthewan, Athasart
;
Varakulsiripunth, Ruttikorn
;
Hamamoto, Kazuhiko
In this research deep learning convolution neural network (CNN) has been implemented for binary classification GBM brain tumor. The dataset from REMBRANDT database that comprise of 155 MRI images has been utilized in this research. The transfer learning by pretrained network namely GoogleNet and AlexNet have been conducted to classify the GBM brain tumor form normal brain. The advantage of classification by transfer learning is the manual segmentation and feature extraction were replaced by automatic procedure that reduce the error from human handcraft. The prediction result by using Googlenet pretrained network shown accuracy result 80.85% and Alexnet pretrained network obtained accuracy 93.62%. The GBM brain tumor classification by deep learning pretrained network obtained good result of accuracy and can be implemented in practical to help the medical staff for earlier diagnosis for further treatment and increasing survivor rate of patients. The manual adjustment for segmentation and feature extraction are improved by automatic classification using deep learning pretrained network is main advantage of this research. The future work we will try to implement with increasing images dataset to improve accuracy, robustness testing and prevent overfitting problem.
Some of the metrics are blocked by your
consent settings
Brain Tumor Classification using Supervised Support Vector Machine
(2023-01-01)
Pilaoon, Pongsak
;
Maneerat, Noppadol
;
Yajima, Kuniaki
;
Varakulsiripunth, Ruttikorn
;
Hamamoto, Kazuhiko
A Glioblastoma (GBM) is a malignant brain tumor earlier detection and diagnosis will increase survival opportunities. This research has developed for binary classification of GBM brain tumors by supervised machine learning from magnetic resonance imaging (MRI). DICOM medical images have been converted into JPEG files and morphological operation has been implemented to separate the brain region from the skull image for preparation and easier for tumor segmentation in preprocessing stage. The global thresholding segmentation has been proposed to segment the brain tumor from the artifact and then the features have been extracted by gray level coefficient matrix feature extraction (GLCM). In this research, a support vector machine has been conducted for binary classification and finally, GBM grade-4 brain tumor is distinguished from normal brain images. The dataset comprises 155 MRI images 80% has been assigned for training and another 20% will be the testing dataset. The experimental output prediction result is 96.875 % accuracy, 95 % sensitivity, and 100% specificity. The performance of classification has been improved and shown better results when compared with previous research work.
Some of the metrics are blocked by your
consent settings
An improved 2DPCA for face recognition under illumination effects
(2015-01-01)
Woraratpanya, Kuntpong
;
Sornnoi, Monmorakot
;
Leelaburanapong, Savita
;
Titijaroonroj, Taravichet
;
Varakulsiripunth, Ruttikorn
Principal component analysis (PCA) is one of the successful techniques for applying to face recognition, but its challenge still remains for solving an illumination effect condition. This paper proposes an improved 2DPCA (I-2DPCA) for overwhelming the illumination effect in face recognition. The proposed method is based on two assumptions. The first assumption is to create the covariance matrix that can effectively decompose the components of illumination effects from the eigenfaces. This avoids the illumination effect problem. The second assumption is to select the suitable eigenvectors that can significantly improve the recognition rate. Based on the Extended Yale Face Database B+ containing 60 illumination conditions, the experimental results show that not only does the proposed method decrease the computing time, but it also improves the recognition rate up to 95.93%.
Some of the metrics are blocked by your
consent settings
Blood warmer using peltiers
(2014-01-01)
Suksantisakul, Park
;
Maneerat, Noppadol
;
Nakprik, Thongchan
;
Varakulsiripunth, Ruttikorn
The purpose of this paper is introduces blood warmer system to easy for anesthesiologists and patients for infusion and facilitate of anesthesiologists work for using blood bag. The system is contained in a rectangular box, inside of which consists of two separate chambers: one housing the electronics inside have; circulating fans, heat sinks, transformer (12v, 100w, 8.3amp) and a PCB with an SG3525 IC for regulation of temperatures in the adjacent chamber and the other the blood warming components inside have is fitted with a shaker, EPDM(Ethylene Propylene Diene Monomer) cushion sheets, a digital thermometer, and two thermo-sensor. Two peltiers and EPDM cushion sheets are employed in the proposed system to quicken the warming of blood bags. Nonetheless, three experiments were conducted with three different blood warming configurations: using one peltier without the EPDM cushion sheets, two peltiers without the EPDM cushion sheets, and two peltiers with the EPDM cushion sheets. The finding indicates that the use of two peltiers together with the plates produces the best results. © 2014 SERSC.
Some of the metrics are blocked by your
consent settings
An ontology-based multi agent autotnotive parts transportation management system
(2013-12-01)
Phutthisathian, Areeworn
;
Maneerat, Noppadol
;
Varakulsiripunth, Ruttikorn
;
Takahashi, Kaoru
;
Kato, Yasushi
This paper presents an ontology-based multi agent automotive parts transportation system. The system is used to decide for a route of transportation by Dijkstra's algorithm and Ontology Concept to find the shortest path. This system collects the traffic data and vehicle's position for a suitable path decision. So the user can control and monitor the automotive parts that are transported to production lines, especially in the traffic jam. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
Feature extraction from retinal fundus image for early detection of diabetic retinopathy
(2013-12-01)
Sreng, Syna
;
Takada, Jun Ichi
;
Maneerat, Noppadol
;
Isarakorn, Don
;
Varakulsiripunth, Ruttikorn
Automated detection of lesions in retinal fundus image can be aid in the detection of diabetic retinopathy. Exudates are the early sign of diabetic retinopathy so the proper detection of these lesions is an essential task in an automatic retinal screening. On the research work leading to automatic analysis of exudate detection, the knowledge of Optic Disk (OD) location is very useful. An efficient algorithm is presented to detect the OD and exudate which are the most important features for early detection of diabetic retinopathy. From a retinal fundus image, the proposed method first preprocesses and estimates the histogram of retinal background, then filters out the bright pixels in intensity image. They include OD, and non-OD (exudates and noise). Next, an OD boundary is determined and eliminated after applying blob boundary measurement and morphological reconstruction. Finally, exudates are extracted by applying the maximum entropy thresholding to filter out the bright pixels from the green component of retinal image which OD region inside is eliminated. The proposed technique has been tested first on 100 images from hospital. Experimental results show that 93% and 89% of OD and exudate were detected correctly, respectively. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
Load distribution using modified RED for multipath TCP communication
(2013-01-01)
Prabhavat, Sumet
;
Varakulsiripunth, Ruttikorn
;
Utsumi, Satoshi
;
Kato, Yasushi
Demand for large bandwidth network connections is growing continuously while network connections via multiple paths can exist because of high degree of connectivity. Utilization of the multiple paths is an economical solution for provisioning large network capacity to meet the users' requirements but it brings about key issues such as load imbalance (causing a large queue and buffer overflow at a gateway) and packet reordering problems that have significant impacts on TCP connections. In this paper, we propose a load distribution model for TCP transmission, called Load Distribution over Multipath with Extended Drop Slope Random Early Detection (LDM/ExRED). LDM/ExRED takes advantages of RED mechanism to maintain a small queue, avoid buffer overflow at a gateway, and mitigate packet reordering problem. In addition, it reduces peak buffer occupancy by using our modified RED (i.e., ExRED). Our work will be evaluated and compared to the other existing models by simulations under realistic traffic conditions. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
Automatic microaneurysms detection through retinal color image analysis
(2013-01-01)
Yunuch, Preeyaporn
;
Maneerat, Noppadol
;
Isarakorn, Don
;
Pasaya, Bundit
;
Panjaphongse, Ronakorn
This paper proposes an automatic system to diagnose the diabetic retinopathy symptom, which can cause a loss of vision by analysis the abnormality in retinal image. Digital image processing system is developed for the retinal image analysis which helps ophthalmologists to identify diabetic patients. The retinal images derived from ophthalmologists are used to analysis by using HSV, area identification and eccentricity techniques to distinguish diabetic retinopathy symptoms from normal diabetic patients. First color bar is evaluated by using HSV method and then using the eccentricity technique with area of pixel to find out the abnormality of Microaneurysms (MAs). The accuracy result of experiment is around 93% when compares to the analysis of ophthalmologists. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
Automatic exudate extraction for early detection of Diabetic Retinopathy
(2013-01-01)
Sreng, Syna
;
Takada, Jun Ichi
;
Maneerat, Noppadol
;
Isarakorn, Don
;
Pasaya, Bundit
Diabetic Retinopathy (DR) is the most common cause of blindness in diabetic patients, but early detection and timely treatment can prevent this problem. Exudates have been found to be one of the signs and serious DR anomalies so the proper detection of these lesions and the treatment should be done immediately to prevent loss of vision. The aim of this study is to automatically detect these lesions in fundus images. To achieve this goal, the proposed method first preprocesses to improve the quality of fundus image, and then Optic Disc (OD) is detected and eliminated to prevent the interference to the result of exudate detection by combination of 3 methods; image binarization, Region Of Interest (ROI) based segmentation and Morphological Reconstruction (MR). Next, exudates are detected by applying the maximum entropy thresholding to filter out the bright pixels from the result of OD region eliminated. Since the result contains some noises which appear as bright light at the edge of fundus area in some images, that affect is considered and eliminated to improve the result of false positive. Finally, exudates are extracted by using MR. The proposed technique has been tested on 100 fundus images from hospital. Experimental results show that 91 % of exudate is extracted correctly with the average process of 3.92 second per image. © 2013 IEEE.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results