Panrasee Ritthipravat

A modified approach to fuzzy Q learning for mobile robots

A modified approach to fuzzy Q-learning is presented in this paper. A reward sharing mechanism is added to increase the learning speed and to allow treatment of each fuzzy rule as a separate learning node. A new method of exploration is also proposed to increase the learning performance. Two basic robot behaviours which are a goal-seeking and an obstacle avoidance behaviour are simulated to show the promise of the proposed techniques. The goal-seeking behaviour is implemented on a real robot. The experimental results show that this method is practical for a real-world problem.

Nasopharyngeal carcinoma segmentation using a region growing technique

PurposeThis paper proposes a new image segmentation technique for identifying nasopharyngeal tumor regions in CT images. The technique is modified from the seeded region growing (SRG) approach that is simple but sensitive to image intensity of the initial seed.MethodsCT images of patients with nasopharyngeal carcinoma (NPC) were collected from Ramathibodi hospital, Thailand. Tumor regions in the images were separately drawn by three experienced radiologists. The images are used as standard ground truth for performance evaluation. From the ground truth images, common sites of nasopharyngeal tumor regions are different from head to neck. Before the segmentation, each CT image is localized: above supraorbital foramen (Group I), below oropharynx (Group III), or between these parts (Group II). Representatives of the CT images in each part are separately generated based on the Self-Organizing Map (SOM) technique. The representative images contain invariant features of similar NPC images. For a given CT slice, a possible tumor region can be approximately determined from the best matching representative image. Mode intensity within this region is identified and used in the SRG technique.ResultsFrom 6,606 CT images of 31 NPC patients, 578 images contained the tumors. Because NPC images above the supraorbital foremen were insufficient for study (6 images from 1 subject), they were excluded from the analysis. The CT images with inconsistent standard ground truth images, metastasis cases, and bone invasion were also disregarded. Finally, 245 CT images were taken into account. The segmented results showed that the proposed technique was efficient for nasopharyngeal tumor region identification. For two seed generation, average corresponding ratios (CRs) were 0.67 and 0.69 for Group II and Group III, correspondingly. Average PMs were 78.17 and 82.47%, respectively. The results were compared with that of the traditional SRG approach. The segmentation performances of the proposed technique were obviously superior to the other one. This is because possible tumor regions are accurately determined. Mode intensity, which is used in place of the seed pixel intensity, is less sensitive to the initial seed location. Searching nearby tumor pixels is more efficient than the traditional technique.ConclusionA modified SRG technique based on the SOM approach is presented in this paper. Initially, a possible tumor region in a CT image of interest is approximately localized. Mode intensity within this region is determined and used in place of the seed pixel intensity. The tumor region is then searched and subsequently grown. The experimental results showed that the proposed technique is efficient and superior to the traditional SRG approach.

Automatic segmentation of nasopharyngeal carcinoma from CT images: Region growing based technique

This paper describes a framework for automatic nasopharyngeal carcinoma segmentation from CT images. The proposed technique is based on the Region Growing Method. It is automatic segmentation in which an initial seed is generated without human intervention. The seed is generated from a probabilistic map representing the chances of it being tumor. This map is created from three probabilistic functions based on location of the tumor, intensities, and non-tumor region respectively. The pixel in which the probability is the highest will be selected as potential seeds. Only one representative of these seeds will be selected as an initial seed. Then the seed will be used for region growing subsequently. The experimental results showed that the potential seeds and initial seed were correctly determined with a percentage accuracy of 81.60% and 95.10%. The seed was grown in preprocessed CT images for identifying the nasopharyngeal carcinoma region. The results showed that, perfect match and corresponding ratio were 71.31% and 53.00% respectively

Automatic Segmentation of Nasopharyngeal Carcinoma from CT Images

This paper presents an automatic segmentation technique for identifying nasopharyngeal carcinoma regions in CT images. The proposed technique is based on the region growing method by which an initial seed is automatically generated. A probabilistic map representing a chance of being the tumor pixel in each CT image will be created and used for initial seed determination. This map is generated from three probabilistic functions established upon location of the tumor considered, intensities of the tumor pixels, and asymmetry of organs respectively. A representative of potential tumor pixels will be selected as an initial seed. The experimental results showed that seeds were correctly determined with the percent accuracy of 84.32%. These seeds were grown in preprocessed CT images for identifying the nasopharyngeal carcinoma regions subsequently. The results showed that, for no metastasis cases, perfect match and corresponding ratio were 85.03% and 52.44% respectively and 29.26% and 28.03% correspondingly for metastasis cases. This resulted from a single seed generated in each CT image. It was unable to identify more than one tumor region.

Comparison and analysis of expertness measure in knowledge sharing among robots

Robot expertness measures are used to improve learning performance of knowledge sharing techniques. In this paper, several fuzzy Q-learning methods for knowledge sharing i.e. Shared Memory, Weighted Strategy Sharing (WSS) and Adaptive Weighted Strategy Sharing (AdpWSS) are studied. A new measure of expertise based on regret evaluation is proposed. Regret measure takes uncertainty bounds of two best actions, i.e. the greedy action and the second best action into account. Knowledge sharing simulations and experiments on real robots were performed to compare the effectiveness of the three expertness measures i.e. Gradient (G), Average Move (AM) and our proposed measure. The proposed measure exhibited the best performance among the three measures. Moreover, our measure that is applied to the AdpWSS does not require the predefined setting of cooperative time, thus it is more practical to be implemented in real-world problems.

Adaptive Thresholding based on SOM Technique for Semi-Automatic NPC Image Segmentation

This paper studies Self-Organizing Map (SOM) based adaptive thresholding technique for semi-automatic image segmentation. CT images of patients with nasopharyngeal carcinoma are considered in the study. The thresholds are determined from histogram of a topological map created from SOM method. With this proposed technique, initial tumor pixel must be manually selected. Pixels which are in the same threshold level are considered as tumor pixels. The experimental results showed that our proposed technique is effective for NPC image segmentation. In addition, it can properly handle tumor heterogeneity generally found in the NPC images.

Dealing with missing values for effective prediction of NPC recurrence

This paper aims to investigate missing data techniques for effective prediction of nasopharyngeal carcinoma (NPC) recurrence. The techniques include listwise deletion, imputations by mean, k-nearest neighbor, and expectation maximization. The completed data are used to predict the presence or absence of NPC recurrence in each year by means of logistic regression, multilayer perceptron with backpropagation training, and naive bayes. Five year predictions are carried out. Validity of each predictive model is assured by 10-fold cross validation. Their results are compared in order to determine proper missing data treatment and the most efficient prediction technique. The results showed that EM imputation was superior to the other missing data techniques because it can be efficiently applied to all predictive models. The multilayer perceptron with backpropagation training gave the highest prediction performance and it was the most robust to the data completed by different missing data techniques.

Prediction of nasopharyngeal carcinoma recurrence by neuro-fuzzy techniques

Neuro-fuzzy techniques for prediction of nasopharyngeal carcinoma recurrence are mainly focused in this paper. A technique, named Generalized Neural Network-type Single Input Rule Modules connected fuzzy inference method is proposed. In the study, clinical data of patients with nasopharyngeal carcinoma were collected from Ramathibodi hospital, Thailand. In total, 495 records were taken into account. Relevant factors were extracted and employed in developing predictive models. The results showed that the proposed technique was superior to the other neuro-fuzzy techniques, stand-alone neural network, logistic regression and Cox proportional hazard model. Accuracy and AUC above 80% and 0.8 could be achieved. To show validity of the proposed technique, two nonlinear problems, i.e., function approximation and the XOR classification problems, are studied. Simulation results showed that the proposed technique could simplify the problem by converting the original nonlinear input into the lower complexity one. In addition, it can solve the XOR problem whereas the traditional approach cannot tackle this problem.

Artificial Neural Networks in Cancer Recurrence Prediction

This paper aims to review the use of artificial neural networks (ANNs) in prediction of cancer recurrence. The sources of publications were randomly selected from PUBMED database, IEEE explore, and the google search engine with the keywords for searching as “recurrence” or “relapse” or “disease free” + “neural network” + “cancer”. Increasing of the predictive performance was considered. In addition, handling incomplete data and feature selection techniques usually employed in this application were examined.

Comparison of neuro-fuzzy based techniques in nasopharyngeal carcinoma recurrence prediction

This paper aims to compare neuro-fuzzy based techniques for effective prediction of nasopharyngeal carcinoma (NPC) recurrence. The techniques include an artificial neural network (ANN), adaptive neuro-fuzzy inference systems (ANFIS), the functional-type single input rule modules connected fuzzy inference method (F-SIRMs method) and the functional and neural network type SIRMs method (F-NN-SIRMs method). All models are produced to predict the presence or absence and timing of the NPC recurrence. Five years predictions are carried out. Validity of each predictive model is assured by 10-fold cross validation. The results show that the F-NN-SIRMs method is superior to the other techniques in a sense that it provides the higher prediction performance.