Xuechen Li

An Automatic Rib Segmentation Method on X-Ray Radiographs

In this paper, an automatic rib recognition method based on image processing and data mining is presented. Firstly, multiple template matching and graph based methods are used to detect rib center line; then, the support vector machine is used to build a rib relative position model and identify the error recognition results; finally, decision trees are employed to refine the center line recognition result. The JSRT database is employed to test our method. The result of rib recognition is over 92% for sensitivity and 98% for specificity.

A Solitary Feature-Based Lung Nodule Detection Approach for Chest X-Ray Radiographs

Lung cancer is one of the most deadly diseases. It has a high death rate and its incidence rate has been increasing all over the world. Lung cancer appears as a solitary nodule in chest x-ray radiograph (CXR). Therefore, lung nodule detection in CXR could have a significant impact on early detection of lung cancer. Radiologists define a lung nodule in CXR as “solitary white nodule-like blob.” However, the solitary feature has not been employed for lung nodule detection before. In this paper, a solitary feature-based lung nodule detection method was proposed. We employed stationary wavelet transform and convergence index filter to extract the texture features and used AdaBoost to generate white nodule-likeness map. A solitary feature was defined to evaluate the isolation degree of candidates. Both the isolation degree and the white nodule likeness were used as final evaluation of lung nodule candidates. The proposed method shows better performance and robustness than those reported in previous research. More than 80% and 93% of lung nodules in the lung field in the Japanese Society of Radiological Technology (JSRT) database were detected when the false positives per image were two and five, respectively. The proposed approach has the potential of being used in clinical practice.

Automatic Fish Recognition and Counting in Video Footage of Fishery Operations

This paper presents an accurate and automatic algorithm to recognize and count fish in the video footages of fishery operations. The unique character of the approach is that it combines machine learning techniques with statistical methods to fully make use the benefits of these algorithms. The approach consists of three major stages including video data preparation such as noise deduction, preliminary fish recognition with artificial neural network to classify image areas into either fish or non-fish, and fine fish recognition and counting with statistical shape models. Experiment results of tuna recognition and counting using the proposed method are presented with performance validation and discussion.

Intelligent tuna recognition for fisheries monitoring

Integrated video camera systems have been installed on fishing boats to trial for fishery monitoring in some countries. Currently, substantial amount of video footage is manually analyzed off the boats after each trip. Automatic processing of the videos is important for saving time and manpower. In this paper, an intelligent tuna recognition method is proposed. The method includes four steps. Firstly, the video is pre-processed by suppressing fast moving objects such as human. Secondly, the color and texture features are extracted to describe tuna, deck and other objects. Thirdly, support vector machine and statistic shape model are employed to identity and recognize tuna. Finally, a prior-knowledge based post-processing method is used to refine the recognition result. The experiment has showed that the proposed method is accurate and robust in tuna recognition. The method can also be used for other fish recognition applications, benefiting fisheries monitoring by providing efficient and automatic fish recognition.

Rib locating on chest direct radiography image using watershed algorithm and correlation matching

A rib locating method on chest direct radiography (DR) image using watershed algorithm and correlation matching is presented in this paper. Firstly, the body and spine are located by employing watershed algorithm; second, the body model is selected to remove other bones outside body; thirdly, the models of left and right ribs are resized and rotated to fit ribs of each side respectively; finally, the rib regions are extracted, each one of which contains only one rib. 70 DR images are used to test the method. The experiment result shows that the average error rate, accuracy, and sensitivity are respectively 0.067, 0.828 and 0.862.