Research On Medical Information Processing And Its Application In Computer Aided Diagnosis

Rapid developing medical science and technology have brought massive medical information. Further processing of medical information (such as feature extraction, visualization in scientific computing, etc.) provides additional meaningful qualitative or quantitative descriptions of physiological characteristics. This is helpful in establishing new diagnostic criteria, and even automated computer-aided diagnosis. Our researches focus on medical information processing and its application in computer aided diagnosis, and consist of two parts: analysis of amplitude-integrated EEG in the newborn based on nonlinear dynamic methods and virtual visualization of human gastrointestinal tract based on nonlinear finite element method.Amplitude-integrated EEG is different from the standard EEG. It reflects the amplitude variations from the maximum to the minimum of EEG background activities. Amplitude-integrated EEG is widely used in intensive care in response to clinical demands of continuous monitoring of brain function because of its ability to record stably for hours and even for days with less data than EEG. Present researches and clinical applications of amplitude-integrated EEG focus only on observing the integrated amplitude. In this study, approximate entropy is introduced into the analysis of neonatal amplitude-integrated EEG to evaluate the irregularity of the signal for the first time. A new indicator is established based on approximate entropy and Three-Sigma criteria to quantitatively analyze amplitude-integrated EEGs in the newborn. An automatic classification algorithm based on BP neural network is also reported. Experiments with amplitude-integrated EEGs of 120 cases (32 normal and 88 abnormal of full-term infants, and 57 cases of preterm infants) are conducted to validate the effectiveness of the proposed method. Statistical results show that the recognition rates of the proposed new indicator in full-term abnormal and preterm infants are both increased by about 10%; using the BP neural network based classification algorithm, the recognition rates in full-term abnormal and preterm infants are increased by about 30% and 12% respectively.Gastrointestinal cancer is a common neoplastic disease with high mortality rate in western countries. Early detection and treatment of gastrointestinal polyps can significantly reduce the incidence and mortality of gastrointestinal cancer. As a non-invasive detection approach, virtual reality based visualization methods (such as virtual endoscopy, etc.) can not only avoid risks caused by invasive detections, but also simulate and even enrich the traditional detection methods. This study includes three aspects. First, three indicators (polyp distortion, sampling loss rate, and blind area in virtual endoscopy) are designed for the current lack of quantitative evaluation of various virtual visualization algorithms. In addition, two improved colon straightening methods (based on improved electrical field model and finite element method respectively) are proposed in order to overcome the drawbacks of current colon straightening approaches. Experiments on both simulation models and human colorectal CT data sets show the effectiveness of the proposed methods. Especially using the finite element based colon straightening method, the three evaluation indicators are reduced by 18.70%, 75.65%, 12.12%, respectively. The straightened colon is helpful in the implementation of virtual flattening and eversion with reduced distortions. Finally, we originally propose virtual insufflation in human gastrointestinal tract, and implement it with both electrical field model based and finite element method based algorithms. The virtual insuffaltion simulates the air pressure in the outer wall of hollow organs, breaks through the limitations of the real inflation. It can inflate both pipe and non-pipe shaped hollow organs, and even non-closed surfaces, while maintains the structure of polyps and folds unchanged. Inflated organs can help to increase visual fields in virtual endoscopy. They are also conductive to implement virtual flattening, eversion and other virtual visualizations bescuase of their regular shapes.