Abstract： Diseases prognosis is often influenced by multiple factors, and some intricate non-linear relationships exist among those factors. Artificial neural network （ANN）, an artificial intelligence model, simulates the work mode of biological neurons and has a b capability to analyze multi-factor non-linear relationships. In recent years, ANN is increasingly applied in clinical medical fields, especially for the prediction of disease prognosis. This article focuses on the basic principles of ANN and its application in disease prognosis research.
Objective To investigate the value of back propagation (BP) neural network for recognizing gastric cancer cell. Methods A total of 510 cells was selected from 308 patients. There were 210 gastric adenocarcinoma cells and 300 non-cancer gastric cells. Ten morphological parameters were measured for each cell. These data were randomly divided into two groups: training dataset (A) and test dataset (B). A three-layer BP neural network was built and trained by using dataset A. The network was then tested with dataset A and B.Results For data A, the sensitivity of network was 99%, specificity 99%, positive predictive value 98%, negative predictive value 99%, and accuracy 99%. For data B, the sensitivity of network was 99%, specificity 97%, positive predictive value 96%, negative predictive value 99%, the accuracy 98%. With receiver operator characteristic (ROC) curve evaluation, the area under ROC curve was 0.99.Conclusion The model based on BP neural network is very effective. A BP neural network can be used for effectively recognizing gastric cancer cell.
目的 建立基于反传（BP）神经网络技术的甲状腺癌诊断模型，并评估该模型的临床应用价值。方法 回顾性分析2010年1月至2011年8月期间南京市鼓楼医院收治的甲状腺癌患者103例及甲状腺良性病变患者51例，提取其超声图像的9个特征，循建模规则，建立基于BP神经网络技术的甲状腺癌诊断模型，依此模型对2011年9月至2011年12月期间收治的根据超声图像特征疑为甲状腺癌的42例患者进行术前诊断，其结果与术后病理诊断结果（术后病理诊断为甲状腺癌32例，甲状腺良性病变10例）进行对比研究。结果 甲状腺癌诊断模型对建模样本的诊断准确率为95.45%（147/154）；术前样本的诊断准确率为90.48%（38/42）；所有样本的诊断准确率为94.39% （185/196）。结论 从本组有限的病例结果初步得出，基于BP神经网络技术的甲状腺癌诊断模型具有较高的可行性及可靠性，可望成为一种全新的甲状腺癌辅助诊断方法。
Generally, P-wave is the wave of low-frequency and low-amplitude, and it could be affected by baseline drift, electromyography (EMG) interference and other noises easily. Not every heart beat contains the P-wave, and it is also a major problem to determine the P-wave exist or not in a heart beat. In order to solve the limitation of suiting the diverse morphological P-wave using wavelet-amplitude-transform algorithm and the limitation of selecting the pseudo-P-wave sample using the wavelet transform and neural network, we presented new P-wave detecting method based on wave-amplitude threshold and using the multi-feature as the input of neural networks. Firstly, we removed the noise of ECG through the wavelet transform, then determined the position of the candidate P-wave by calculating modulus maxima of the wavelet transform, and then determine the P-wave exist or not by wave-amplitude threshold method initially. Finally we determined whether the P-wave existed or not by the neural networks. The method is validated based on the QT database which is supplied with manual labels made by physicians. We compared the detection effect of ECG P-waves, which was obtained with the method developed in the study, with the algorithm of wavelet threshold value and the method based on "wavelet-amplitude-slope", and verified the feasibility of the proposed algorithm. The detected ECG signal, which is recorded in the hospital ECG division, was consistent with the doctor's labels. Furthermore, after detecting the 13 sets of ECG which were 15min long, the detection rate for the correct P-wave is 99.911%.
Multi-layer perceptron (MLP) neural network belongs to multi-layer feedforward neural network, and has the ability and characteristics of high intelligence. It can realize the complex nonlinear mapping by its own learning through the network. Bipolar disorder is a serious mental illness with high recurrence rate, high self-harm rate and high suicide rate. Most of the onset of the bipolar disorder starts with depressive episode, which can be easily misdiagnosed as unipolar depression and lead to a delayed treatment so as to influence the prognosis. The early identification of bipolar disorder is of great importance for patients with bipolar disorder. Due to the fact that the process of early identification of bipolar disorder is nonlinear, we in this paper discuss the MLP neural network application in early identification of bipolar disorder. This study covered 250 cases, including 143 cases with recurrent depression and 107 cases with bipolar disorder, and clinical features were statistically analyzed between the two groups. A total of 42 variables with significant differences were screened as the input variables of the neural network. Part of the samples were randomly selected as the learning sample, and the other as the test sample. By choosing different neural network structures, all results of the identification of bipolar disorder were relatively good, which showed that MLP neural network could be used in the early identification of bipolar disorder.
A new method based on convolution kernel compensation (CKC) for decomposing multi-channel surface electromyogram (sEMG)signals is proposed in this paper. Unsupervised learning and clustering function of self-organizing map (SOM) neural network are employed in this method. An initial innervations pulse train (IPT) is firstly estimated, some time instants corresponding to the highest peaks from the initial IPT are clustered by SOM neural network. Then the final IPT can be obtained from the observations corresponding to these time instants. In this paper, the proposed method was tested on the simulated signal, the influence of signal to noise ratio (SNR), the number of groups clustered by SOM and the number of highest peaks selected from the initial pulse train on the number of reconstructed sources and the pulse accuracy were studied, and the results show that the proposed approach is effective in decomposing multi-channel sEMG signals.
To explore the self-organization robustness of the biological neural network, and thus to provide new ideas and methods for the electromagnetic bionic protection, we studied both the information transmission mechanism of neural network and spike timing-dependent plasticity (STDP) mechanism, and then investigated the relationship between synaptic plastic and adaptive characteristic of biology. Then a feedforward neural network with the Izhikevich model and the STDP mechanism was constructed, and the adaptive robust capacity of the network was analyzed. Simulation results showed that the neural network based on STDP mechanism had good rubustness capacity, and this characteristics is closely related to the STDP mechanisms. Based on this simulation work, the cell circuit with neurons and synaptic circuit which can simulate the information processing mechanisms of biological nervous system will be further built, then the electronic circuits with adaptive robustness will be designed based on the cell circuit.
Based on bioelectrical impedance theory and pattern recognition algorithm, we in this study measured varieties of people's bioelectrical impedance in hands and identified different people according to their bioelectrical impedance. We designed a bioelectrical impedance collection circuit with AD5933 chip to measure the impedance in different people's hands, and we obtained the bioelectrical impedance spectrum for each person under 1-100 kHz electrical stimulation. We calculated the segmentation slopes of bioelectrical impedance spectrum, and took the slopes as characteristic parameters. In order to promote the recognition rate and prevent the overfitting of the model, we divided the people into the training set and the test set, and designed a 3 layer back propagation neural network model to train and test the samples. The results showed that back propagation neural network model could identify the test set effectively. The recognition rate of the training sets was as high as 97.62%, recognition rate of validation sets was 88.79%, recognition rate of test sets was 86.34%, and the synthetical recognition rate was 94.22%. It gives a clue that the network can perfectly recognize people in the training network as well as strangers that comes from the outside of the tests. Our work can verify the feasibility and reliability of using bioelectrical impedance and pattern recognition algorithm for identification, and can provide a simple and supplementary way to identify people.
Polyvinyl alcohol (PVA) hydrogel was made for simulating human's soft tissue in our experiment. The image acquisition device is composed of an optical platform, a camera and its bracket and a light source. In order to study the law of soft tissue deformation under flexible needle insertion, markers were embedded into the soft tissue and their displacements were recorded. Based on the analysis of displacements of markers in X direction and Y direction, back propagation (BP) neural network was employed to model the displacement of Y direction for the markers. Compared to the experimental data, fitting degree of the neural network model was above 95%, the maximum relative error for valid data was limited to 30%, and the maximum absolute error was 0.8 mm. The BP neural network model was beneficial for predicting soft tissue deformation quantitatively. The results showed that the model could effectively improve the accuracy of flexible needle insertion into soft tissue.
There are a great number of uncertainties in medical practice, causing considerable difficulties in medical activities such as diagnosis and prognostic prediction. Neural-fuzzy system (NFS) combines the advantages of artificial neural networks and fuzzy logic very well, and has become a new type of artificial intelligence model which is capable of acquiring knowledge from data and expressing it in the form of fuzzy rules. Because of its strong capability of classification and processing fuzzy information, NFS is more and more used in medical practice. Adaptive neural-fuzzy inference system (ANFIS) is one of the most popular forms of NFS. This review focuses on the use of ANFIS in medical practice.