|
Of gastric cancer in China is a common malignant tumor, a serious threat to people's health. Surgery, discriminant gastric cancer and normal tissue with the naked eye doctors are required to have an extremely rich clinical experience, and the accuracy is not high. Traditional pathological examination is not only time consuming, but also for patients with a certain amount of pain and suffering. In recent years, laser Raman spectroscopy with its advantages of non-invasive, high-resolution, high sensitivity and high automation, medical testing and diagnostic caused more and more attention of scholars at home and abroad. This study is intended to explore the the surgery real-time, in situ, accurate detection of gastric cancer new methods, new technology, discriminant gastric experimental foundation for the future to help doctors surgery. Firstly, to build a 785nm near-infrared laser as an excitation light near-infrared Raman spectroscopy system. The model of human gastric cancer SGC-7901 in nude mice to the injection anesthesia laparotomy simulation surgery, and acquisition nude cancerous and normal tissues during surgery, in the body of the laser Raman spectrum of 205. Preprocessing the collection of Raman spectroscopy, the difference between the Raman spectra of normal and cancerous tissue. The results showed that: compared to normal tissue, cancerous tissue, Raman Spectroscopy, 870cm-1 (the hydroxyproline CC stretching vibration), 1450cm-1 (CH2 stretching or bending vibration), 1660 cm-1 amide Ⅰ C = O stretching vibration and water HOH variable angular vibration) at the lower intensity, while the symmetric stretching at 1007cm-1 alanine CC ring breathing vibration), 1050 cm-1 (deoxyribose in the CO stretching vibration), 1093cm-1 (phosphoric acid diester group PO2-symmetric stretching vibration), 1209cm-1 (the Phenylalanine C-C6H5 stretching vibration) at a higher intensity. In addition, the cancerous tissue spectroscopy 1330-1 (the phospholipid CH chain and nucleic acid bases CH2 vibration) at 1297cm-1 and 1331cm-1 bimodal normal tissue Raman spectroscopy is a single peak or only one acromion. Analysis of the differences between cancerous and normal tissues, we found that these differences are mainly reflected in the nucleic acids, proteins and water vibration band intensity. Cancerous tissue increase in DNA content may lead to cancerous tissue Raman spectroscopy 1050,1093 cm-1 line enhancement, and the breakdown of collagen or scattering spectrum mask may be the reason for the cancerous tissue is reduced at 870 cm-1 . We can put these features as a sign of cancer to determine whether the organization. Raman spectra of normal and cancerous tissue using support vector machine classification, classification analysis were selected linear plot kernel function, Gaussian radial basis function, polynomial kernel function, different penalty factor and nuclear parameters, 10 re-cross-validation method for training and testing. The results showed that: when the kernel function using Gaussian radial basis kernel function, penalty parameter C = 10 kernel parameter sigma = 5, the best classification results. The overall sensitivity, specificity, and accuracy were 95.73%, 70.73% and 90.73%, respectively. In order to deepen the understanding of the process of carcinogenesis, support vector machine method different growth stages of cancer tissue Raman spectroscopy staging. When the kernel function using Gaussian radial basis kernel function, the penalty parameter C = 100, nuclear parameter sigma = 15, the best classification results. The overall sensitivity, specificity, and accuracy were 98.82%, 98.73% and 98.78%, respectively. Classification results can be seen, with the support vector machine algorithm to classify, with good sensitivity, accuracy and objectivity, identify cancerous tissue in guiding surgical significance.
|