A Novel Wrist Pulse Signal Processing And Lung Cancer Detection Using ISW Algorithm And Gaussian SVM

Lung cancer is one of the major disease leading to the high incidence of death worldwide.In China the number of lung cancer patients has been increasing due to the environment pollution.Currently,Computed Tomography(CT)is the major method for lung cancer detection,and this usually requires people to join in the queue of the Medical imaging and radiotherapy room in hospital.Thus,the inconvenient CT detection could not monitor the healthy status of the potential lung cancer patients in daily life.It is more than 70% lung cancer patients that are in the metaphase and terminal when the first time diagnose.However,more than 60% of the lung cancer patients who are diagnosed in the early stage could continue the lifetime approximately five years under the proper treatment.In recent years,the improvement of wearable devices,Internet of things and artificial intelligence provides the main technique of designing the Smart health(and Smart home),which supports the application of the proposed method,designed mainly for care giving service at home.It is both important and quite helpful for users who might have lung cancer to be warned conveniently at an earlier stage during daily life so that they can go to the specialists for further medical examinations to get proper treatment as early as possible.Such methods using signal processing and machine learning algorithms that are the essential part of the abovementioned smart system.To be the best of our knowledge,this is the first time that the wrist pulse signal has been used to diagnose lung cancer.Although the existing literature shows a few work of identifying the patients having specific disease,most of the related work is still in the stage of designing hardware acquisition device and traditional pulse waveform objectification.On the one hand,the traditional pulse diagnosis theory is complex and is still to be improved.On the other hand,there is a lack of research on capturing the pathological characteristics of pulse signal.In this work,a novel lung cancer detection method is designed having insight on the Jin's pulse diagnosis using the real dataset collected by doctors from Shandong Institute of traditional Chinese Medicine.Due to the period(in time domain)of the wrist pulse signal implying pathological characteristics of lung cancer is uncertain,the time-series based method of de-noising and segmenting the pulse signal is not suitable for processing such pulse signal.That is,under the condition that the abnormal signals with pathological characteristics of lung cancer are preserved,the noise and baseline drift are removed,and the continuous signals with pathological characteristics of lung cancer are accurately segmented.Moreover the methods mentioned in the literature are lack of standard for pulse signal characteristics extraction,and there is no research which have mentioned the pulse signal characteristics of lung cancer patients combining with the theory of Jin's pulse diagnosis theory.Therefore,this work proposed a novel method of pulse signal processing and lung cancer recognition using iterative sliding window(ISW)and classification model of Gaussian support vector machine,supported by project of the National Natural Science Foundation of China(No.61572231),the project of the 11 th session of the China-Slovenia Scientific and technological Cooperation program(No.11-3),and the Shandong Provincial key research & development program(No.2017GGX10141).First of all,an ISW algorithm is proposed,which improves the method of removing baseline wander of the pulse signal.Moreover the continuous pulse signal carrying pathological characteristics is accurately divided into the single periods.The ISW could greatly improved the accuracy of recognizing the valley of the single period,and could solve the problem in the existing literature of that of obvious new distortion introduced by removing baseline wander removal.Secondly,in this study,new wrist pulse signal features are extracted having sight on Jin's pulse diagnosis.The invalid signal is recognized by cubic support vector machine.Finally,principal component analysis and Gaussian support vector machine are utilized to select features and identify lung cancer patients' wrist pulse signal.Other seven classifiers are applied in the experiment to further explore the best signal acquisition position using the wrist pulse signal collected at three different positions of radial artery.The result demonstrate that Gaussian support vector achieves the highest precision of 96.15% of recognizing lung cancer patients' signal,collected at the Cun,a position at radial artery.