Classification of electrocardiograms using digital signal processing and the classifier of the vector support machine
Jazyk: | ruština |
---|---|
Rok vydání: | 2022 |
Předmět: | |
DOI: | 10.18720/spbpu/3/2023/vr/vr23-633 |
Popis: | Тема вÑпÑÑкной квалиÑикаÑионной ÑабоÑÑ: «ÐлаÑÑиÑикаÑÐ¸Ñ ÑлекÑÑокаÑдиогÑамм Ñ Ð¿Ð¾Ð¼Ð¾ÑÑÑ ÑиÑÑовой обÑабоÑки Ñигнала и клаÑÑиÑикаÑоÑа маÑÐ¸Ð½Ñ Ð²ÐµÐºÑоÑной поддеÑжки». Рданной ÑабоÑе иÑÑледÑеÑÑÑ Ð²Ð¾Ð·Ð¼Ð¾Ð¶Ð½Ð¾ÑÑÑ Ð¿ÑÐ¸Ð¼ÐµÐ½ÐµÐ½Ð¸Ñ Ð¼ÐµÑода маÑÐ¸Ð½Ñ Ð¾Ð¿Ð¾ÑнÑÑ Ð²ÐµÐºÑоÑов Ð´Ð»Ñ ÐºÐ»Ð°ÑÑиÑикаÑии аÑиÑмии по ÑлекÑÑокаÑдиогÑаммам. ÐÑновной ÑелÑÑ Ð´Ð°Ð½Ð½Ð¾Ð¹ ÑабоÑÑ ÑвлÑеÑÑÑ Ð¿Ð¾ÑÑÑоение алгоÑиÑма клаÑÑиÑикаÑии аÑиÑмии по ÑлекÑÑокаÑдиогÑамме, в коÑоÑом бÑдÑÑ Ð¸ÑполÑÐ·Ð¾Ð²Ð°Ð½Ñ ÑазлиÑне меÑодики маÑинного обÑÑÐµÐ½Ð¸Ñ Ð´Ð»Ñ ÑÐ¾Ð·Ð´Ð°Ð½Ð¸Ñ Ð¼Ð¾Ð´ÐµÐ»Ð¸ маÑÐ¸Ð½Ñ Ð¾Ð¿Ð¾ÑнÑÑ Ð²ÐµÐºÑоÑов. ÐадаÑами данной ÑабоÑÑ ÑвлÑеÑÑÑ Ð¸Ð·ÑÑение Ñ Ð°ÑакÑеÑиÑÑик ÑлекÑÑокаÑдиогÑамм, изÑÑение ÑазлиÑнÑÑ Ð¼Ð¾Ð´ÐµÐ»ÐµÐ¹ клаÑÑиÑикаÑии, изÑÑение маÑÐ¸Ð½Ñ Ð¾Ð¿Ð¾ÑнÑÑ Ð²ÐµÐºÑоÑов, изÑÑение ÑазлиÑнÑÑ Ð¼ÐµÑодик ÑлÑÑÑÐµÐ½Ð¸Ñ Ð¼Ð¾Ð´ÐµÐ»Ð¸ клаÑÑиÑикаÑии, поÑÑÑоение данной модели и оÑенка ее ÑабоÑÑ. ÐÐ»Ñ ÑоÑмиÑÐ¾Ð²Ð°Ð½Ð¸Ñ Ð¼Ð¾Ð´ÐµÐ»Ð¸ клаÑÑиÑикаÑии аÑиÑмии по ÑлекÑÑокаÑдиогÑамме бÑл вÑбÑан меÑод конвейеÑизаÑии, в коÑоÑом обÑединÑÑÑÑÑ Ð½ÐµÑколÑко ÑÑапов пÑедобÑабоÑки даннÑÑ Ð¸ поÑледним ÑÑапом ÑÑавиÑÑÑ Ð¼Ð¾Ð´ÐµÐ»Ñ Ð¼Ð°ÑÐ¸Ð½Ñ Ð¾Ð¿Ð¾ÑнÑÑ Ð²ÐµÐºÑоÑов. ÐÑедобÑабоÑка даннÑÑ ÑоÑÑÐ¾Ð¸Ñ Ð¸Ð· двÑÑ ÑÑапов. ÐеÑвÑм ÑÑапом ÑвлÑеÑÑÑ Ð¿Ñименение ÑилÑÑÑа СавиÑкого-ÐÐ¾Ð»Ð°Ñ Ð´Ð»Ñ ÑÐ³Ð»Ð°Ð¶Ð¸Ð²Ð°Ð½Ð¸Ñ Ð²Ñ Ð¾Ð´ÑÑего Ñигнала и ÑÐ´Ð°Ð»ÐµÐ½Ð¸Ñ ÑÑмов из Ñигнала. ÐÑоÑÑм ÑÑапом ÑвлÑеÑÑÑ Ð¼Ð°ÑÑÑабиÑование, коÑоÑое позволÑÐµÑ Ð¿ÑивеÑÑи вÑе Ð²Ñ Ð¾Ð´ÑÑие к Ð¾Ð´Ð½Ð¾Ð¼Ñ Ð²Ð¸Ð´Ñ, Ñем ÑамÑм знаÑиÑелÑно ÑпÑоÑÑиÑÑ ÑабоÑÑ Ð¼Ð¾Ð´ÐµÐ»Ð¸. Так же пÑименÑÑÑÑÑ Ð¼ÐµÑÐ¾Ð´Ñ Ð´Ð»Ñ ÑлÑÑÑÐµÐ½Ð¸Ñ Ð¼Ð¾Ð´ÐµÐ»Ð¸, Ñакие как поиÑк по ÑеÑке Ð´Ð»Ñ 5 вÑбоÑа опÑималÑнÑÑ Ð³Ð¸Ð¿ÐµÑпаÑамеÑÑов, а Ñак же меÑод One-Vs-Rest Ð´Ð»Ñ Ð¿ÐµÑÐµÑ Ð¾Ð´Ð° Ð¾Ñ Ð±Ð¸Ð½Ð°Ñной клаÑÑиÑикаÑии к мÑлÑÑиклаÑÑовой. ÐÑогом данной ÑабоÑÑ ÑвлÑеÑÑÑ Ð¾Ð±ÑÑÐµÐ½Ð½Ð°Ñ Ð¼Ð¾Ð´ÐµÐ»Ñ, коÑоÑÐ°Ñ Ð¾Ð±ÑединÑÐµÑ ÑÑÐ°Ð¿Ñ Ð¿ÑепÑоÑеÑÑинга и маÑÐ¸Ð½Ñ Ð¾Ð¿Ð¾ÑнÑÑ Ð²ÐµÐºÑоÑов в единÑÑ ÑÑÑноÑÑÑ. ÐÑенка ÑабоÑÑ Ð´Ð°Ð½Ð½Ð¾Ð¹ модели пÑоводилаÑÑ Ñ Ð¿Ð¾Ð¼Ð¾ÑÑÑ Ð¼ÐµÑода пеÑекÑеÑÑной пÑовеÑки, коÑоÑÑй показал ÑоÑноÑÑÑ Ð¼Ð¾Ð´ÐµÐ»Ð¸ в 94 пÑоÑенÑа. ÐаннÑÑ Ð¼Ð¾Ð´ÐµÐ»Ñ Ð¼Ð¾Ð¶Ð½Ð¾ иÑполÑзоваÑÑ Ð² медиÑинÑкой пÑакÑике Ð´Ð»Ñ Ð¾Ð±Ð»ÐµÐ³ÑÐµÐ½Ð¸Ñ Ð¿Ð¾Ð»ÑÑÐµÐ½Ð¸Ñ Ð´Ð¸Ð°Ð³Ð½Ð¾Ð·Ð° аÑиÑмии, пÑиÑем не пÑоÑÑо Ð´Ð»Ñ Ð¾Ð¿ÑÐµÐ´ÐµÐ»ÐµÐ½Ð¸Ñ ÐµÑÑÑ Ð½Ð°ÑÑÑÐµÐ½Ð¸Ñ ÑиÑма или неÑ, а Ð´Ð»Ñ Ð¾Ð¿ÑÐµÐ´ÐµÐ»ÐµÐ½Ð¸Ñ ÐºÐ¾Ð½ÐºÑеÑного Ñипа аÑиÑмии. Так же даннÑÑ Ð¼Ð¾Ð´ÐµÐ»Ñ Ð¼Ð¾Ð¶Ð½Ð¾ иÑполÑзоваÑÑ ÐºÐ°Ðº один из оÑновнÑÑ ÐºÐ¾Ð¼Ð¿Ð¾Ð½ÐµÐ½Ñов ÐÐ, коÑоÑое можно иÑполÑзоваÑÑ Ð´Ð»Ñ Ð°Ð½Ð°Ð»Ð¸Ð·Ð° болÑÑÐ¸Ñ ÐºÐ°ÑдиогÑамм, ÑÐ°ÐºÐ¸Ñ ÐºÐ°Ðº в пÑибоÑÐ°Ñ ÑÑÑоÑного мониÑоÑиÑованиÑ. Такое ÐÐ Ð¼Ð¾Ð¶ÐµÑ ÑилÑно облегÑиÑÑ Ð¾Ð¿ÑÐµÐ´ÐµÐ»ÐµÐ½Ð¸Ñ Ð¿Ð°Ñологий в ÑабоÑе ÑеÑдÑа и бÑÐ´ÐµÑ Ð¾ÐºÐ°Ð·ÑваÑÑ ÑеÑÑезнÑÑ Ð¿Ð¾Ð¼Ð¾ÑÑ Ð²ÑаÑÑ Ð´Ð»Ñ Ð¿Ð¾ÑÑÐ°Ð²Ð»Ð½Ð¸Ñ Ð´Ð¸Ð°Ð³Ð½Ð¾Ð·Ð°. ÐÑновнÑм вÑводом данной ÑабоÑÑ ÑвлÑеÑÑÑ ÑÑвеÑждение, ÑÑо меÑод маÑÐ¸Ð½Ñ Ð¾Ð¿Ð¾ÑнÑÑ Ð²ÐµÐºÑоÑов, вкÑпе Ñ Ð¸ÑполÑзованием ÑÐ°ÐºÐ¸Ñ Ð¼ÐµÑодов пÑедобÑабоÑки даннÑÑ , как ÑилÑÑÑ Ð¡Ð°Ð²Ð¸Ñкого-ÐÐ¾Ð»Ð°Ñ Ð¸ маÑÑÑабиÑованиÑ, а Ñак же пÑи иÑполÑзовании меÑодов опÑÐµÐ´ÐµÐ»ÐµÐ½Ð¸Ñ Ð¾Ð¿ÑималÑнÑÑ Ð³Ð¸Ð¿ÐµÑпаÑамеÑÑов Ñ Ð¿Ð¾Ð¼Ð¾ÑÑÑ Ð¿Ð¾Ð¸Ñка по ÑеÑке и пеÑÐµÑ Ð¾Ð´ Ð¾Ñ Ð±Ð¸Ð½Ð°Ñной клаÑÑиÑикаÑии к мÑлÑÑиклаÑÑовой Ñ Ð¿Ð¾Ð¼Ð¾ÑÑÑ Ð¼ÐµÑода One-Vs-Rest, показÑÐ²Ð°ÐµÑ Ñ Ð¾ÑоÑие ÑезÑлÑÑаÑÑ Ð´Ð»Ñ Ð·Ð°Ð´Ð°Ñи клаÑÑиÑикаÑии аÑиÑмий по ÑлекÑÑокаÑдиогÑаммам. ТоÑноÑÑÑ Ñакой модели ÑоÑÑоавлÑÐµÑ 94 пÑоÑенÑа. Topic of the final qualification work: "Classification of electrocardiograms using digital signal processing and the classifier of the vector support machine." In this paper, we investigate the possibility of using the support vector machine method for classifying arrhythmias from electrocardiograms. The main goal of this work is to build an algorithm for classifying arrhythmia by electrocardiogram, which will use various machine learning techniques to create a support vector machine model. The objectives of this work are to study the characteristics of electrocardiograms, the study of various classification models, the study of the support vector machine, the study of various methods for improving the classification model, the construction of this model and the evaluation of its work. To form a classification model for arrhythmia according to the electrocardiogram, a pipelinerization method was chosen, in which several stages of data preprocessing are combined, and the support vector machine model is set as the last stage. Data preprocessing consists of two stages. The first step is to apply the Savitsky-Golay filter to smooth the incoming signal and remove noise from the signal. The second stage is scaling, which allows you to bring all inputs to the same view, thereby greatly simplifying the work of the model. Methods are also used to improve the model, such as grid search to select the optimal hyperparameters, as well as the One-Vs-Rest method for the transition from binary classification to multiclass. The result of this work is a trained model that combines the stages of preprocessing and support vector machines into a single entity. The performance of this model was evaluated using the cross-validation method, which showed an accuracy of 94 percent of the model. 7 This model can be used in medical practice to facilitate the diagnosis of arrhythmia, and not just to determine whether there is a rhythm disorder or not, but to determine the specific type of arrhythmia. Also, this model can be used as one of the main components of software that can be used to analyze large cardiograms, such as in 24-hour monitoring devices. Such software can greatly facilitate the determination of pathologies in the work of the heart and will provide serious assistance to the doctor for making a diagnosis. The main conclusion of this work is the assertion that the support vector machine method, coupled with the use of data preprocessing methods such as the Savitsky-Golay filter and scaling, as well as when using methods for determining optimal hyperparameters using grid search and the transition from binary classification to multiclass using the One-Vs-Rest method, shows good results for the task of classifying arrhythmias from electrocardiograms. The accuracy of this model is 94 percent. |
Databáze: | OpenAIRE |
Externí odkaz: |