Kronik obstrüktif akciğer hastalığı teşhisi için makine öğrenmesi tabanlı yeni bir yöntem geliştirilmesi = Development of a new machine learning-based method for the diagnosis of chronic obstructive pulmonary disease

Abstract

Geçmiş ve Amaç: Kronik Obstrüktif Akciğer Hastalığı KOAH nefes almayı zorlaştıran ve zamanla kötüleşen bir tür akciğer hastalığıdır. Normalde, ciğerlerimizdeki hava yolları ve hava keseleri elastik veya esnektir. Nefes aldığınızda, hava yolları hava keselerine hava getirir. Hava kesecikleri küçük bir balon gibi hava ile dolar. Nefes verdiğimizde hava keseleri şişer ve hava dışarı çıkar. KOAH'ınız varsa, bir veya daha fazla sorun nedeniyle solunum yollarına daha az hava girer ve çıkar. KOAH kalıcı bir hastalık olarak, zamanla tedavi edilmezse tedavisi zorlaşır, bu hastalığın en büyük nedeni sigara, zararlı partiküller, dışardaki çevresel etkenler ve gazlar sebebi sonucu meydana gelmektedir. Özellikle küçük çocuk, engelli veya ilerlemiş seviyedeki hastaların spirometre aracını kullanma zorlukları ve hastaneye erişimindeki zorlukları, teşhis sürecinin kolaylaştırılması ve kısaltılmasını zorunlu kılmaktadır. Bu sorunların önüne geçebilmek, KOAH'ta erken teşhis ve daha kolay takip için KOAH teşhisinde Fotopletismografi (PPG) sinyalinin kullanımının faydalı olacağı değerlendirilmektedir. PPG, deri yüzeyinden vücudun herhangi bir yerinde ölçülebilen bir biyosinyaldir. Kalbin her atışında oluşan PPG sinyal ölçümü kolay bir sinyaldir. Literatürde PPG sinyalinin vücuda ait oldukça fazla bilgi içerdiği bilgisine yer verilmiştir. Bu çalışmada PPG sinyalinin KOAH teşhisini kullanılabilmesi için makine öğrenmesi yöntemiyle KOAH tahmin modelleri geliştirilmiştir. Yöntemler: Bu çalışmada PPG sinyalinin KOAH teşhisinde kullanılabilmesi için bir sistem tasarımı yapılmıştır. Çalışmanın amacı "PPG ile KOAH teşhis edilebilir mi?" ve "Edilebilir ise en az kaç saniyelik sinyal yardımı ile bu işlem yapılabilir?" bunu belirlemektir. Bu amaç doğrultusunda çalışma için 14 bireyden (8 KOAH, 6 Sağlıklı) ortalama 7-8 saatlik PPG kaydı alınmıştır. Alınan kayıtlar 2, 4, 8, 16, 32, 64, 128, 256, 512 ve 1024 saniyelik parçalara ayrılmıştır. Her saniye grubu için işlemler yapılmış ve hangi saniyelik sinyallerle daha performanslı teşhis yapılabildiği tespit edilmeye çalışılmıştır. Hasta bireye ait 8 saatlik kayıtlar 2 saniyelik parçalara ayrıldığında her parçaya hasta etiketi verilmiştir. Aynı işlem sağlıklı birey için yapıldığında tüm parçalara Sağlıklı etiketi verilmiştir. Her bir sinyal grubu öncellikle 0.1-20 Hz sayısal filtreleme yöntemiyle temizlenmiştir. Ardından 25 adet zaman domeninde özellik çıkarılmıştır. En sonunda oluşan veri setleri (2, 4, 8. sn) karar ağaçları makine öğrenmesi yöntemleri ile sınıflandırılmıştır. Çalışmada PPG sinyali gürültüden temizlenmiş ve PPG'nin üç alt frekans bantlarına sahip yeni PPG sinyalleri elde edilmiştir. Dört sinyalden her birinden 25 adet özellik çıkarılmıştır. Toplam 100 adet özellik çıkarılmıştır. Buna ek olarak yaş, kilo ve boy özellik olarak kullanılmıştır. Performansın artırılması amacıyla Fisher özellik seçme algoritması kullanılmıştır. Sonuçlar: Geliştirilen PPG tahmin modelleri tüm bireyler için doğruluk oranı 0.95 performans değerlerine sahiptir. Özellik seçme algoritmasında kullanılar sınıflandırma algoritmalari performans artışına yardımcı olmuştur. Çözüm: Elde edilen sonuçlara göre 2 saniyelik veri grubu ile en yüksek performans değerlerine ulaşılmış olup 0.99 duyarlık, 0.99 özgüllük ve %98.99 doğruluk oranı elde edilmiştir. Elde edilen sonuçlara göre PPG bazlı KOAH tahmin modelleri pratikte kullanılabileceği düşünülmektedir.

Chronic Obstructive Pulmonary Disease (COPD) is a widespread respiratory disorder disease. COPD is a common disease, characterized by airway and / or persistent respiratory symptoms and airway restriction caused by significant exposure to harmful particles or gases, which can be prevented and treated. COPD constitutes a significant portion of chronic respiratory diseases. COPD is one of the most important causes of morbidity and mortality and with each passing day it continues causing increasing substantial economic and social burden. With the expected prolongation of the life expectancy and increased exposure worldwide, the burden of COPD is predicted to increase further. According to the World Health Organization (WHO) it is the fourth leading cause of death worldwide. Approximately 2.9 million people die annually due to COPD worldwide. The main pathological characteristic of chronic obstructive pulmonary disease (COPD) is chronic respiratory obstruction. COPD is a permanent, progressive disease and is caused by harmful particles and gases entering the lungs. Chronic coughing and shortness of breath are the most important symptoms of the disease. Cigarette smoke is one of the main casual factors of these symptoms. The cigarette smoke fills the airways and alveoli with harmful gases and substances, in course of time, these harmful substances cause damage to the structure of the bronchi and alveoli, resulting in the development and progression of COPD. Since there is not sufficient information about the disease, diagnosis and thus the treatment of the disease are delayed. The diagnosis of the disease is made by the specialist doctor according to the report obtained from the spirometer device, which is the standard method in diagnosis. This method can be applied only in hospitals with the help of a technician. However, monitoring the disease after diagnosis is also important for tracking the damages caused by the disease to the body. Therefore, it is vital to have an early and fast diagnosis process. As COPD is a progressive disease with permanent damages, the earlier it is diagnosed and treated, the less damage it causes to the individuals. Same as with any other disease, monitoring of the disease during the treatment is possible with advanced equipment and only in hospitals. This process is very difficult and time consuming. At present, COPD diagnosis can be made with an apparatus called a spirometer. After measurement of Forced vital Capacity (FVC) and Forced expiratory volume in one second; (FEV 1), by evaluating the FEV 1/FV C ratio, the person can be diagnosed by a specialist doctor. If the FEV 1/FVC is < 70% the patient is considered to have COPD. The difficulty of using the spirometer apparatus and the difficulties in accessing the hospital, especially for young children, disabled or patients in advanced stages of illness, necessitates facilitating and shortening the diagnosis process. Due to the disadvantages associated with current methods of diagnosing chronic obstructive pulmonary disease (COPD), it is necessary to develop systems that are easier to use and to follow up with patients. To address these issues and expedite the diagnosis of COPD, the use of photoplethysmography (PPG) signals is being considered. PPG is a biologic signal that can be measured in any location near the heart in the body. Its potential usefulness lies in its ability to facilitate faster diagnosis and easier patient monitoring. Heart signals convey vital information about the body and illness. Therefore, based on the obtained results, it has been evaluated that it can be used in the diagnosis of COPD. In this study, a PPG signal-based COPD diagnosis algorithm is proposed. It is expected that the developed method will create an infrastructure for the production of the cost-effective portable apparatus for the disease diagnosis. In recent years, in the diagnosis of diseases; various researches areas are being carried out on usability of some new and helpful classifier, decision making software and tools. One of these areas is artificial intelligence applications. It is clear that these systems will provide advantages such as assistance in making the diagnosis, shortening the diagnosis time, efficiency and increased productivity which will be beneficial in the medical field. In this study, the intention is to make diagnosis of COPD disease with the machine learning algorithm only by using the PPG signal belonging to a patient. One of the general aims of this study is to make easier the diagnosis of COPD by using the machine learning method, to assist in determination of the COPD diagnosis. In addition, the improvement of parameters such as diagnosis duration, efficiency and time are among the objectives. This study was carried out by using the PPG signal in compliance with the principles of the Global Initiative for Chronic Obstructive Lung Disease (GOLD). Background and Purpose: COPD, which is named with the initials of the words Chronic Obstructive Pulmonary Disease is a main public health issue both globally and in our country, which continues to increase due to poor awareness of the disease and lack of necessary preventive measures. COPD disease is the result of a blockage of the air sacs known as alveoli withinside the lungs; it's miles a persistent sickness that reasons proceedings difficulty in breathing, cough and shortness of breath. COPD is characterised with breathing signs and symptoms and airflow challenge because of anomalies in the airways and alveoli that occurs as the result of significant exposure to the harmful particles and gases. The spirometry test (breath measurement test), which is being used for the diagnosis of the COPD is creating difficulties in reaching the hospitals, especially in patients with disabilities or advanced disease, and in children. With facilitate the diagnostic treat and prevent these problems, it's far evaluated that using Photoplethysmography (PPG) signal in the diagnosis of COPD disease would be beneficial in order to simplify and speed up the diagnosis process and make it more convenient for monitoring. A PPG signal includes numerous components, including volumetric changes in arterial blood that are related to heart activity, fluctuations in venous blood volume that modify the PPG signal, a Direct Current (DC) component that shows the optical properties of the tissues, and modest energy changes in the body. PPG has typically received the usage of a pulse oximeter, which illuminates the pores and skin and measures adjustments in mild absorption. PPG occurring with every heart rate is an easy signal to measure. PPG signal is modeled by machine learning to predict COPD. Methods: In this study, a system design was made so that the PPG signal can be used in the diagnosis of COPD. The aim of the study was "Can COPD be diagnosed with PPG?" and "If possible, this operation can be done with the help of at least how many seconds of signal?" is to determine it. For this purpose, an average of 7-8 hours of PPG recording was taken from 14 individuals (8 COPD, 6 healthy) for the study. The received recordings are divided into 2, 4, 8, 16, 32, 64, 128, 256, 512 and 1024 second segments. Operations have been made for each second group and it has been tried to determine which second signals can be diagnosed with better performance. When the 8-hour recordings of the patient were divided into 2-second pieces, each piece was given a patient label. When the same process is done for a healthy individual, all parts are labeled as Healthy. Each signal group was first cleaned with 0.1-20 Hz digital filtering method. Then, features were extracted in 25 time domains. Finally, the resulting data sets (2, 4, 8. sec) were classified by decision trees using machine learning methods. During the studies, the PPG signal was filtered to remove noise, resulting in a new PPG signal consisting of three (3) low-frequency bands. From each of the four (4) extracted signals, twenty-five (25) features were obtained, resulting in a total of one hundred (100) features. Furthermore, weight, height, and age were also used as characteristics. The Fisher method was employed in the feature selection process to enhance performance. The purpose of using this method is to improve the performance effectiveness of the feature selection process. Results; According to the results of this study, it was concluded that COPD diagnosis can be made based on machine learning with PPG signal and biomedical signal processing techniques. In addition, it was determined how long a PPG record is needed. Just with only 2 second record the accuracy rate of 95.31% was obtained. These results are an indication that COPD diagnosis has practical diagnostic methods. According to the results obtained, the highest performance values were achieved with the data group of 2 seconds, and 0.99 sensitivity, 0.99 specificity and 98.99% accuracy rate were obtained. The enhanced PPG prediction models have demonstrated an impressive accuracy rate of 0.95 for all individuals. The utilization of classification algorithms in the feature selection process has significantly contributed to the improvement in overall performance. As a conclusion, the important parts of the study can be summarized as, (1) easy-to-use, (2) artificial intelligence-based, (3) very low-cost (4) reliable biomedical system that makes decisions based on signal measurement data has been developed. We hope this study will open up new horizons for COPD diagnosis.