Course Content

Abstract :

The low weight of fetus at birth is considered one   of the most critical problems in pregnancy care, affecting the newborn’s health and leading it to death in more severe cases. This condition is responsible for the high infant mortality rates worldwide. In health, artificial intelligence techniques, especially those based on machine learning (ML), can early predict problems related to the fetus’ health state during entire gestation, including at birth. Hence, this project  proposes an analysis of several ML techniques capable of predicting whether the fetus will born with less weight for its gestational age. The importance of the early diagnosis of problems related to  fetal development relies on the possibility of an increase in the gestation days through timely intervention. Such intervention would allow an improvement in fetal weight at birth, associated with a decrease in neonatal morbidity and mortality.

So in this project we are going to predict the fetal birth weight in early stage also classified them as low birth weigh is weight is less than 2.5kg ,normal birth weight if weight is greater than 2.5kg and less than 4.5kg and abnormal birth weight if weight is greater than 4.5kg. Here we used machine learning techniques and algorithms like Linear Regression and Random Forest Regressor to predict the fetal birth weight and Random Forest Regressor is predicting more accurately than Linear Regression.

 

 

Introduction :

Nowadays, the quantity of newborns with low birth weight has increased. Intrauterine growth restriction (IUGR) is a disorder in which the fetus is smaller in stature compared with others with the same gestational age. This condition limits the growth of the baby’s body and organs. Neonates with IUGR can have problems at birth, such as low oxygen level, low Apgar score, respiratory difficulties due to meconium aspiration or hypoglycemia. Severe cases can cause fetal death or long-term growth problems. Maternal hypertension is the primary cause of IUGR, although there are still significant difficulties in determining the different types of hypertension during the pregnancy-puerperal cycle. In this sense, persistent arterial hypertension is, independently of its etiology, responsible for the most severe fetal growth disorders . The health services area is one of the most benefited by the diffusion of machine learning (ML) approaches. ML belongs to a subfield of artificial intelligenc.e (AI) that uses data collected on a large scale to recognize patterns and trends that would be undetectable by human observers . In this way, it allows a system to incorporate new information and learn from experience, automatically correcting parameters and increasing the ability to recognize patterns and predictions of behavior. The larger the databases, the more powerful is the learning ability of those systems. This paradigm allows AI systems to learn, e.g., thousands of diagnoses and medical guidelines on the most diverse diseases .. In examining their patients, clinicians can insert case characteristics into the system and generate the most probable diagnostic assumptions . In this sense, ML algorithms allow reducing the number of possible diagnoses in more complex cases, producing more precise indications of exams and treatments. Therefore, ML based algorithms represent a powerful tool to assist the health professional in the decision-making process. Regarding the Big Data paradigms, it is possible, through innovative ML techniques, to analyze a significant amount of data from pregnant women and to provide obstetricians/gynecologists with more detailed information on fetal health during and after pregnancy . The early identification of increased risk of complications due to low birth weight can provide guidance and treatment adapted to risk profiles and individual needs. Besides, the health area also tends to adhere firmly to the Internet of Things (IoT) paradigms for monitoring patients through various devices that can transmit data about the clinical condition of the patient that needs constant monitoring. There is also the ease of data storage and remote access through cloud computing . Therefore, constant monitoring of fetal health through information and communication technologies can reduce mortality and morbidity rates, contributing to a better quality of life for both mother and baby.

Fetal weight is an essential factor to predict the short and long-term health consequences . According to birth weight (BW), the neonates are defined by the World Health Organization (WHO) as three groups, namely low birth weight (LBW, BW < 2500g), normal birth weight (NBW, 2500g ≤ BW < 4000g) and high birth weight (HBW, BW ≥ 4000 g) which is also called macrosomia . Low birth weight is connected with fetal and neonatal mortality and inhibited growth, it can also cause long-term diseases in their childhood, such as mental retardation and learning disabilities . Macrosomia can cause perinatal asphyxia and death, moreover, for maternities, the risk of caesarean section, prolonged labor, abnormal haemorrhage, and perineal trauma The associate editor coordinating the review of this manuscript and approving it for publication was Yonghong Peng. increases . In the long term, macrosomia is more likely to be associated with obesity, diabetes, and heart disease . Therefore, it is significant to estimate fetal weight accurately during pregnancy and identify low birth weight fetuses or macrosomia correctly. Once the risk has been identified, the maternal or neonatal morbidity and mortality can be reduced by taking appropriate clinical decisions and precautions.

So in this project we are going to predict the fetal birth weight in early stage also classified them as low birth weigh ,normal birth weight and abnormal birth weight based on predicted birth weight. Here we used machine learning techniques and algorithms like Linear Regression and Random Forest Regressor  to predict the fetal birth weight and Random Forest Regressor is predicting more accurately than Linear Regression.

 

 

Objective:

The main aim of this project is to predict the fetal birth weight in advance with the help of machine learning techniques and algorithms and classify them based on predicted weight whether it is low birth weight, normal birth weight  and abnormal  birth weight.

 

Problem Definition:

Low birth weight is the most influential factor in the determination of neonatal morbidity and mortality. According to the World Health Organization (WHO), newborns having a birth weight below 2,500g are considered small for gestational age (SGA) ) care. This cut-off point, adopted for international comparison, is based on epidemiological observations that newborns are weighing less than 2,500g present 20 times more chances to die than newborns with more weight. Birth weight is often used as a proxy for fetal weight. Problems with this practice have recently been brought to light. We explore whether data available at birth can be used to predict estimated fetal weight using several machine learning algorithms.

 

 

 

Proposed System:

 

We proposed a system to predict fetal birth weight using machine learning techniques and algorithms like linear regression and random forest regressor on based on some parameters and classify them based predicted birth weight as low, normal and abnormal birth weight.