Page 53 - AIH-1-1

Artificial Intelligence in Health AI model for cardiovascular disease prediction

number of major vessels colored by fluoroscopy, and defect values of the original image, respectively. The normalization
type (Thal) (Table 2). The workflow of the machine learning considered the input values for each attribute obtained
model for vascular disease prediction and classification is using the min-max normalization technique.
given in Figure 3.
The normalization ensures that the values stay in the
However, machine learning is a method of analyzing data range of 0 – 1 to prevent a surge to higher values. Min-
samples and drawing main conclusions using mathematical max normalization technique was used because of the
and statistical approaches, which allows machines to learn NN activation function with a recommended range of
without having to be programmed. The collected data 0.1 – 0.9 to avoid saturation. The data pre-processed was
were pre-processed using the data imputation of the KNN then divided into two: a training set of 70% and a testing
method to fill in the missing values and data normalization set of 30% ratio. The training allows the NN to develop
before the classification process. The attribute average a relationship between the input data and the target. The
(mean) for a column with a missing value was calculated testing set was used to test how efficiently and accurately
and this was used to fill in the missing values. This technique the NN can predict the target. The number of attributes
is direct and simple to implement using Equation II. was reduced from 14, 10, and 8, and also, GA was used for

X −1 = x − min( )x (II) data selection. Furthermore, the number of datasets varied
from 2000, 1500, 1000, and 500, and their performance
max ( ) − min( )x x was evaluated.
where x is the normalized size, x is the original value, A multilayer feed-forward back propagation NN was
1
and min(x) and max(x) are the minimum and maximum used with 14, 10, and 8 input layers, various hidden layers,

Table 2. CVD dataset attributes
Key attribute Patient attribute ID
Age Year
Sex Value 1: Male; value 0: Female
Chest pain type Value 1: Typical type 1 angina; value 2: Typical type angina;
value 3: Non-angina pain; value 4: Asymptomatic
Fasting blood sugar Value 1: >120 mg/dl; value 0: <120 mg/dl
Resting electrographic results (Restecg) Value 0: Normal; value 1: Having ST-T wave abnormality;
value 2: Showing probable or definite left ventricular hypertrophy
Exercise-induced angina (Exang) Value 1: Yes; value 0: No
Slope of the peak exercise ST segment (Slope) Value 1: Unsloping; value 2: Flat; value 3: Down-sloping
Number of major vessels colored by fluoroscopy (CA) Value 0–3
Defect type (Thal) Value 3: Normal; value 6: Fixed defect; value 7: Reversible defect
Resting blood pressure The blood pressure of patients admitted to the hospital was measured in mmHg
Serum cholesterol Measured in mg/dl
Thalach Maximum heart rate achieved
Oldpeak – ST ST depression induced by exercise
Abbreviation: CVD: Cardiovascular disease.

Figure 3. The overview of the machine learning model for vascular disease prediction.

Volume 1 Issue 1 (2024) 47 https://doi.org/10.36922/aih.1746