Punjab Engineering College, Chandigarh
Predicting heart disease has been one of the most challenging jobs in the medical industry in recent decades. Heart disease claims the lives of about one person per minute in the modern era. Processing vast amounts of data in the healthcare industry requires the use of data science. Since predicting cardiac illness is a difficult undertaking, it is necessary to automate the process in order to minimize risks and notify patients well in advance. The Kaggle machine learning repository's dataset on heart disease is used in this paper. The suggested work uses logistic regression to categorize patients' risk level and forecast their likelihood of developing heart disease. As a result, this research provides a thorough analysis of the machine learning algorithm's performance in detecting the risk of a heart attack. The testing results confirm that, when compared to other ML algorithms used, the Logistic Regression approach has the greatest accuracy of 90.81%.
Cardiovascular diseases (CVDs) are the leading cause of death worldwide, responsible for 17.9 million deaths annually, or nearly one-third of all global deaths. Over 85% of these are due to heart attacks and strokes. The global CVD mortality rate stands at 233 cases per 100,000 people, with low- and middle-income countries seeing a rapid increase in cases and deaths. Premature CVD deaths are projected to rise to over 23 million annually by 2030. To combat this, health systems need strengthening, and efforts must focus on prevention, early detection, and treatment through evidence-based strategies. The bright side is that heart attacks are remarkably avoidable, and simple lifestyle changes (including cutting back on alcohol and tobacco use, eating healthfully, and exercising) along with prompt treatment greatly increase the likelihood of survival. Despite this, the multifactorial nature of many contributing risk factors, such as diabetes, hypertension, increased cholesterol, and so on, makes it challenging to identify high-risk patients. Information mining and machine learning take center stage here. Because machine learning (ML) tactics are more common in design recognition and characterization than other traditional factual methodologies, experts and academics have also turned to ML strategies to construct screening devices.
A coronary artery spasm, commonly referred to as a heart attack, typically happens when a blood clot obstructs the heart's blood supply. Tissue dies when it is oxygen-starved and lacks blood. With heart attacks contributing to almost 25% of all fatalities globally in 2023, they are the leading cause of mortality globally. Anxiety, exhaustion, dizziness, tightness or pain in the arms, back, neck, or chest, and an irregular pulse are common signs of a heart attack. Not every person experiencing a respiratory failure will exhibit the same symptoms or level of severity. Some suffer from mild anguish, while others have more severe agony. Some folks have no symptoms at all. For a few, unexpected heart failure may be the main symptom. However, the greater the number of symptoms and indicators you have, the more likely it is that you are experiencing respiratory failure.
While some heart attacks happen suddenly, many people experience warning signs and symptoms hours, days, or even weeks in advance. The earliest warning signs could be weight or sporadic chest pain that is relieved by rest and triggered by activity (angina). A brief reduction in the blood supply to the heart causes angina.
When obstruction occurs in one or more of your coronary veins, you may experience cardiovascular failure. Gradually, the accumulation of fatty deposits, like as cholesterol, forms deposits known as plaques that can narrow the conduits (atherosclerosis). The majority of cardiovascular failures are caused by a disease known as coronary conduit sickness.
A plaque rupture that releases various chemicals and cholesterol into the bloodstream can occur during a myocardial infarction. At the location of the fracture, a blood coagulation forms. If coagulation is too high, it can obstruct blood flow via the coronary artery, depriving the heart of oxygen and nutrients (ischemia).
The coronary artery may be completely or partially blocked.
Depending on the type you've had, there may be differences in determination and therapy.
Blockages resulting from atherosclerosis do not trigger heart attacks in all cases. In the absence of obstructive coronary artery disease, a heart attack caused by other heart and blood vessel disorders is referred to as a myocardial infarction (MINOCA). MINOCA is more prevalent among women, younger individuals, and members of racial and ethnic minorities, such as Asian, Black, and Hispanic/Latino persons. Different conditions can affect the heart differently and lead to MINOCA.
Although little plaques in your arteries may not obstruct your blood vessels, they may fracture or lose part of their outer covering. Blood clots may form on these plaques as a result. Your coronary arteries may then become blocked by the blood clots. Small plaque formation is more prevalent in women, smokers, and those with other blood vessel disorders. Even in the absence of plaque accumulation, a sudden and severe spasm (tightening) of your coronary artery might obstruct blood flow through it. One of the risk factors for a coronary spasm is smoking. If you smoke, you can be more susceptible to spasms brought on by extremely cold temperatures or stressful circumstances. Cocaine and other drugs can also induce myocardial spasm. When a blood clot enters your bloodstream and lodges in your coronary artery, it is known as a coronary artery embolism. This may stop your artery's blood flow. This is particularly likely in those with atrial fibrillation or illnesses like thrombocytopenia or pregnancy that increase the risk of blood clots.
When a tear develops inside your coronary artery, it can lead to spontaneous coronary artery dissection, or SCAD. The ripped tissue may subsequently obstruct your artery or a blood clot may form at the site of the rupture. Pregnancy, high levels of physical activity, and stress can all contribute to SCAD. Marfan syndrome sufferers and women under 50 years old who are pregnant are more likely to experience this illness.
The likelihood of developing coronary artery disease and experiencing a heart attack is increased by specific risk factors.
Factors under control include:
Lifestyle choices include things like smoking, not getting enough exercise, eating a poor diet that includes a lot of foods high in sodium or saturated fat.
Additional medical disorders include high blood pressure, high blood sugar, diabetes, high blood triglycerides, high blood cholesterol, and preeclampsia (high blood pressure during pregnancy), being overweight or obese.
It is referred to as metabolic syndrome if a person possesses three or more of these heart disease-related risk factors. This significantly raises the chance of having a heart attack.
Factors not under control include:
Age: Men are more likely to develop heart disease after 45, and women are more likely to do so after 55 (or after menopause).
Early heart disease in the family: If your mother, sister, or father were diagnosed with coronary artery disease before the age of 65, or if your brother or father was diagnosed with the condition before the age of 55, you are at an increased risk of developing early heart disease viral and bacterial infections
Prevention:
By treating any known coronary artery disease or altering behaviors that increase your risk of heart attack, you can reduce your chance of having one. Heart disease can be prevented by adopting heart-healthy lifestyle practices, such as eating a balanced diet, exercising regularly, giving up smoking, controlling stress, and keeping a healthy weight. These modifications can reduce the risk of a heart attack even in those with pre-existing coronary artery disease.
The damage your heart sustains after a respiratory failure is often linked to complications, which might lead to:
Arrhythmias, or irregular heartbeats: Electrical "short circuits" can result in irregular heartbeats, some of which can be fatal and cause death.
Heart disappointment: You may have so much damage to your heart from a respiratory failure that your heart's remaining muscle is unable to pump enough blood out of your heart. Cardiovascular breakdown can occur suddenly or develop gradually because of widespread, lifelong heart damage.
Abrupt heart failure: An electrically unsettling influence that creates an irregular heartbeat (arrhythmia) causes your heart to quit beating suddenly. Acute heart failure increases the risk of cardiac failure, which can result in death without immediate medical attention.
The leading cause of illness and death worldwide is heart disease:
Here's when data mining methods and machine learning algorithms come in handy.
An approach for supervised learning classification that estimates the likelihood of a target variable is called logistic regression. There would only be two classes conceivable due to the dichotomous character of the dependent or target variable.
An elucidation of the standard logistic function can precede an explanation of logistic regression. Any real number between zero and one can be entered into the logistic function, which is a sigmoid function. It is described as
The following are some major benefits of employing logistic regression:
A confusion matrix, commonly called an error matrix, is a type of standard table structure used to visualize algorithm performance, especially when dealing with statistical classification difficulties.
A supervised learning one is sometimes called a matching matrix in unexpected learning. In the matrix, a hypothetical class is represented by each row, while instances in the real class are represented by each column (or vice versa). The name alludes to how easy it is to ascertain whether the two classes are being confused by the system, or if they are typically confused with one another.
That's a specific type of contingency table where the two dimensions (a variable is a combination of each class and dimension in the dimension table) have the same set of "classes" in both "real" and "approximate" dimensions.
False positive (FP): A test result that falsely indicates the existence of a particular condition or characteristic
Until a call threshold is considered, logistic regression does not become a classification method. Setting the threshold value is an important logistic regression component that is dependent on the classification problem.
The evaluation of "the price | the worth" of the threshold value is adversely affected by the exactness and recall values. While it would be perfect if every exactness and recall were the same, this rarely happens. We take into account the following elements to establish the threshold in the case of a Precision-Recall trade-off: -
Low Precision/High Recall: We choose a call price that has a low exactitude accuracy or a high recall value when we want to reduce the number of false negatives without lowering the number of false positives. For instance, in a cancer diagnostic application, we frequently classify any affected patient as unaffected without considering whether the patient has received a de jure cancer diagnosis. This may be the case since cancer can be detected in other medical problems, but not in a candidate who has already been rejected.
High Precision/Low Recall: We frequently select a call with a high precision value or a low recall value when we wish to lower the number of false positives without also lowering the number of false negatives. If, for example, we are inclined to classify customers according to how they will respond to a made-to-order advertisement, we want to be 100% sure that the customer will respond favorably to the promotion because, should the customer respond unfavorably, the consumer might miss out on a possible sale.
The following steps are involved in predicting the risk of a heart attack:
Split the data into test and trainsets
Using the data, train the logistic regression model to:
Importing: Bring in the necessary libraries and the CSV file. The required libraries are as follows:
Pre-processing and cleaning of data: Pre-processing refers to the modifications made to our data before it is sent into the algorithm. One technique used to transform the raw data into a clean informational index is called data preprocessing.
The actions needed to prepare the data are as follows:
The following are potential obstacles in this step:
RESULTS OF EXPERIMENTS AND DISCUSSION
A Kaggle dataset of 3,700 rows (datapoints) and 13 columns (attributes) was utilized. 165 (54%) of the 1,998 datapoints had Cardio Vascular Disease (CVD), while the remaining 1,702 (46%) did not.
A selection from the first three rows of the data
Exploratory Analysis
The initial stage involves examining the distribution of various attributes, which is best illustrated using histograms.
Plot the graphs between each column (attribute) and the proportion of patients with and without CVD after that.
Examine the percentages of good and negative cases in each category to obtain additional understanding.
It was challenging to draw conclusions because of the unbalanced nature of the data set, however these are the conclusions that could be inferred from the observations:
The distribution of CHD patients' ages revealed another intriguing trend: the number of sick people grew progressively older, peaking at 63 years old.
Checking the correlation between the various features and the objective variable, as well as between them, is the last stage. This will disclose any co-linearity among the features and provide a good assessment of how strong the features are as coronary heart disease predictors.
* Model Results
The model's accuracy is determined to be 91%.
The confusion matrix illustration:
Classification Report:
Other ML algorithms like K-nearest neighbor, Random Forest Classifier, Decision Tree Classifier, Support Vector Machine and XG Boost Machine can also be used to detect the heart attack risk. The dataset was also tested with these algorithms and the performance was compared. The table summarizes and compares the outcomes of different algorithms:
CONCLUSION
Make the classification or prediction based on the test results and show the patient's actual values—which show whether they have CVD—as well as the classification or prediction the Logistic Regression model made. I have seen that the model misdiagnosed certain patients as having cardiovascular disease (CVD) when in fact they did not, and it misclassified as cancer-free those who did have CVD. This approach is good, but I want it to be even better in terms of how it affects the lives of other people. It should be as accurate as feasible and perform on par with physicians, if not better. Therefore, there is still some fine-tuning that needs to be done on each model. There are a lot of possible improvements that may be looked into to make this expectation framework more flexible and accurate.
REFERENCES
Muskan Khepar, Ashish Sharma, A Step Towards Preventive Care: Leveraging Machine Learning for Heart Disease Risk Prediction, Int. J. of Pharm. Sci., 2024, Vol 2, Issue 10, 566-573. https://doi.org/10.5281/zenodo.13920479
10.5281/zenodo.13920479
