Literature Review
MAALRAWAILI
literatureviewCardiovasculardiseases.docx
Student Name: MANEAF AFET D ALRAWAIL
University ID: 421103847
Literature Review of Machine Learning Application in Cardiovascular Disease
Introduction
In the last ten years, the world has witnessed an increase in the number of people suffering from cardiovascular diseases. Fortunately, the world has also experienced a surge in technological developments. Technological developments have made life easier. One of the sectors that have greatly benefitted from the surge in technological development is health. Due to the advancement of technology, health delivery has improved; through computerized health systems physicians can now detect diseases easily and faster compared to the past. Technological developments in the last ten years have seen the development and adoption of artificial intelligence and consequentially machine learning has become possible. Since artificial intelligence is still a new technology, it has not been adopted fully in the health sector. However, some physicians believe artificial intelligence with special emphasis being placed on machine learning should be adopted in the fight against cardiovascular diseases. The above statement sets the foundation of this paper’s thesis statement; machine learning should be adopted in the diagnosis and treatment of cardiovascular diseases.
Cardiovascular diseases
Cardiovascular diseases refer to diseases that affect the function and or the structure of the heart. Some of the common cardiovascular diseases are arrhythmias, Marfan syndrome, coronary heart diseases, heart attack, pericardial disease, deep vein thrombosis, and stroke. Cardiovascular diseases are the main cause of death in the United States.
In the diagnosis of cardiovascular diseases, there are eight main techniques and methods used. The first method is through a blood test and chest x-ray. The second most common diagnosis method is the use of the electrocardiogram, a machine that records the electrical signals of the heart making it possible for physicians to identify irregularities in a heart’s rhythm and structure. The third most common cardiovascular disease diagnosis technique is the use of Holter monitoring. A Holter monitor is a portable device used in the detection of heart rhythm irregularities more so those that are not identified during an electrocardiogram exam. Another common diagnosis method is the use of an echocardiogram which is the ultrasound of one’s chest to capture the structure and the function of a person’s heart.
The fifth most common cardiovascular disease diagnosis technique is the stress test. The test involves having individuals raise their hearts by either medicine or exercise while heart tests and imaging are being conducted. The sixth diagnosis method is the use of a cardiac cathetization which involves the insertion of a sheath into a vein or artery to aid doctors in x-ray imaging. Another common diagnosis method is the cardiac computerized tomography scan which involves an x-ray machine rotating around a person capturing the chest and heart images of a person. The last common cardiovascular diagnosis method is through the use of the cardiac magnetic resonance imaging machine which helps in the production of heart and chest pictures through the use of a magnetic field and as a result, a physician can evaluate a patient’s heart.
Cardiovascular diseases can be treated in three main ways. Top on the list is through lifestyle changes. Patients identified with cardiovascular diseases can adopt healthy lifestyles such as the consumption of low sodium and low-fat diets, participating in regular exercises, and refraining from smoking and alcohol consumption. The second treatment method for cardiovascular diseases is the use of medications. Depending on one’s disease, physicians can prescribe appropriate medication. The last treatment method for cardiovascular diseases is through surgery or medical procedures. Depending on the disease and the extent of damage to one’s heart, a doctor can recommend specific surgeries. In most cases, heart surgeries are conducted when medications are not enough, and when one’s condition is severe.
Problem statement
In some cases, doctors can accurately predict a person’s cardiovascular state by simply looking at their test results. In some cases, it can be challenging to accurately predict however, through computerized systems predictions can be made faster and more accurate due to the elimination of biases and human error. Physicians can save the lives of those with cardiovascular disease through the use of artificial intelligence and machine learning. Cardiovascular patients can get an early diagnosis and consequentially get treated early before their health situations deteriorate.
Unfortunately, some people are against the use of artificial intelligence more so machine learning in the diagnosis and treatment of cardiovascular patients. Those against the technique cite that computers cannot be trusted to make accurate health diagnoses and decisions without human intervention. Opponents of machine learning in cardiovascular disease treatment claim that computerized systems are bound to make errors that can be fatal in the diagnosis of patients and this can hurt the treatment of patients.
Literature Review
To gain a deeper understanding of machine learning and the treatment of cardiovascular diseases a literature review will be conducted. The review will be on ten scholarly articles published in the last five years. The articles will be accessed from four electronic databases: google scholar, PubMed, EBSCOhost, and ScienceDirect. The main aim of the literature review will be to understand the general view of the use of machine learning in handling the cardiovascular disease. Also, the review will help reveal whether it is advisable for machine learning to be adopted in health practices more so for the diagnosis and treatment of patients suffering from cardiovascular diseases.
Clinical applications of machine learning in cardiovascular disease and its relevance to cardiac imaging
The first article under review is an article published in 2018 by Subhi J Al’Aref and twenty-one other researchers. The researchers were keen on discovering the relevance of machine learning in the treatment of cardiovascular diseases. The researchers in their publication start by stating that machine learning which is a part of artificial intelligence has transformed the key aspects of human life. According to Al’Aref et al., (2020), through artificial intelligence computerized systems can acquire information through the extraction of patterns from large databases. The researchers identify that machine learning is increasingly being adopted in the medical community despite there not being a guiding framework for its use. Furthermore, the researchers identify that one of the main areas in the medical sector that machine learning has been adopted is in the domain of cardiovascular diseases.
The researchers studied four main domains of machine learning applications in cardiovascular treatment. The first two domains studied were non-invasive imaging techniques such as coronary computed tomography angiography and coronary artery calcium scoring. The other domains highlighted were electrocardiography and echocardiography. The researchers also reviewed the limitations associated with the use of a machine learning algorithm in the cardiovascular field.
The researchers identified that the diagnosis and treatment of cardiovascular diseases are possible as there are clear indicators of coronary diseases. Through identifying patterns of cardiovascular diseases such as stroke and heart attack and through the availing of lifestyle data, computerized systems can identify the extent of a coronary illness. It is also possible to predict when the disease will worsen if all associated variables remain constant. Consequentially, the researchers identified that the clinical application of machine learning in the handling of coronary diseases is relevant more so to cardiac imaging.
As far as the limitations of machine learning in its clinical application is concerned, Al’Aref and his team of researchers identified that for effective diagnosis and prediction, computerized systems need to have access to many medical databases. Furthermore, the researchers found that machine learning was difficult due to the different variables influencing cardiovascular diseases. The researchers concluded that human intervention is still necessary for the diagnosis and treatment of cardiovascular diseases as much as machine learning is possible. According to Al’Aref et al., (2020), medical practitioners end up doing double work when machine learning is in effect. They have to feed data into the systems and also go through the data and predictions to ensure that there are no errors.
The study ‘s publication is ideal as it focuses on the application of machine learning in the medical sector more so in the domains of cardiovascular diseases. The publication not only shares the advantages of machine learning in health care but it also highlights the limitations of the technology. An analysis of the article reveals that machine learning is an effective technology more so in the diagnosis of simple cardiovascular diseases that do not have many variables influencing them. Furthermore, the technology is not yet advanced enough to run on its own without human intervention. Based on the article, there is a need for more refining of artificial intelligence more so machine learning for more adoption in the health sector.
Artificial Intelligence, Machine Learning, and Cardiovascular Disease
The second article under study is an article published in 2020 by Pankaj Mathur and three other researchers. The researchers were interested in knowing how effective artificial intelligence is in the diagnosis and management of cardiovascular diseases. The four researchers start their publication by mentioning that artificial intelligence-based applications have been adopted in many fields of technology, science, and medicine. They also mention that the study of machines in the health sector has been on-going since the 1960s. The biggest and most recent development has been the discovery of algorithms that aid machines learn and mimic the functions of the human brain.
According to Mathuri et al., (2020), artificial intelligence-based systems have found their use in the imaging of the cardiovascular system, the prediction of cardiovascular risks, and in the identification of newer drug targets. According to the article, machine learning has enhanced the human understanding of congenital heart diseases and the various phenotypes of heart failure. Based on the study there are three main benefits of the use of machine learning and artificial intelligence in cardiovascular diseases. First, machine learning applications have led to the establishment of newer and more reliable treatment strategies for cardiovascular diseases. Secondly, the applications have led to the adoption of newer drug therapies for cardiovascular diseases. Lastly through machine learning applications new and better post-marketing surveys of prescription drugs for cardiovascular drugs have been conducted.
The researchers also mentioned the challenges of the use of artificial intelligence in the clinical treatment of cardiovascular diseases. Four challenges were identified with the first being data privacy. According to the article, data privacy is not guaranteed once computerized systems are introduced to hospitals as patient confidential data is at risk of being leaked. Secondly, the computerized systems at times use outdated and poorly selected data in developing patterns and trends. Consequentially, the diagnosis and the predictions made can at times be misleading. Thirdly, there is a risk of selection bias. For machine learning to happen, computers are granted access to specific databases to extract information. The selection of the databases for use is not based on a criterion rather it is based on a clinic’s preferences. Consequentially, there are high risks of bias in the diagnosis and the prediction of cardiovascular trends. Lastly, there is the challenge of the unintentional continuance of stereotypes and historical biases in the data in use in the machines, and this at times leads to erroneous conclusions.
The reviewed article is appropriate as it focuses on artificial intelligence in the diagnosis and management of cardiovascular diseases. An analysis of the article’s findings reveals that artificial intelligence more so machine learning is a transformative intelligence that has immense potential to revolutionize the delivery of healthcare. The analysis further reveals that at the moment the risks associated with the technology outweigh the benefits of the technology. There is a need for data security to be enforced and for a guiding framework on the access of medical databases to be established.
Can machine-learning improve cardiovascular risk prediction using routine clinical data?
The third article under review is an article published in 2017 by Stephen Weng and four other researchers. The publication was about a study carried out earlier. The researchers believed that the approaches in existence are not effective at predicting cardiovascular risks and as a result individual that would benefit from early treatment fail to get the treatment. Furthermore, some individuals receive unnecessary interventions. The researchers wanted to confirm whether indeed machine learning offers the health sector opportunities to improve accuracy. The researchers assessed whether machine-learning can improve the prediction of cardiovascular risks.
Weng and his research team conducted a prospective cohort study. They used the clinical data of over 350,000 patients from the United Kingdom. The researchers compared four machine learning algorithms to the established predictions of cardiovascular events by the American College of Cardiology Guidelines. The four machine learning algorithms compared were logistic regression, neural networks, random forest, and gradient boosting machines. According to Weng et al., (2020), machine learning algorithms significantly improved the prediction of cardiovascular diseases. Due to the improved cardiovascular risk predictions, the researchers concluded that machine learning can help increase the number of individuals in need of preventive treatment while avoiding the unnecessary treatment of other individuals.
The article was appropriate for review because it is less than five years and because it touches on how machine learning is used in the handling of cardiovascular conditions. An analysis of the study’s findings reveals that machine learning offers an improvement in the diagnosis of cardiovascular diseases. The technology makes it possible for a more accurate risk prediction of cardiovascular diseases to be made. Based on the study, as much as technology has its challenges, its benefits outweigh the cons and challenges.
Machine Learning in Predicting Coronary Heart Disease and Cardiovascular Disease Events: Results from The Multi-Ethnic Study of Atherosclerosis (MESA)
The fourth article under review is an article published in 2018 by Rine Nakanishi and nine other researchers. The publication was about a quantitative study that they conducted to establish whether machine learning will lead to improved prediction of coronary heart diseases compared to standard coronary artery calcium and clinical risk assessments of coronary heart diseases. The researchers collected data from over 6000 asymptomatic patients undergoing coronary artery calcium scanning. The machine learning systems in place utilized all available data to predict the likelihood of the patient’s health deteriorating. According to Nakanishi et al., (2018). Machine learning was found to be superior to clinical risk factor assessment.
The article is appropriate as it focuses on machine learning and cardiovascular diseases. An analysis of the article reveals that the algorithm in machine learning systems speeds up the analysis of symptoms and signs of cardiovascular diseases. Due to the algorithm, machines can make better and more accurate predictions compared to humans. The article was good however it had one main limitation; it did not touch on the challenges and risks of adopting machine learning in the risk prediction of cardiovascular diseases.
Machine learning techniques for classification of diabetes and cardiovascular diseases
The fifth article under review is a study published by three researchers; Alic, Gurbeta, and Badnjevic. The researchers published their article in 2017 following a study that they had conducted earlier. The researchers conducted a comparative analysis of papers that were identified. The papers were published between 2008 and 2017. The researchers were interested in knowing how machine learning techniques help in the classification of cardiovascular diseases and diabetes. According to Alic Gurbeta and Badnjevic (2017), the most commonly used artificial neural network technique is the use of the levendberg-Marquardt learning algorithm. Whereas the most commonly used Bayesian networks were the Naïve Bayesian network.
The article was appropriate as it touched on how machine learning techniques make it possible for physicians and health practitioners to classify diabetes and cardiovascular diseases. An analysis of the study and its findings reveal that machine learning is more effective at classifying cardiovascular diseases and diabetes compared to humans. The algorithms dictating machine learning are effective and lead to fast diagnosis and consequentially fast classification.
Moving beyond regression techniques in cardiovascular risk prediction: applying machine learning to address analytic challenges
The sixth article under review is an article published in the European Heart Journal in June 2016. The article was published by three researchers: Goldstein, Navar, and Carter. The research focused on finding ways in which medical institutions can move from using regression techniques in the risk prediction of cardiovascular diseases and instead adopt the use of machine learning. According to Goldstein, Navar, and Carter (2016), risk prediction is important in cardiology research. The majority of risk models used in clinical cardiology are based on regression models. Unfortunately, the models use a small number of predictors as much as there robust and useful. The researchers' primary aim of the study was to find findings that would help push their agenda of having as many clinics and hospitals adopt the use of machine learning technology.
The researchers conducted a review of the data that they derived from their institution’s electronic health records. Also, the researchers walked through the difficulties and the challenges that are in the modeling of health records. Afterward, the researchers introduced machine learning. Consequentially, the researchers realized that machine learning helped to solve challenges not addressed by typical regression approaches. An analysis of the article, reveals that machine learning improves the capability of physicians. The technology offers a faster and more accurate way in which physicians can identify individuals at risk of cardiovascular diseases, diagnose the diseases, and future predictions of the diseases. Based on the article findings, health practitioners must adopt machine learning to improve healthcare delivery.
Machine Learning Strategy for Gut Microbiome-Based Diagnostic Screening of Cardiovascular Disease
The seventh article under review is an article published in September 2020 by Aryal and four other researchers. The article is appropriate as it details how machine learning helps in the fast screening of cardiovascular diseases. The researchers start their publication by highlighting that cardiovascular diseases are the number of cause of death. They also identify that other than environmental and genetic factors, the gut microbiota is a new factor influencing cardiovascular diseases. The main aim of the study was to establish whether machine learning was effective at the diagnostic screening of cardiovascular diseases through the identification of gut microbiota (Aryal et al., 2020).
The researchers analyzed the data of 478 cardiovascular disease patients and 473 non-cardiovascular disease human subjects. They used five supervised algorithms including decision tree, support vector machine, random forest, elastic net, and neural networks. The researchers identified that machine learning made it possible for the early diagnosis of cardiovascular diseases to happen. Based on the analysis of the article, it is possible to claim that machine learning offers faster diagnosis as compared to other diagnosis models. Through fast diagnosis, health practitioners can offer effective interventions on time to curb the worsening of cardiovascular diseases.
Physician-Friendly Machine Learning: A Case Study with Cardiovascular Disease Risk Prediction
The eighth article under review was published in 2019 in an attempt to convince as many people as possible that machine learning is critical in the fight against cardiovascular diseases. The article was about a study conducted by Meghana Padmanabhan and three other researchers. According to Padmanabhan et al., (2019), machine learning is often considered as a highly sophisticated technology that should only be used by trained experts. Consequentially, many physicians and healthcare providers are hesitant of adopting the technology and as a result, keep on using outdated techniques. The study focused on eliminating outdated perceptions.
To make it possible for the researchers to push their agenda their compared auto-machine learning techniques with a graduate student using several diagnostic metrics of cardiovascular diseases. According to the results of the study, automatic machine learning takes an hour to produce accurate diagnostic classifiers for cardiovascular diseases. On the other hand, it takes the graduate student a month to come up with the same qualifiers. The study findings confirm that machine learning is more effective than human interventions are in the fight against cardiovascular diseases. Machine learning makes it easy to conduct a diagnosis of cardiovascular diseases. Furthermore, it gives more accurate classifiers. Based on the analysis of the article, there is an immediate need for health care providers to adopt the use of machine learning technology to ease healthcare delivery.
Cardiovascular disease risk prediction using automated machine learning: A prospective study of 423,604 UK Biobank participants
The ninth article under review was published in 2019 and was a result of a joint effort by Ahmed Alaa and four other researchers. The publication was about a study conducted to test whether machine learning techniques could improve cardiovascular risk prediction in comparison to traditional approaches and whether non-traditional variables had the potential to increase the accuracy of cardiovascular risk predictions. According to the researchers, cardiovascular risk prediction models in place are based on some limited traditional predictors.
To conduct the test, the researchers used data from over 400,000 people without cardiovascular diseases at baseline in the United Kingdom biobank. The researchers developed a machine learning model for predicting cardiovascular disease risk based on 473 variables. According to Alaa et al., (2019), the machine learning-based model used was derived using auto prognosis. The researchers compared the machine learning model with a well-established risk prediction algorithm pegged on the Framingham score. The findings of the study revealed that the machine learning auto prognosis improved risk prediction significantly compared to the Framingham score. The auto prognosis was able to predict over 300 cases more in comparison to the Framingham score.
The article was appropriate because it fit the description for the literature review; it focused on machine learning and cardiovascular diseases and it also was less than five years since it was published. An analysis of the article confirms that machine learning is significantly superior to other metrics of diagnosing and classifying cardiovascular diseases. Based on the findings of the study, healthcare givers need to adopt the use of machine learning not only for handling cardiovascular diseases but also in the handling of other health complications to improve health outcomes.
Effective Heart Disease Prediction Using Hybrid Machine Learning Techniques
The tenth article under review is an article published in 2019 by three researchers; Mohan, Thirumalai, and Srivastava. The three researchers start by mentioning that cardiovascular diseases are one of the leading causes of mortality in the world. The researchers' highlight that the prediction of cardiovascular diseases is one of the most critical challenges in clinical data analysis. According to Mohan. Thirumalai and Srivastava (2019), machine learning is effective in assisting healthcare decision making. To prove their claims, the researchers proposed comparing machine learning in heart decision prediction with traditional heart disease prediction models.
The study findings revealed that machine learning produced an accurate prediction of 88.7%. The machine learning algorithm used was the hybrid random forest. An analysis of the article revealed that it was accurate as it touched on machine learning and heart diseases. Furthermore, the findings of the article support the need for wide-scale adoption of machine learning in the fight against cardiovascular diseases. Through the adoption, it will be possible for early predictions to be made and also for early diagnosis to be made. Through risk prediction and early diagnosis, it will be possible for healthcare providers to ensure that as few people possibly lose their lives due to cardiovascular diseases.
Findings and conclusion
An analysis of the ten articles under review has two main findings. Top on the list is that machine learning is effective at the prediction of cardiovascular diseases and also in the diagnosis of cardiovascular diseases. The algorithms implemented in machine learning ensure that technology is fast and more effective at diagnosing and classifying cardiovascular diseases. Through early detection and diagnosis of cardiovascular diseases, early and effective treatment can be issued leading to the saving of lives. The second finding is that the benefits of machine learning outweigh the challenges and the disadvantages of the technology. As much as machine learning has many advantages, it also has many limitations. Some of the limitations are that it is prone to bias and data privacy is at risk. To make machine learning more effective than it is, measures to reduce the challenges associated with it should be done away with.
Based on the review, there is a need for hospitals and healthcare providers to adopt the use of machine learning. On the other hand, there is a need for health sector regulators and privacy regulators to develop a guiding framework on the use of machine learning in not only the fight against cardiovascular diseases but also in the fight against other diseases. Through the adoption of artificial intelligence more so machine learning, it will be possible to improve health outcomes
References
Al’Aref, S. J., Anchouche, K., Singh, G., Slomka, P. J., Kolli, K. K., Kumar, A., ... & Berman, D. S. (2019). Clinical applications of machine learning in cardiovascular disease and its relevance to cardiac imaging. European heart journal, 40(24), 1975-1986. Retrieved from https://doi.org/10.1093/eurheartj/ehy404
Alić, B., Gurbeta, L., & Badnjević, A. (2017, June). Machine learning techniques for the classification of diabetes and cardiovascular diseases. In 2017 6th Mediterranean Conference on Embedded Computing (MECO) (pp. 1-4). IEEE. Retrieved from https://ieeexplore.ieee.org/abstract/document/7977152
Aryal, S., Alimadadi, A., Manandhar, I., Joe, B., & Cheng, X. (2020). Machine Learning Strategy for Gut Microbiome-Based Diagnostic Screening of Cardiovascular Disease. Hypertension, 76(5), 1555-1562. Retrieved from https://doi.org/10.1161/HYPERTENSIONAHA.120.15885
Goldstein, B. A., Navar, A. M., & Carter, R. E. (2017). Moving beyond regression techniques in cardiovascular risk prediction: applying machine learning to address analytic challenges. European heart journal, 38(23), 1805-1814. Retrieved from https://doi.org/10.1093/eurheartj/ehw302
Mathur, P., Srivastava, S., Xu, X., & Mehta, J. L. (2020). Artificial Intelligence, Machine Learning, and Cardiovascular Disease. Clinical Medicine Insights: Cardiology, 14, 1179546820927404. Retrieved from https://doi.org/10.1177/1179546820927404
Mohan, S., Thirumalai, C., & Srivastava, G. (2019). Effective heart disease prediction using hybrid machine learning techniques. IEEE Access, 7, 81542-81554. Retrieved from https://ehjournal.biomedcentral.com/articles/10.1186/s12940-017-0310-9?optIn=false
Nakanishi, R., Dey, D., Commandeur, F., Slomka, P., Betancur, J., Gransar, H., ... & Budoff, M. (2018). Machine learning in predicting coronary heart disease and cardiovascular disease events: results from the multi-ethnic study of atherosclerosis (mesa). Journal of the American College of Cardiology, 71(11S), A1483-A1483. Retrieved from https://www.jacc.org/doi/full/10.1016/S0735-1097%2818%2932024-2
Padmanabhan, M., Yuan, P., Chada, G., & Nguyen, H. V. (2019). Physician-friendly machine learning: A case study with cardiovascular disease risk prediction. Journal of clinical medicine, 8(7), 1050. Retrieved from https://doi.org/10.3390/jcm8071050
Weng, S. F., Reps, J., Kai, J., Garibaldi, J. M., & Qureshi, N. (2017). Can machine-learning improve cardiovascular risk prediction using routine clinical data?. PloS one, 12(4), e0174944. Retrieved from https://doi.org/10.1371/journal.pone.0174944
