Diseases like diabetes, heart disease, kidney disease, thyroid disease, and other diseases are increasing in frequency, and people are suffering globally. Specifically, thyroid and heart diseases affect many people and, without proper treatment, become serious health issues. Different thyroid and heart disease disorders can be detected early with specific symptoms. Here, the authors provide a thorough literature review of the different popular approaches for disease classification using specific symptoms for early identification and treatment using machine learning. This chapter also outlines the different advantages and limitations of specific approaches for disease symptom detection. The experimental results in existing literature has shown significant results on eight disease benchmark datasets using three state-of-the-art algorithms, including the reduced error pruning (REP) tree, random tree, and C4.5 decision tree algorithm.
TopIntroduction
To ensure accurate results, it is critical to handle medical data analysis with precision. Medical data analysis centers around finding unknown patterns that can be used for disease detection and treatment; however, traditional approaches lack accuracy and precision in their prediction.
The following are the main steps traditionally followed when applying machine learning in medicine:
- 1.
Digital Knowledge Base:
The patient case sheet, or profile, is collected from the medical institutes, and the data is digitalized. The patient's profile includes all demographic and clinical data. Demographic data includes the patient’s name, age, gender, height, weight, etc. At the same time, clinical data includes diagnoses, symptoms, and medications provided as treatment.
Digitalized medical data is pre-processed for better representation as input to the machine learning pipeline. The data undergoes complexity analysis, and different complexities are identified, such as high dimensionality, normalization, feature space selection, outlier, and missing values analysis. The regularized data is prepared and given as input to machine learning algorithms, computer-aided diagnosis is performed, and an effective treatment strategy is identified.
The results generated by the machine learning analysis are synthesized, medical experts perform a thorough discussion, and finally, clinical decisions are made. The new outcomes are incorporated into the knowledge base.
In supervised learning, a predictive model is built using the training data, and this model is then evaluated on the test data, and, finally, the prediction of unknown instances is made. In unsupervised learning, clustering techniques are applied to the raw data, and a model is built for clustering new cases. One of the most common syndromes worldwide, thyroid disorders and diseases, result from thyroid malfunction (Prerana, 2015), in which the thyroid gland causes lower or higher production of thyroid hormones. The thyroid gland releases triiodothyronine (T3) and thyroxine (T4) into the bloodstream, which are considered principal hormones. The rate of metabolism is regulated by the functions of these thyroid hormones.
In some cases, thyroid hormone secretion is a problem with metabolism control. The most common issues of thyroid disease are hypothyroidism, low or inactive thyroid hormone, and hyperthyroidism, elevated thyroid hormone levels. Early diagnosis is always recommended to prevent the acute effects and improve treatment to maintain the normal thyroid hormone level (Tyagi, 2018).
Autoimmune-mediated thyroid disease is due to immune system dysregulation, genetic predisposition, and environmental factors. Due to variability in disease progression and onset heterogeneity, prognosis and diagnosis are unpredictable. Autoimmunity pre-disposition is highly correlated with genetics and occurs due to a mechanical defect resulting in a loss of self-tolerance. Autoimmune disease prevalence is highly complex, and the diseases are variably represented (Prerana, 2015).
Autoimmune thyroid disease benefits from personalized healthcare for a casual molecular mechanism as an alternative to treating special symptoms. Standard patient care generates multiple clinical data types, including laboratory results and MRI images. Additionally, omics data, like patients’ proteomic, transcriptomic, and genomic profiles, have recently become available. Several omics data types can be rapidly used in machine learning models to identify the exact details of autoimmune thyroid disease. Data mining methods have the ability to identify clinically relevant patterns to estimate autoimmune risks, initial and ongoing management, diagnosis, treatment response, and results and observations (Franco, 2013).
Recently, new technology in data mining classification algorithms developed in medical science allows researchers to aid the expert advisory system in diagnosing a greater variety of diseases with increased accuracy. Due to certain errors during the general diagnosis process, medical professionals are using these newly developed systems based on artificial intelligence. These systems assist doctors in reducing the cost and time needed for effective treatment and diagnosis (Stafford, 2020).