CIS4517 Research and Development Project Proposal 2026 | Edge Hill University

CIS4517 Research and Development Project

Aim

To develop a reliable method for the early detection of symptoms associated with diseases such as cancer and diabetes, with the goal of enabling timely and effective treatment for patients.

Objectives

• To develop predictive models that accurately identify symptom patterns of cancer and diabetes based on genetic factors, medical history, and lifestyle factors.
• To validate these predictive models through rigorous testing using diverse and representative datasets.
• To refine the predictive models by analyzing the impact of genetic factors, medical history, and lifestyle on the accuracy of symptom prediction.

8. Focused review of relevant literature, public datasets and open source packages and rationale for study (at the end of this section, list 8-10 key references)

Predictive analytics to diagnose the disease with the help of machine learning (ML) and artificial intelligence (AI) in order to identify different patterns in medical data helps us detecting the diseases early and accurate. This technique is very much good for some dangerous or non-curable diseases at the later stages such as cancer and diabetes, where early diagnosis can help in improving the expected outcomes of the patient. This literature review highlights some literature points by observing the public datasets, and some other open-source surveys that helps in predicting analytics to diagnose the disease.

Literature review

Machine Learning in Healthcare

According to Topol, E. J. (2019), he has discussed about the deep medicines and how Artificial Intelligence can make the healthcare human again. He has also discussed in his writing about the transformative potential of AI in the domain of healthcare, and also enhancing the role of predictive analytics in the very early disease diagnosis.

According to Esteva et al., (2017) discussed about the Dermatologist-level classification of skin cancer with deep neural networks. This study helps in demonstrating the capability of models of the deep learning in order to classify the skin cancer with accuracy comparable related to dermatologists.

Specific Disease Prediction

According to Rajkomar et al., (2018), they have discussed about the Scalable as well as accurate deep learning within the electronic health records. This research helps in highlighting the applications and uses of deep learning within the electronic health records (EHRs) for predicting the patient outcomes; it also helps in showcasing scalability and accuracy.

As per, Choi et al., (2016) discussed about the Doctor AI; they have also predicted the models by predicting clinical events through recurrent neural networks. They have also discussed about the Machine Learning for Healthcare Conference. This paper also discussed about various Doctor AI which is built by using a model using recurrent neural networks (RNNs) to predict the clinical events that helps in using the temporal data for appropriate as well as accurate predictions.

Methodological Advances

According to Miotto et al., (2017) discussed about the Deep learning for healthcare they have discussed about the opportunities and challenges. This paper review and it also provides an overview of deep learning applications with in the healthcare also discussed the challenges as well as potential solutions.

As per, Shickel et al., (2018), this paper discussed about the Deep HER which a survey on the latest and different advances on deep learning techniques for electronic health record-based clinical prediction. This paper also highlights recent advances in deep learning techniques for EHR-based predictions, providing insights into various architectures and their performance.

Diabetes Prediction

This is about type 2 Diabetes, as per Meng et al. (2016) employed logistic regression as well as neural networks on the basis of electronic health records (EHR) in order to predict the onset of type 2 diabetes. The models shown in the paper showed that early prediction is very much on the basis high accuracy, and if the models are trained and tested properly with high accuracy and it is possible to detect type 2 diabetes very easily. 

Rationale for Study:

1. Early Detection and Improved Outcomes: Early diagnosing of different diseases such as cancer, diabetes, and cardiovascular conditions can help in improving the outcomes of patients significantly by giving then proper and on a timely manner treatment. It also helps in creating the models that can help in checking the high risk for individuals for the clinical symptoms and also help in reducing the healthcare cost.
2. Integration of Diverse Data Sources: there is a growing availability of the health records which are electronic so it is good to provide the information on the basis of this rich data and predict the models on the basis of real and available data.
3. Addressing Healthcare Challenges: it also helps in addressing the healthcare challenge like cost, volume of patients are also increasing, and there is a huge shortage of healthcare professionals.
4. Technological Advancements: technology is changing every day, there are various algorithms that are coming and training the dataset and also every time new technology came, it is better than the previous one, also resolving previous technology limitations. So this, way technology can be used in order to diagnose the real problem and issues related to human beings. 

References: 

Meng, X., Zhang, Y., Li, Z., Cheng, L., & Ji, Y. (2016). Predicting the onset of type 2 diabetes mellitus using logistic regression and neural network models. Journal of Biomedical Informatics, 60, 1-7. doi:10.1016/j.jbi.2016.01.004 .

Topol, E. J. (2019). High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine, 25(1), 44-56. doi:10.1038/s41591-018-0300-7 .

Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115-118. doi:10.1038/nature21056 .

Rajkomar, A., Oren, E., Chen, K., Dai, A. M., Hajaj, N., Hardt, M., … & Dean, J. (2018). Scalable and accurate deep learning with electronic health records. npj Digital Medicine, 1(1), 18. doi:10.1038/s41746-018-0029-1 .

9. Detail study design and methods and justify them if possible

• Data collection 
• Data pre-processing 
• Feature selection 
• Model development with the help of different learning models like logistic regression, random forest, and various other algorithms such as CNN, RNN & others. 
• Model evaluation 
• Validation 

10. Detail a proposed time schedule for the project, with key dates and the milestones of each phase of the project  

Total Duration: 4 months

Month 1: Project Planning and Initial Setup

• Finalize project proposal and objectives.
• Gather and review relevant literature.

Milestone: Project proposal approval and literature review completion.

Month 2: Data Collection and Pre-processing

Feature Selection and Model Development

• Identify and acquire public datasets (Breast Cancer Wisconsin, Pima Indians Diabetes, MIMIC-III).
• Perform initial data cleaning and pre-processing.
• Handle missing values, normalize data, and conduct exploratory data analysis.
• Document data pre-processing steps.
• Apply feature selection techniques (RFE, PCA).
• Finalize feature set for modelling.
• Develop baseline machine learning models (logistic regression, decision trees).
• Begin developing advanced models (random forests, neural networks).

Milestone: Completion of data collection and pre-processing and Feature selection and initial model development.

Month 3: Model Training and Evaluation

Model Validation and Sensitivity Analysis

• Train models using cross-validation.
• Evaluate models using performance metrics (accuracy, precision, recall, F1-score, AUC-ROC).
• Refine and optimize models based on evaluation results.
• Document model development and evaluation process.
• Validate the final model on an independent test dataset.
• Conduct sensitivity and robustness analysis.
• Iterate on model improvements based on validation results.
• Prepare detailed report on model validation and sensitivity analysis
  Milestone: Completion of model training and initial evaluation and Final model validation and robustness testing

Month 4: Integration and Finalization

• Integrate the predictive model into a user-friendly application or interface (if applicable).
• Conduct usability testing and gather feedback.
• Make final adjustments based on user feedback.
• Milestone: Integration and usability testing completion.

Month 4: Documentation and Presentation

• Compile all documentation, including methodology, results, and conclusions.
• Prepare final project report and presentation.
• Present findings to supervisors, peers, and stakeholders.
• Milestone: Submission of final report and project presentation.