VIETNAM NATIONAL UNIVERSITY, HANOI INTERNATIONAL SCHOOL ------------------------------- GRADUATION PROJECT EXPLORING THE APPLICATION OF ARTIFICIAL INTELLIGENCE TECHNIQUES IN DATA ANALYSIS FOR CANCER DETECTION: A SYSTEMATIC ANALYSIS HOANG THANH NHAT HANOI – YEAR 2024 VIETNAM NATIONAL UNIVERSITY, HANOI INTERNATIONAL SCHOOL ------------------------------- GRADUATION PROJECT EXPLORING THE APPLICATION OF ARTIFICIAL INTELLIGENCE TECHNIQUES IN DATA ANALYSIS CANCER DETECTION: A SYSTEMATIC ANALYSIS SUPERVISOR : Ph. Dinh-Toi Chu STUDENT NAME : Thanh Nhat Hoang COHORT : BDA2020A MAJOR : Business Data Analytics HANOI – YEAR 2024 ACKNOWLEDGEMENTS Upon completing my graduation project as stipulated by the International School, Vietnam National University, Hanoi (VNUIS), I have acquired significant hands-on experience in data analysis and assessment within the biomedical data science domain. My project took me close to four months to complete. This initiative has brought to light the significance of the fundamental information I gained during my university education, which has been essential to my success and will present me with many prospects for growth in the future.
Apart from utilizing my academic expertise, I conducted a thorough exploration of the subject of biomedicine, focusing especially on cancer and the application of Artificial Intelligence in the detection of this health issue. The most important lesson I've learned from my graduation assignment is this. With profound gratitude and respect, I want to increase my heartfelt regard for my supervisor, Ph. Dinh Toi-Chu, for his dedicated guidance.
From the very beginning of my thesis work, he has been an enthusiastic mentor, guiding me and introducing me to new knowledge. Throughout the process of completing my project, his support, along with the enthusiastic support and guidance MSc. Hue Vu Thi, a member of the VNUIS’ Center for Biomidicine and Community Health, was invaluable in overcoming numerous challenges and enabling me to successfully complete this project. i DECLARATION I, Thanh Nhat Hoang, a student majoring in Business Data Analytics at the International School, Vietnam National University, Hanoi (VNUIS), hereby declare that: 1.
This graduation report titled "Exploring the Application of Artificial Intelligence Techniques in Data Analysis Cancer Detection: A Systematic Analysis" is the result of my research and hard work under the guidance of Ph. Dinh Toi Chu. I have independently conducted the data collection, analysis, and evaluation of relevant documents. All sources and references in this report have been clearly, honestly, and accurately cited according to the regulations of the VNUIS.
I commit that there is no plagiarism or copyright infringement during the implementation of this report. I take full responsibility under the law and the school's regulations for the honesty and accuracy of the contents in this report. Any errors or violations (if any) will be my personal responsibility, and I commit to accepting any disciplinary actions from the school. This report has not been submitted or published at any other organization, institution, or agency other than the VNUIS.
With all sincerity and respect, please allow me to extend my heartfelt thanks to all the teachers for their support and guidance, especially Ph. Dinh Toi Chu and MSc. Hue Vu Thi, a member of the the VNUIS’ Center for Biomidicine and Community Health, throughout the process of completing this graduation report. Hanoi, 2024 Thanh Nhat Hoang ii ABSTRACT In the current context in the world in general and in Vietnam in particular, there are very few published articles and research on the use of Artificial Intelligence (AI) technology in early detection cancer.
Therefore, this is the goal that I want to do this research through an extensive systematic study, through an extensive systematic study, this research investigates the AI utilization approaches in data analysis and diagnosis for cancer. AI has improved early detection, improved treatment strategies, and enhancing the prognosis for cancer patients, revolutionizing the field. The research looks at different AI models and evaluates how well they detect diseases such skin, lung, breast, and prostate cancers. These models include existing AI models, which have been developed by researchers, followed by Deep Learning (DL) models, Neural Network (NN), and hybrid methods.
The literature review addresses the shortcomings of conventional techniques by describing the development of AI in healthcare and its crucial role in cancer detection and treatment. The methodology in this work outlines the systematic review process, including data sources, selection criteria, and analytical techniques. Results indicate that AI models significantly enhance the accuracy and speed of cancer detection, with CNNs excelling in image analysis and DL methods handling large datasets effectively. Hybrid models combine AI techniques to refine detection accuracy and efficiency.
Despite these advancements, challenges like the requirement for extensive training data, high computational resources, and ethical concerns regarding patient data privacy persist. The report emphasizes the importance of standardized evaluation metrics to ensure the reliability and generalizability of AI applications in clinical settings. It concludes by discussing AI's potential to transform oncology, offering recommendations for future research to address current limitations. By integrating AI more effectively into clinical practice, significant advancements in early cancer detection and treatment can be achieved, ultimately improving patient care and outcomes.
iii TABLE OF CONTENTS ACKNOWLEDGEMENTS. iii TABLE OF CONTENTS .iv LIST OF KEYWORDS .vi LIST OF FIGURES. vii LIST OF TABLES. AI (AI) technology and applications in health .2 Cancer, cancer data analysis, and the limitations .3 AI technology aids in the detection and treatment of cancer.
AI Algorithms and Models in Medical Data and Cancer Simulation. Used in diagnosis to detect early signs of cancer. Systematic Review Method for Medical and Cancer Data Analysis. Systematic review analysis in medicine.
Systematic review analysis in the field of cancer. Reasons for conducting a systematic review study. Process of Systematic Review. Selection and exclusion standards.
Data sources and techniques for searching .5 Research screening and selection process. AI in Detection of Cancer: A Systematic Review .2 Characteristics of research subjects. Effectiveness of AI in Early Cancer Detection. AI Application in Cancer Detection Using Existing Models.
AI Detection in Cancer: Analysis of DL and NN Approaches. AI Detection in Cancer: Analysis of Hybrid Approaches .3 AI model Advantages and Disadvantages. Advantages and disadvantages of Existing AI Models. Advantages and disadvantages of DL and NN model.
Advantages and disadvantages of the Hybrid approach.4 Additional research potential on the use of AI in oncology. CONCLUSION AND RECOMMENDATIONS. 76 Finally, testing and applying AI in cancer diagnosis needs to be tested and tested for rigorous accuracy before being put into clinical trials. 77 v LIST OF KEYWORDS Term/Acronym Definition/Description KNN K-Nearest Neighbors CNN Convolutional Neural Network AI Artificial Intelligence DNN Deep Neural Network ML Machine Learning PRISMA Preferred Reporting Items for Systematic Review and Meta-analysis CT Computed Tomography Scan PC Prostate Cancer AUC Area Under Curver ACC Accuracy SEN Sensitivity DL Deep Learning NN Neural Network ANN Artificial Neural Network RNN Recurrent Neural Network SVM Support Vector Machine RF Random Forest KNN K-Nearest Neighbors FP False Positive FN False Negative vi LIST OF FIGURES Figure 2.1 Research Design Diagram .Error! Bookmark not defined.
Search results and data screening based on PRISMA diagram. Year of publication of the article. Types of data are used. Proportion type of Cancer.
Article punlished by countries. 73 vii LIST OF TABLES Table 3. Characteristics of articles selected for the study. Analysis of application AI in detection based on Existing Models approaches in Cancer.
Analysis of application AI in detection based on DL and NN approaches in Cancer. Analysis of application AI in detection based on Hybrid approaches in Cancer. Advantages and disadvantages of Existing AI Models. Advantages and Disadvantages of the DL and NN approaches.
Advantages and Disadvantages of the Hybrid approaches. This technology was first proposed by Alan Turing (Father of AI) in 1950, who conducted an experiment called the “Turing Test” on the philosophy that the development of intelligent behavior of computers related to cognition is to achieve human performance levels, thereby distinguishing humans from machines [1, 3]. Through the above experiment, he described AI technology as the human brain but reimagined and more complex [2]. Then in 1956, the term AI (AI) was officially born under the announcement of John McCarthy in a conference on this topic[3].
Over time, under the interest and support of many people around the world for AI, this technology is increasingly researched and developed more deeply, divided into smaller branches such as machine learning, deep learning, and algorithms, e. Besides, AI is also strongly applied in research and support in disease diagnosis and treatment [5]. In the medical field, AI has contributed to improving and developing disease diagnosis, healthcare, and treatment activities using methods of AI, including artificial neural networks, hybrid intelligence systems e. To easily classify AI in the health field, it is divided into two small parts: the virtual part including health care software, machine learning, and algorithms, e.c, and the physical part such as robots, and smart medical tools [1].
AI applications in the medical field are extremely prominent, demonstrated through many different methods and aspects such as: supporting disease diagnosis and treatment, administrative applications, and record management. patients, Healthcare, medical research [6, 7]. Furthermore, data sources in the health sector in general are becoming more complex with more and more research being conducted and 1 data sources increasingly dense [7]. This shows that AI will be applied more and more and reflects its importance in the current context of supporting humans [7].
However, because of these factors, ethical and social issues in the application of AI have also been raised and discussed about information security and patient safety, showing that the management of Using AI is extremely important to ensure the research process, as well as for patient privacy [6]. Among the above applications of AI, the application of AI in cancer diagnosis is one of the top concerns of researchers and doctors, with great promise in improving accuracy, identify and detect cancer early, thereby improving the survival rate for patients [8, 9]. The rapid and remarkable development of AI in recent years has encouraged scientists to research, improve, and develop AI applications in cancer diagnosis and treatment. AI's early cancer diagnosis technology is applied in most cancers such as breast cancer, cervical cancer, myeloma.
By developing detection of cancer methods grounded in deep learning and machine learning, AI algorithms, along with a huge data warehouse, this technology has achieved outstanding results in terms of accuracy. as well as optimizing disease diagnosis time compared to previous traditional methods [8, 9]. Thereby, AI applications in cancer diagnosis include: early diagnosis, disease condition assessment, thereby supporting doctors in providing appropriate and timely treatment [8, 9]. However, to be able to apply AI most accurately in diagnosis requires a large amount of data to be used for analysis and evaluation [10].
Not only that, when using AI we will have to pay attention to ethical issues [8]. In short, the use of AI brings outstanding results in cancer diagnosis with full promise of further developments in the future of AI.2 Cancer, cancer data analysis, and the limitations Traditionally, data on cancer was generated based on common imaging techniques (radiological scans, ultrasound, computerized scans) [11], histological tests and stains [12], analysis of bodily fluids (blood, urine, liquid biopsy) which tests for tumor signals [13] or are used in cytology [14] to detect features that presuppose the presence of tumors, and text-based medical records indicated on file. Jiang et al. outlined five main types of oncologic data: data from imaging, phenotypic analysis, molecular interactions, and text 2 [15].
Newer additions to cancer databases included “omics” data that provides molecular evidence for cancer research and detection (genomics, epigenomics, proteomics, transcriptomics, and metabolomics) [15]. However, advances in testing methodologies, data science, and storage allowed for a larger quantity of data – “big data” – to be generated and processed, which is no longer economical or feasible to assess manually. Furthermore, human interpreters struggle with reduced accuracy and attention to detail when working long hours with repetitive tasks.