Source-linked AI summary
Artificial Intelligence Algorithms for the Detection of Pathologies Related to Lung Cancer through Image Analysis using Convolutional Neural Networks and Data Augmentation: a systematic mapping of the literature
Pablo Ramirez Amador
TL;DR
Lung-cancer image interpretation is complex, while early diagnosis is important for patients. This systematic mapping reviews AI and deep-learning applications, emphasizing CNN-based image analysis, and concludes that these systems can detect related pathologies while unresolved data, privacy, and clinical-validation challenges remain.
Problem
Lung cancer has low survival, early diagnosis is important, and interpreting CT and other medical images is complex and often requires trained experts.
Method
The paper systematically maps literature on AI and deep learning for lung-cancer image analysis, examining CNNs, Data Augmentation, and related detection systems.
Results
The reviewed systems can detect lung-cancer-related pathologies through image analysis or deep-learning mechanisms, with image datasets and detection or machine-learning techniques commonly used.
Takeaways & Limitations
Maintaining an up-to-date and diverse lung-image database can support algorithm training and validation and help specialists compare similar cases.
Takeaways & Limitations
The mapping identifies unresolved problems that require further work, including clinical-practice challenges and issues addressed in the proposed doctoral research.
Abstract
from arXiv · showhide
Lung cancer is one of the leading causes of death worldwide, and its early diagnosis is crucial to improving patients prognosis and quality of life. However, the process of interpreting medical images for the detection of lung cancer is complex and requires trained experts. In this context, artificial intelligence (AI) and deep learning (DL) emerge as potential tools to automate and optimize image analysis. The objective of this work is to review the most recent and relevant applications of AI and DL in the field of radiology for the detection of lung cancer. To this end, an exhaustive search was carried out in scientific databases such as PubMed,IEEEXPLORE, Scopus and Web of Science, and 96 articles published from 2015 to the present addressing the use of AI and DL in biomedical engineering were selected. Emphasis is placed on the use of convolutional neural networks (CNN) with transfer learning and Data Augmentation as promising techniques to improve the accuracy and efficiency of the image interpretation process. The results show that the use of AI and DL can offer an effective alternative for the early diagnosis of lung cancer, with high sensitivity and specificity. However, current limitations and challenges that must be addressed to guarantee its responsible and safe application in clinical practice are also identified, such as the lack of standardized data, the ex plainability of the models, patient privacy, and the ethical and social implications. It is concluded that the use of AI and DL can have a positive impact on the care of patients with lung cancer, but further research and regulation are required to ensure its quality and reliability.
Introduction
Lung cancer has low survival, making early diagnosis important, but CT interpretation is complex and depends on trained experts. The section presents AI, especially pre-trained CNNs with Data Augmentation, as an approach for detecting NSCLC from CT images.
- Introduction: Early diagnosis is crucial because lung cancer has low survival and NSCLC is the most common type, with slow growth and poor chemotherapy response.
- Introduction: CT provides three-dimensional thoracic images that can reveal pulmonary abnormalities such as nodules or masses.
- Introduction: Interpreting medical images is complex and often requires highly trained experts, limiting diagnostic availability and access.
- Introduction: Pre-trained CNNs with Data Augmentation are proposed to detect NSCLC from CT images while increasing training-data variety and reducing overfitting risk.Transfer learning reuses CNNs trained on generic datasets, while transformations such as rotation, scaling, and noise generate additional images.
- Introduction: The work aims to compare pre-trained CNNs, evaluate Data Augmentation’s influence on accuracy and robustness, and assess feasibility for NSCLC detection.
2. Research questions
The mapping uses critical reading of selected literature and is guided by research questions.
- 2. Research questions: Critical reading of the selected material guides the systematic mapping analysis.
- 2. Research questions: The mapping’s guiding questions are presented in Table 1.
3. Review methods
The review follows a structured systematic-search process to identify, filter, classify, and analyze literature on AI and deep learning for lung-cancer image detection. It defines the principal concepts and examines findings, implications, limitations, and challenges.
- 3. Review methods: The search targeted articles from 2015 onward in PubMed, IEEEXPLORE, Scopus, and Web of Science using AI, deep learning, CNN, Data Augmentation, lung cancer, diagnosis, and image-analysis terms.
- 3. Review methods: The extracted studies are analyzed by variables including publication year, country, journal, CNN type, Data Augmentation use, and maturity level.
- 3. Review methods: The review defines AI, DL, CNNs, Data Augmentation, lung cancer, diagnosis, and image analysis for consistent use throughout the article.
- 3. Review methods: Data Augmentation generates transformed images through rotation, scaling, or noise to increase training-data variety and improve robustness to test-image variation.
- 3. Review methods: The review protocol includes planning, execution, and reporting, with database searching followed by inclusion and exclusion criteria.
- 3. Review methods: A total of 96 articles met the inclusion criteria, and their titles, years, authors, journals, objectives, methods, results, and conclusions were extracted and classified.
4. Search for works
The search process established a review protocol, constructed topic-specific search terms, and extracted data from the selected literature. The reported search stage initially obtained 167 articles.
- 4. Search for works: The search procedure began by defining research questions, search strings, and inclusion and exclusion criteria.
- 4. Search for works: The review protocol included establishing the search and selection procedure before article classification and synthesis.
- 4. Search for works: The search and filtering process initially obtained 167 articles.
5 Synthesis of extracted data
The synthesis organizes the selected literature around CNN suitability, Data Augmentation benefits, and clinical-application challenges. The review selected 59 and 35 articles during its study process.
- The synthesis evaluates which CNN type is most suitable for lung-cancer image analysis.
- It examines the benefits of Data Augmentation for increasing the quantity and variety of training data.
- It also considers challenges and limitations affecting AI diagnosis in clinical practice.
6. Conclusions
The conclusions characterize the mapped literature on CNN- and Data Augmentation-based detection of lung-cancer-related pathologies. They report image-analysis and deep-learning applications, emphasize dataset development for specialist support, and identify unresolved problems for future doctoral research.
- The mapping separately examined CNN and Data Augmentation studies before analyzing systems that combined them in one evaluation framework.
- These systems can warn of favorable or unfavorable diagnoses and provide additional control functionalities using deep-learning techniques.
- AI algorithms can detect lung-cancer-related pathologies through image analysis or deep-learning mechanisms.
- Up-to-date, diverse lung-image databases can broaden training and validation cases and help specialists compare similar cases.
- The mapping identifies unresolved problems that the planned doctoral thesis will attempt to address.
Bibliografía
The bibliography brings together systematic reviews, mapping studies, and research on deep learning, CNNs, transfer learning, segmentation, uncertainty, and lung-cancer-related medical-image analysis.
- The references include reviews and mapping studies of deep learning and medical-image analysis.
- Several cited works address CNN architectures, transfer learning, and medical-image classification or detection.
- The cited applications span lung, thyroid, breast, liver, vascular, and other biomedical imaging problems.
- Other references cover segmentation, registration, vascular applications, and reconstruction in medical imaging.
- The bibliography also includes work on federated learning and predictive uncertainty in medical-image analysis.