Recent advancement in cancer diagnosis using machine learning and deep learning techniques: A comprehensive review

Area of Science:

• Oncology
• Artificial Intelligence
• Medical Imaging

Background:

• Cancer is a leading cause of mortality globally, necessitating improved early detection and prevention strategies.
• Advanced stage cancer diagnosis often has limited impact on patient survival.
• Machine learning (ML) and deep learning (DL) offer enhanced efficiency and reduced error rates in cancer diagnosis compared to human capabilities.

Purpose of the Study:

• To review, analyze, and classify recent advancements in ML and DL for cancer detection and diagnosis.
• To focus on six major cancer types: breast, lung, liver, skin, brain, and pancreatic cancer.
• To assess the performance of various state-of-the-art techniques using key performance indicators.

Main Methods:

• Comprehensive literature review of ML and DL-based cancer detection methods published in the last six years.
• Analysis of different data modalities and feature extraction techniques.
• Classification of state-of-the-art techniques and examination of their results on benchmark datasets.

Main Results:

• Significant progress in ML and DL applications for cancer segmentation and classification over the past decade.
• Evaluation of techniques based on accuracy, area under the curve, precision, sensitivity, and dice score.
• Identification of trends and performance benchmarks for various cancer detection methods.

Conclusions:

• ML and DL have revolutionized cancer diagnosis, offering powerful tools for early detection.
• Continued research is needed to address challenges and further enhance the capabilities of AI in oncology.
• Future work should focus on refining algorithms, validating on diverse datasets, and clinical integration for improved patient care.