Application of Artificial Intelligence in Early Detection and Prognostic Prediction of Multi-Organ Cancers: A Systematic Review of Research Gaps

Artificial intelligence, which includes machine learning and deep learning, has made big strides in detecting cancer and predicting patient outcomes. It's being used in radiology, digital pathology, and molecular data, and it's getting really good at it. But even with all this progress, it's still not being used as much as it could be in regular clinical practice. And when it comes to actually helping patients, the evidence is a bit spotty - it works really well for some types of cancer, but not as well for others. This study looks at the main problems that stop artificial intelligence from being used to detect and predict cancers that affect many parts of the body. It also suggests specific areas where more research is needed to fill these gaps. We did a thorough review of recent studies from top scientific databases like PubMed, Scopus, and Web of Science. We only looked at studies that used AI for detecting, diagnosing, or predicting cancers, and that also considered how well the AI systems worked in real-life situations, how fair they were, and how useful they were for doctors. We grouped the results into categories and summarized the evidence, the importance of each problem, and potential solutions. We found ۱۱ main areas where more research is needed: AI systems that don't work well outside the lab, biased datasets, AI systems that are too complex to understand, not enough evidence that AI can save lives, predictive models that don't consider what doctors actually need, problems combining different types of data, some types of cancer being neglected, difficulties in using liquid biopsies and biomarkers, not enough data from clinical trials, AI systems that can't adapt to changing regulations, and unfair access to AI-powered healthcare. The most important and achievable goals are to do more studies that compare AI to traditional methods, test AI systems in real-life situations with diverse groups of people, design AI systems that are easier to understand, and create large datasets that combine imaging and biomarker data from many different sources.

کلیدواژه ها:

artificial intelligence ، deep learning ، cancer detection ، prognostic prediction ، external validation ، explainable AI ، multi-omics ، health equity