Current and Future Applications of Artificial Intelligence in Oncology

Applications of AI

Nearly every area of ​​our lives and our patients’ lives is being affected by the increasing use of artificial intelligence (AI), whether we are explicitly aware of it or not. In healthcare, AI-based tools have already been implemented clinically in various specialties, including radiology, pathology, and dermatology. With a significant increase in capabilities and related applications, the use of AI has expanded from primarily diagnosis and screening to prognosis, treatment monitoring, and even treatment selection. With the increasing availability of “big data” sets, the use of AI in research settings has expanded to include new applications such as radiological and histological evaluation, drug discovery and development, and novel biomarker and genomic prediction algorithms. Such applications have further opened up the possibility of developing tumor-agnostic therapies targeting novel genomic and/or microenvironment signatures. Finally, the use of AI at a systems level has the potential to improve healthcare delivery in response to the growing diverse community affected by cancer.

The digitization of histology slides is also expanding the use of AI in pathology. The Cancer Genome Atlas is one of the largest biorepositories, containing more than 10,000 digital pathology images from more than 20 cancer types, along with associated clinical and genomic data. Several studies are using this and other repositories to develop diagnostic, prognostic, and predictive AI models that identify subtle histological features and patterns. This approach allows new digital pathology and genomic biomarkers to be developed and, once validated, used for diagnosis and treatment monitoring. For example, with the increasing use of immunotherapy and recently approved antibody-drug conjugates (e.g., trastuzumab deruxtecan, Enhertu), AI technologies can aid in treatment selection by accurately quantifying PD-1 expression in the tumor microenvironment and tumor cell expression of low or very low levels of HER2 in digital pathology images. Finally, the use of AI techniques to analyze genomic biomarkers (outside of histology) is an exciting area of ​​development for treatment selection and could be crucial in advancing new screening technologies that rely on genomic patterns detected in extracellular DNA, i.e. the much-needed “cancer screening blood test.”

One of the most exciting, yet challenging areas of AI use in oncology is treatment selection and monitoring. Currently, several platforms are used for early drug discovery, processing clinical, genomic, and proteomic data to identify therapeutic targets and select relevant molecules for further development. In the clinical setting, AI tools can help predict treatment resistance based on patient and tumor characteristics, as well as ex vivo testing of biopsy samples. Incorporating new biomarker datasets beyond bulk tumor sequencing, such as single-cell sequencing to distinguish tumor and microenvironment components, may reduce the need for tissue samples in future AI-based treatment prediction tools. Personalized drug dosing is another exciting potential use of AI in oncology, with tools currently under development leveraging large datasets that include host factors (e.g., body mass index, comorbidities, functional status), patient-reported outcomes, and side-effect profiles in addition to traditional clinicopathological features.

AI holds great promise, but several key challenges must be addressed to continue to scale AI technology in clinical implementation. First, datasets, including digitized images, must be standardized in terms of variables, quality, processing and storage procedures, and other parameters to maximize the potential of deep learning. Second, many diagnostic and predictive models derived from AI require validation before large-scale clinical implementation is possible. To include broad and diverse datasets in AI modeling, transparency and trust regarding privacy and data use must be well established. The legal and ethical implications of the use of AI in healthcare are evolving and must continue to be centrally addressed as AI applications expand. Related to privacy and trust, it is essential to include underrepresented minority populations in training datasets or to have dedicated datasets that can be included in machine learning to create equity in the application of AI. Finally, the framework within which AI is implemented must enhance, rather than overshadow, oncologists and patient-centered decision making. The potential of AI in oncology is very promising across multiple domains, but human ingenuity is needed to realize this potential to the fullest.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *