Research Assessment #3
Date: September 18, 2020
Subject: Artificial intelligence impact on Oncology
Citation: Shimizu, Hideyuki, and Keiichi I Nakayama. "Artificial intelligence in oncology." Cancer science vol. 111,5 (2020): 1452-1460. doi:10.1111/cas.1437
Assessment:
Artificial intelligence has impacted many fields over the last decade, including the field of oncology. Through deep learning, cancer images, genomics and personalized medicine artificial intelligence (AI) are leading the future in the identification and treatment of cancer in patients.
As a sub-field of machine learning, deep learning has impacted both basic and clinical cancer research and has helped to solve problems that were once deemed unsolvable. Unlike Machine learning which applies mathematical and statistical approaches in order to improve the performance of computers, deep learning is characterized by layering artificial neural networks. Deep learning is a neural network composed of a connected input layer, hidden layer, and output layer which allows deeper layers in the network to automatically create features that are needed to solve the problem at hand. Compared to the best existing machine learning algorithms these new deep learning techniques have made improvements in performance and the field of biomedicine due to its application to the diagnostics of disease. Additionally, deep learning has proved highly accurate for the detection of retinopathy from fundus photographs.
The ability of AI to be applied to cancer image analysis is important because the early detection of cancer is key to saving lives. This is why so many researchers use AI and apply it to clinical radiology and pathology because of its ability to create detailed features. It can even be used in the classification of dermoscopy images and has the ability to determine skin problems (including melanoma) just as precisely as expert dermatologists. In the future, this has the ability to provide a dermatologist level diagnosis, from a smartphone that not only provides convenience for a patient but could be beneficial in the times of a pandemic. Additionally, the interpretation of mammograms for breast cancer screening AUC for AI is .840 compared to .814 for physicians. The ability AI has to detect problems early on is one of great hopefulness for preventative measurement. Deep learning also has the ability to output a map of where potential cancer cells are present and determine the probability of a cancerous cell. Although standard protocol still needs to be established for automatic histopathology analysis, AI can be of great help in reducing the burden on medical staff involved in the assessment of a tumor. In the near future, deep learning will become an important tool for pathologists to improve the accuracy of diagnosis and informative treatment selection.
Although there are problems that occur such as the magnitude of available image collection (less than one million), data augmentation allows for increased accuracy since images can randomly be cropped, tilted, or inverted.
Additionally, deep learning is useful in cancer genetics because it allows not only for single-task learning but multitask learning. Furthermore, it allows for multimodal learning, which is a method that integrates different types of data as inputs. Since cancer is a complex disease that is affected by genetics, environmental factors, and diet, multilayer data is beneficial. The application of AI and deep learning allows large amounts of data to be analyzed and for the best solution to be found.
Nonetheless, artificial intelligence has the ability to affect the area of precision medicine. Currently, disease identification in oncology includes the microscopic anatomy of biological tissues known as histological examination. This procedure is done by a pathologist on the protein or mRNA level. Artificial intelligence helps deep learning-based AI algorithms to detect, diagnose, and predict certain cancer types. Additionally, the inclusion of artificial intelligence into this procedure would help classify the identification further.
Overall as AI is implemented into the field of oncology more preventative measures to prevent the problem in the first place will arise, alongside specific and patient-designed treatment plans.