Oncology is one such field of medical science that is constantly evolving. The word cancer still evokes fear and shock in everyone’s mind. Cancer is a heterogeneous disease which consists of many different subtypes. Early diagnosis and prognosis of this deadly disease have become a necessity in cancer research. The advent of new technologies in the field of medical sciences have helped the medical research community to analyze a vast amount of data sets. The only hindrance is to access the accurate prediction of the disease, which is interesting and challenging. Adaption of Machine learning techniques have now become a powerful tool to discover and identify patterns between complex data sets and effectively predict the outcomes.
Machine Learning Techniques
Machine learning is a branch of Artificial Intelligence which learns from data samples and use that to classify new data, identify new patterns or predict trends. Machine learning has proven to be a boon for biomedical research to help the researchers search through an n-dimensional space for a given set of samples using different algorithms and techniques. Mainly, there are 2 common types of machine learning methods, supervised learning and unsupervised learning. In supervised learning, data is labeled. The model identifies the labels and groups accordingly. Simply put, the model is provided with all the inputs and then told the expected output. In contrast to this, unsupervised learning methods have no label for data. The models identify different features and classify them based on different characteristics. In this method, the input is provided and the computer then learns to find patterns and make logical classification or groupings.
Application To Cancer Research
The elemental goals of cancer prognosis and prediction are different from the goals of cancer diagnosis and detection. In cancer prediction and prognosis there are three center points:
- Prediction of cancer susceptibility ( risk assessment)
- Prediction of cancer recurrence
- Prediction of cancer survivability
Risk assessment is more about trying to predict the likelihood of developing some type of cancer before the occurrence of the disease. In the prediction of cancer recurrence, one tries to predict the likelihood of redevelopment of cancer, right after the apparent resolution of the disease. The last point, prediction of cancer survivability predicts the outcome after the disease has been diagnosed, such as survivability, life expectancy, progression, tumor-drug sensitivity.
Machine learning helps researchers identify and classify tumors based on growth characteristics: where they grow, size, the speed of spread etc., and group them together based on a similar range of predictive outcomes.
Cancer Detection From Data Sets
Gene expression data is very complex as it is highly dimensional and makes it challenging to leverage that data in cancer detection. Researchers have been able to use deep learning to extract meaningful features from the gene expression data which has enabled the classification of breast cancer cells. In the past, Google’s CNN system has showcased the ability to identify deadline skin cancers. Researchers from China have leveraged deep learning for segmenting brain tumors in Magnetic Resonance (MR) imaging which yielded more stable results in comparison to one done manually by physicians as it was more prone to vision errors. Machine learning has also assisted in measuring the size of tumors undergoing treatment and detect other metastases which might have been overlooked.
The biggest benefit of using Artificial intelligence and machine learning is that these next-gen technologies have the capability to read a vast amount of data with utmost accuracy, thus eliminating the manual time-consuming process. The technologies also help to reduce dependence on the limited judgment and skills of a specialized expert. Machine learning is paving the way for the future in medical sciences since research on cancer, its cure and treatment has been a prominent focus for ages.