|
Abstract:
|
Due to the development of machine learning algorithms and increasing computing power, current neural networks algorithms perform much better in prediction, compared to the traditional statistical models. The method of machine learning was initially developed in the discipline of computer science and quickly was applied in many other fields. An interdisciplinary approach is much needed for the predictions in cancer research.
In observational studies, statistical models are often used to control for covariates and tease out the net effects. However, in other situation, predictions are a major concern. For example, cystectomy is a standard procedure to treat muscle-invasive bladder cancer patients. The procedure of cystectomy causing a very high readmission rate (~25%-30%). If the prediction algorithms could predict who is likely to readmitted, with high sensitivity and specificity, at the time of surgery, this could better inform both the surgeons and patients and improve the health outcomes.
This talk will demonstrate how machine learning algorithms will be used in prediction questions.
|
