Natural language processing and machine learning approaches for parsing pathology reports

Gciniwe Dlamini from IBM Research will present this seminar.

Unstructured medical notes, such as pathology reports and medical discharge summaries, contain a lot of information about patients. This information can be mined for potentially actionable insights to improve the care of patients and to improve health systems in general. However, information extraction from these medical texts is a challenge, which has sparked interest from both the natural language processing (NLP) and medical informatics communities. Subsequently, statistical modelling and machine learning approaches are increasingly being used in conjunction with NLP methods to extract relevant content from medical texts effectively and efficiently. In IBM’s recent work, the utility of machine learning in this setting was demonstrated through a case study where machine learning classifiers were built for the automatic labelling of free-text, breast cancer pathology reports, according to their International Classification of Diseases for Oncology (ICD-O) topographical codes across nine classes for breast cancer.

Add event to calendar