Machine learning classification of surgical pathology reports and chunk recognition for information extraction noise reduction

Background and aims: Machine learning techniques for the text mining of cancer-related clinical documents have not been sufficiently explored. Here some techniques are presented for the pre-processing of free-text breast cancer pathology reports, with the aim of facilitating the extraction of information relevant to cancer staging.

Materials and methods: The first technique was implemented using the freely available software RapidMiner to classify the reports according to their general layout: ‘semi-structured’ and ‘unstructured’. The second technique was developed using the open source language engineering framework GATE and aimed at the prediction of chunks of the report text containing information pertaining to the cancer morphology, the tumour size, its hormone receptor status and the number of positive nodes. The classifiers were trained and tested respectively on sets of 635 and 163 manually classified or annotated reports, from the Northern Ireland Cancer Registry.

Results: The best result of 99.4% accuracy – which included only one semi-structured report predicted as unstructured – was produced by the layout classifier with the k nearest algorithm, using the binary term occurrence word vector type with stopword filter and pruning. For chunk recognition, the best results were found using the PAUM algorithm with the same parameters for all cases, except for the prediction of chunks containing cancer morphology. For semi-structured reports the performance ranged from 0.97 to 0.94 and from 0.92 to 0.83 in precision and recall, while for unstructured reports performance ranged from 0.91 to 0.64 and from 0.68 to 0.41 in precision and recall. Poor results were found when the classifier was trained on semi-structured reports but tested on unstructured.

Conclusions: These results show that it is possible and beneficial to predict the layout of reports and that the accuracy of prediction of which segments of a report may contain certain information is sensitive to the report layout and the type of information sought.

D4.3 Prototype Software (Documentation)

The agent-based social simulation component of the TELL ME project (WP4) developed prototype software to assist communications planners to understand the complex relationships between communication, personal protective behaviour and epidemic spread. Using the simulation, planners can enter different potential communications plans, and see their simulated effect on attitudes, behaviour and the consequent effect on an influenza epidemic.

The model and the software to run the model are both freely available (see section 2.2.1 for instructions on how to obtain the relevant files). This report provides the documentation for the prototype software. The major component is the user guide (Section 2). This provides instructions on how to set up the software, some training scenarios to become familiar with the model operation and use, and details about the model controls and output.

The model contains many parameters. Default values and their source are described at Section 3. These are unlikely to be suitable for all countries, and may also need to be changed as new research is conducted. Instructions for how to customise these values are also included (see section 3.5).

The final technical reference contains two parts. The first is a guide for advanced users who wish to run multiple simulations and analyse the results (section 4.1). The second is to orient programmers who wish to adapt or extend the simulation model (section 4.2). This material is not suitable for general users.