Dataset Information

Automatic extraction of cancer registry reportable information from free-text pathology reports using multitask convolutional neural networks.

ABSTRACT: OBJECTIVE:We implement 2 different multitask learning (MTL) techniques, hard parameter sharing and cross-stitch, to train a word-level convolutional neural network (CNN) specifically designed for automatic extraction of cancer data from unstructured text in pathology reports. We show the importance of learning related information extraction (IE) tasks leveraging shared representations across the tasks to achieve state-of-the-art performance in classification accuracy and computational efficiency. MATERIALS AND METHODS:Multitask CNN (MTCNN) attempts to tackle document information extraction by learning to extract multiple key cancer characteristics simultaneously. We trained our MTCNN to perform 5 information extraction tasks: (1) primary cancer site (65 classes), (2) laterality (4 classes), (3) behavior (3 classes), (4) histological type (63 classes), and (5) histological grade (5 classes). We evaluated the performance on a corpus of 95 231 pathology documents (71 223 unique tumors) obtained from the Louisiana Tumor Registry. We compared the performance of the MTCNN models against single-task CNN models and 2 traditional machine learning approaches, namely support vector machine (SVM) and random forest classifier (RFC). RESULTS:MTCNNs offered superior performance across all 5 tasks in terms of classification accuracy as compared with the other machine learning models. Based on retrospective evaluation, the hard parameter sharing and cross-stitch MTCNN models correctly classified 59.04% and 57.93% of the pathology reports respectively across all 5 tasks. The baseline models achieved 53.68% (CNN), 46.37% (RFC), and 36.75% (SVM). Based on prospective evaluation, the percentages of correctly classified cases across the 5 tasks were 60.11% (hard parameter sharing), 58.13% (cross-stitch), 51.30% (single-task CNN), 42.07% (RFC), and 35.16% (SVM). Moreover, hard parameter sharing MTCNNs outperformed the other models in computational efficiency by using about the same number of trainable parameters as a single-task CNN. CONCLUSIONS:The hard parameter sharing MTCNN offers superior classification accuracy for automated coding support of pathology documents across a wide range of cancers and multiple information extraction tasks while maintaining similar training and inference time as those of a single task-specific model.

SUBMITTER: Alawad M

PROVIDER: S-EPMC7489089 | biostudies-literature | 2020 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Automatic extraction of cancer registry reportable information from free-text pathology reports using multitask convolutional neural networks.

Alawad Mohammed M Gao Shang S Qiu John X JX Yoon Hong Jun HJ Blair Christian J J Penberthy Lynne L Mumphrey Brent B Wu Xiao-Cheng XC Coyle Linda L Tourassi Georgia G

Journal of the American Medical Informatics Association : JAMIA 20200101 1

<h4>Objective</h4>We implement 2 different multitask learning (MTL) techniques, hard parameter sharing and cross-stitch, to train a word-level convolutional neural network (CNN) specifically designed for automatic extraction of cancer data from unstructured text in pathology reports. We show the importance of learning related information extraction (IE) tasks leveraging shared representations across the tasks to achieve state-of-the-art performance in classification accuracy and computational ef ...[more]

PMID: 31710668

Similar Datasets

Project description:BackgroundFor ecological studies, it is crucial to count on adequate descriptions of the environments and samples being studied. Such a description must be done in terms of their physicochemical characteristics, allowing a direct comparison between different environments that would be difficult to do otherwise. Also the characterization must include the precise geographical location, to make possible the study of geographical distributions and biogeographical patterns. Currently, there is no schema for annotating these environmental features, and these data have to be extracted from textual sources (published articles). So far, this had to be performed by manual inspection of the corresponding documents. To facilitate this task, we have developed EnvMine, a set of text-mining tools devoted to retrieve contextual information (physicochemical variables and geographical locations) from textual sources of any kind.ResultsEnvMine is capable of retrieving the physicochemical variables cited in the text, by means of the accurate identification of their associated units of measurement. In this task, the system achieves a recall (percentage of items retrieved) of 92% with less than 1% error. Also a Bayesian classifier was tested for distinguishing parts of the text describing environmental characteristics from others dealing with, for instance, experimental settings.Regarding the identification of geographical locations, the system takes advantage of existing databases such as GeoNames to achieve 86% recall with 92% precision. The identification of a location includes also the determination of its exact coordinates (latitude and longitude), thus allowing the calculation of distance between the individual locations.ConclusionEnvMine is a very efficient method for extracting contextual information from different text sources, like published articles or web pages. This tool can help in determining the precise location and physicochemical variables of sampling sites, thus facilitating the performance of ecological analyses. EnvMine can also help in the development of standards for the annotation of environmental features.

Project description:BackgroundLung cancer is the leading cause of cancer deaths worldwide. Clinical staging of lung cancer plays a crucial role in making treatment decisions and evaluating prognosis. However, in clinical practice, approximately one-half of the clinical stages of lung cancer patients are inconsistent with their pathological stages. As one of the most important diagnostic modalities for staging, chest computed tomography (CT) provides a wealth of information about cancer staging, but the free-text nature of the CT reports obstructs their computerization.ObjectiveWe aimed to automatically extract the staging-related information from CT reports to support accurate clinical staging of lung cancer.MethodsIn this study, we developed an information extraction (IE) system to extract the staging-related information from CT reports. The system consisted of the following three parts: named entity recognition (NER), relation classification (RC), and postprocessing (PP). We first summarized 22 questions about lung cancer staging based on the TNM staging guideline. Next, three state-of-the-art NER algorithms were implemented to recognize the entities of interest. Next, we designed a novel RC method using the relation sign constraint (RSC) to classify the relations between entities. Finally, a rule-based PP module was established to obtain the formatted answers using the results of NER and RC.ResultsWe evaluated the developed IE system on a clinical data set containing 392 chest CT reports collected from the Department of Thoracic Surgery II in the Peking University Cancer Hospital. The experimental results showed that the bidirectional encoder representation from transformers (BERT) model outperformed the iterated dilated convolutional neural networks-conditional random field (ID-CNN-CRF) and bidirectional long short-term memory networks-conditional random field (Bi-LSTM-CRF) for NER tasks with macro-F1 scores of 80.97% and 90.06% under the exact and inexact matching schemes, respectively. For the RC task, the proposed RSC showed better performance than the baseline methods. Further, the BERT-RSC model achieved the best performance with a macro-F1 score of 97.13% and a micro-F1 score of 98.37%. Moreover, the rule-based PP module could correctly obtain the formatted results using the extractions of NER and RC, achieving a macro-F1 score of 94.57% and a micro-F1 score of 96.74% for all the 22 questions.ConclusionsWe conclude that the developed IE system can effectively and accurately extract information about lung cancer staging from CT reports. Experimental results show that the extracted results have significant potential for further use in stage verification and prediction to facilitate accurate clinical staging.

Project description:Introduction/backgroundCancer registries in the US collect timely and systematic data on new cancer cases, extent of disease, staging, biomarker status, treatment, survival, and mortality of cancer cases. Existing methodologies for accessing local cancer registry data for research are time-consuming and often rely on the manual merging of data by staff registrars. In addition, existing registries do not provide direct access to these data nor do they routinely provide linkage to discrete electronic health record (EHR) data, reports, or imaging data. Automation of such linkage can provide an impressive data resource and make valuable data available for translational cancer research.MethodsThe UPMC Network Cancer Registry collects highly structured, longitudinal data on all reportable cancer patients, from the point of the diagnosis throughout treatment and follow-up/outcomes. Using commercial registry software, we collect data in compliance with standards governed by the North American Association of Central Cancer Registries. This standardization ensures that the data are highly structured with standard coding and collection methods, which support data exchange among central cancer registries and the Centers for Disease Control and Prevention.ResultsAt the UPMC Hillman Cancer Center and University of Pittsburgh, we explored the feasibility of linking this well-curated, structured cancer registry data with unstructured text (i.e., pathology and radiology reports), using the Text Information Extraction System (TIES). We used the TIES platform to integrate breast cancer cases from the UPMC Network Cancer Registry system and then combine these data with other EHR data as a pilot use case that can be replicated for other cancers.ConclusionsAs a result of this integration, we now have a single searchable repository of information for breast cancer patients from the UPMC registry, combined with their pathology and radiology reports. The system that we developed is easily scalable to other health systems and cancer centers.

Dataset Information

Automatic extraction of cancer registry reportable information from free-text pathology reports using multitask convolutional neural networks.

Publications

Automatic extraction of cancer registry reportable information from free-text pathology reports using multitask convolutional neural networks.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets