Evaluation of Automated Public De-Identification Tools on a Corpus of Radiology Reports

An Extensible Evaluation Framework Applied to Clinical Text Deidentification Natural Language Processing Tools: Multisystem and Multicorpus Study

BACKGROUND

Clinical natural language processing (NLP) researchers need access to directly comparable evaluation results for applications such as text deidentification across a range of corpus types and the means to easily test new systems or corpora within the same framework. Current systems, reported metrics, and the personally identifiable information (PII) categories evaluated are not easily comparable.

OBJECTIVE

This study presents an open-source and extensible end-to-end framework for comparing clinical NLP system performance across corpora even when the annotation categories do not align.

METHODS

As a use case for this framework, we use 6 off-the-shelf text deidentification systems (ie, CliniDeID, deid from PhysioNet, MITRE Identity Scrubber Toolkit [MIST], NeuroNER, National Library of Medicine [NLM] Scrubber, and Philter) across 3 standard clinical text corpora for the task (2 of which are publicly available) and 1 private corpus (all in English), with annotation categories that are not directly analogous. The framework is built on shell scripts that can be extended to include new systems, corpora, and performance metrics. We present this open tool, multiple means for aligning PII categories during evaluation, and our initial timing and performance metric findings. Code for running this framework with all settings needed to run all pairs are available via Codeberg and GitHub.

RESULTS

From this case study, we found large differences in processing speed between systems. The fastest system (ie, MIST) processed an average of 24.57 (SD 26.23) notes per second, while the slowest (ie, CliniDeID) processed an average of 1.00 notes per second. No system uniformly outperformed the others at identifying PII across corpora and categories. Instead, a rich tapestry of performance trade-offs emerged for PII categories. CliniDeID and Philter prioritize recall over precision (with an average recall 6.9 and 11.2 points higher, respectively, for partially matching spans of text matching any PII category), while the other 4 systems consistently have higher precision (with MIST's precision scoring 20.2 points higher, NLM Scrubber scoring 4.4 points higher, NeuroNER scoring 7.2 points higher, and deid scoring 17.1 points higher). The macroaverage recall across corpora for identifying names, one of the more sensitive PII categories, included deid (48.8%) and MIST (66.9%) at the low end and NeuroNER (84.1%), NLM Scrubber (88.1%), and CliniDeID (95.9%) at the high end. A variety of metrics across categories and corpora are reported with a wider variety (eg, F2-score) available via the tool.

CONCLUSIONS

NLP systems in general and deidentification systems and corpora in our use case tend to be evaluated in stand-alone research articles that only include a limited set of comparators. We hold that a single evaluation pipeline across multiple systems and corpora allows for more nuanced comparisons. Our open pipeline should reduce barriers to evaluation and system advancement.

De-identification of clinical free text using natural language processing: A systematic review of current approaches

BACKGROUND

Electronic health records (EHRs) are a valuable resource for data-driven medical research. However, the presence of protected health information (PHI) makes EHRs unsuitable to be shared for research purposes. De-identification, i.e. the process of removing PHI is a critical step in making EHR data accessible. Natural language processing has repeatedly demonstrated its feasibility in automating the de-identification process.

OBJECTIVES

Our study aims to provide systematic evidence on how the de-identification of clinical free text written in English has evolved in the last thirteen years, and to report on the performances and limitations of the current state-of-the-art systems for the English language. In addition, we aim to identify challenges and potential research opportunities in this field.

METHODS

A systematic search in PubMed, Web of Science, and the DBLP was conducted for studies published between January 2010 and February 2023. Titles and abstracts were examined to identify the relevant studies. Selected studies were then analysed in-depth, and information was collected on de-identification methodologies, data sources, and measured performance.

RESULTS

A total of 2125 publications were identified for the title and abstract screening. 69 studies were found to be relevant. Machine learning (37 studies) and hybrid (26 studies) approaches are predominant, while six studies relied only on rules. The majority of the approaches were trained and evaluated on public corpora. The 2014 i2b2/UTHealth corpus is the most frequently used (36 studies), followed by the 2006 i2b2 (18 studies) and 2016 CEGS N-GRID (10 studies) corpora.

CONCLUSION

Earlier de-identification approaches aimed at English were mainly rule and machine learning hybrids with extensive feature engineering and post-processing, while more recent performance improvements are due to feature-inferring recurrent neural networks. Current leading performance is achieved using attention-based neural models. Recent studies report state-of-the-art F1-scores (over 98 %) when evaluated in the manner usually adopted by the clinical natural language processing community. However, their performance needs to be more thoroughly assessed with different measures to judge their reliability to safely de-identify data in a real-world setting. Without additional manually labeled training data, state-of-the-art systems fail to generalise well across a wide range of clinical sub-domains.

Exploring Freely Available Data Tools to Support Open Data and Open Science

Librarians support researchers by promoting open science and open data practices. This article explores five freely available tools that support and facilitate open science practices. Open Science Framework provides a platform for project management, data sharing, and data storage. OpenRefine cleans and formats data. DMPTool has templates for data management and sharing plans that comply with funder mandates. The NIH Common Data Elements is a repository for standardized data elements, and finally, the NLM Scrubber is a tool for de-identifying clinical text data. Information professionals can add these tools to their repertoire and share them with researchers at their institution.

Feasibility of Using the Privacy-preserving Large Language Model Vicuna for Labeling Radiology Reports

Background Large language models (LLMs) such as ChatGPT, though proficient in many text-based tasks, are not suitable for use with radiology reports due to patient privacy constraints. Purpose To test the feasibility of using an alternative LLM (Vicuna-13B) that can be run locally for labeling radiography reports. Materials and Methods Chest radiography reports from the MIMIC-CXR and National Institutes of Health (NIH) data sets were included in this retrospective study. Reports were examined for 13 findings. Outputs reporting the presence or absence of the 13 findings were generated by Vicuna by using a single-step or multistep prompting strategy (prompts 1 and 2, respectively). Agreements between Vicuna outputs and CheXpert and CheXbert labelers were assessed using Fleiss κ. Agreement between Vicuna outputs from three runs under a hyperparameter setting that introduced some randomness (temperature, 0.7) was also assessed. The performance of Vicuna and the labelers was assessed in a subset of 100 NIH reports annotated by a radiologist with use of area under the receiver operating characteristic curve (AUC). Results A total of 3269 reports from the MIMIC-CXR data set (median patient age, 68 years [IQR, 59-79 years]; 161 male patients) and 25 596 reports from the NIH data set (median patient age, 47 years [IQR, 32-58 years]; 1557 male patients) were included. Vicuna outputs with prompt 2 showed, on average, moderate to substantial agreement with the labelers on the MIMIC-CXR (κ median, 0.57 [IQR, 0.45-0.66] with CheXpert and 0.64 [IQR, 0.45-0.68] with CheXbert) and NIH (κ median, 0.52 [IQR, 0.41-0.65] with CheXpert and 0.55 [IQR, 0.41-0.74] with CheXbert) data sets, respectively. Vicuna with prompt 2 performed at par (median AUC, 0.84 [IQR, 0.74-0.93]) with both labelers on nine of 11 findings. Conclusion In this proof-of-concept study, outputs of the LLM Vicuna reporting the presence or absence of 13 findings on chest radiography reports showed moderate to substantial agreement with existing labelers. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Cai in this issue.

Automated deidentification of radiology reports combining transformer and "hide in plain sight" rule-based methods

OBJECTIVE

To develop an automated deidentification pipeline for radiology reports that detect protected health information (PHI) entities and replaces them with realistic surrogates "hiding in plain sight."

MATERIALS AND METHODS

In this retrospective study, 999 chest X-ray and CT reports collected between November 2019 and November 2020 were annotated for PHI at the token level and combined with 3001 X-rays and 2193 medical notes previously labeled, forming a large multi-institutional and cross-domain dataset of 6193 documents. Two radiology test sets, from a known and a new institution, as well as i2b2 2006 and 2014 test sets, served as an evaluation set to estimate model performance and to compare it with previously released deidentification tools. Several PHI detection models were developed based on different training datasets, fine-tuning approaches and data augmentation techniques, and a synthetic PHI generation algorithm. These models were compared using metrics such as precision, recall and F1 score, as well as paired samples Wilcoxon tests.

RESULTS

Our best PHI detection model achieves 97.9 F1 score on radiology reports from a known institution, 99.6 from a new institution, 99.5 on i2b2 2006, and 98.9 on i2b2 2014. On reports from a known institution, it achieves 99.1 recall of detecting the core of each PHI span.

DISCUSSION

Our model outperforms all deidentifiers it was compared to on all test sets as well as human labelers on i2b2 2014 data. It enables accurate and automatic deidentification of radiology reports.

CONCLUSIONS

A transformer-based deidentification pipeline can achieve state-of-the-art performance for deidentifying radiology reports and other medical documents.

Ensemble Approaches to Recognize Protected Health Information in Radiology Reports

Natural language processing (NLP) techniques for electronic health records have shown great potential to improve the quality of medical care. The text of radiology reports frequently constitutes a large fraction of EHR data, and can provide valuable information about patients' diagnoses, medical history, and imaging findings. The lack of a major public repository for radiological reports severely limits the development, testing, and application of new NLP tools. De-identification of protected health information (PHI) presents a major challenge to building such repositories, as many automated tools for de-identification were trained or designed for clinical notes and do not perform sufficiently well to build a public database of radiology reports. We developed and evaluated six ensemble models based on three publically available de-identification tools: MIT de-id, NeuroNER, and Philter. A set of 1023 reports was set aside as the testing partition. Two individuals with medical training annotated the test set for PHI; differences were resolved by consensus. Ensemble methods included simple voting schemes (1-Vote, 2-Votes, and 3-Votes), a decision tree, a naïve Bayesian classifier, and Adaboost boosting. The 1-Vote ensemble achieved recall of 998 / 1043 (95.7%); the 3-Votes ensemble had precision of 1035 / 1043 (99.2%). F1 scores were: 93.4% for the decision tree, 71.2% for the naïve Bayesian classifier, and 87.5% for the boosting method. Basic voting algorithms and machine learning classifiers incorporating the predictions of multiple tools can outperform each tool acting alone in de-identifying radiology reports. Ensemble methods hold substantial potential to improve automated de-identification tools for radiology reports to make such reports more available for research use to improve patient care and outcomes.

Nuclear Medicine and Artificial Intelligence: Best Practices for Algorithm Development

The nuclear medicine field has seen a rapid expansion of academic and commercial interest in developing artificial intelligence (AI) algorithms. Users and developers can avoid some of the pitfalls of AI by recognizing and following best practices in AI algorithm development. In this article, recommendations on technical best practices for developing AI algorithms in nuclear medicine are provided, beginning with general recommendations and then continuing with descriptions of how one might practice these principles for specific topics within nuclear medicine. This report was produced by the AI Task Force of the Society of Nuclear Medicine and Molecular Imaging.

Data Sharing of Imaging in an Evolving Health Care World: Report of the ACR Data Sharing Workgroup, Part 1: Data Ethics of Privacy, Consent, and Anonymization

Radiology is at the forefront of the artificial intelligence transformation of health care across multiple areas, from patient selection to study acquisition to image interpretation. Needing large data sets to develop and train these algorithms, developers enter contractual data sharing agreements involving data derived from health records, usually with postacquisition curation and annotation. In 2019 the ACR convened a Data Sharing Workgroup to develop philosophies around best practices in the sharing of health information. The workgroup identified five broad domains of activity important to collaboration using patient data: privacy, informed consent, standardization of data elements, vendor contracts, and data valuation. This is Part 1 of a Report on the workgroup's efforts in exploring these issues.