T-staging pulmonary oncology from radiological reports using natural language processing: translating into a multi-language setting

Abstract

Background In the era of datafication, it is important that medical data are accurate and structured for multiple applications. Especially data for oncological staging need to be accurate to stage and treat a patient, as well as population-level surveillance and outcome assessment. To support data extraction from free-text radiological reports, Dutch natural language processing (NLP) algorithm was built to quantify T-stage of pulmonary tumors according to the tumor node metastasis (TNM) classification. This structuring tool was translated and validated on English radiological free-text reports. A rule-based algorithm to classify T-stage was trained and validated on, respectively, 200 and 225 English free-text radiological reports from diagnostic computed tomography (CT) obtained for staging of patients with lung cancer. The automated T-stage extracted by the algorithm from the report was compared to manual staging. A graphical user interface was built for training purposes to visualize the results of the algorithm by highlighting the extracted concepts and its modifying context. Results Accuracy of the T-stage classifier was 0.89 in the validation set, 0.84 when considering the T-substages, and 0.76 when only considering tumor size. Results were comparable with the Dutch results (respectively, 0.88, 0.89 and 0.79). Most errors were made due to ambiguity issues that could not be solved by the rule-based nature of the algorithm. Conclusions NLP can be successfully applied for staging lung cancer from free-text radiological reports in different languages. Focused introduction of machine learning should be introduced in a hybrid approach to improve performance.

Knowledge Graph and Deep Learning-based Text-to-GraphQL Model for Intelligent Medical Consultation Chatbot

“…The pre-trained language model (PLM) is driven by a large amount of corpus and can use these data to realize the semantic representation of knowledge contained in a large amount of text to realize downstream tasks. The downstream tasks include natural language processing tasks such as classification (Li et al, 2019b ; Maltoudoglou et al, 2022 ; Ni et al, 2020a , 2020b ), sequence labeling (Dai et al, 2019 ; Li et al, 2020b ), summarization (Chintagunta et al, 2021 ; Lacson et al, 2006 ; Yuan et al, 2021 ), translation (Névéol et al, 2018 ; Nobel et al, 2021 ; Wang et al, 2019 ), generation (Melamud & Shivade, 2019 ; Peng et al, 2019 ; Xiong et al, 2019 ), etc. As one of the new downstream tasks, the translation task, Zhu et al ( 2020 ) previously found that using the pre-trained language model as contextual embedding instead of direct fine-tuning will produce better results.…”

How Natural Language Processing Can Aid With Pulmonary Oncology Tumor Node Metastasis Staging From Free-Text Radiology Reports: Algorithm Development and Validation

“…For both the training and the validation sets, the substage accuracy scores were calculated separately for the T-stage and the N-stage. T-substage is a subdivision of the T-stage to provide more detail, for example, stage T1 (≤3 cm) contains substage T1c (2 to ≤3 cm) [ 1 , 18 ]. Next to the T- and N-stage, the combined accuracy score (TN-stage) was scored for the training and validation sets.…”

“…NLP has also been used in a recent and ongoing transnational project to extract the stage in pulmonary oncology from free-text radiological chest CT scan reports [ 17 , 18 ]. The overall goal is to build a language-independent algorithm that can extract pulmonary oncology staging according to the TNM classification.…”

“…In prior work, a rule-based NLP algorithm was trained and validated on Dutch radiological reports before it was translated and validated on English reports, which showed an accuracy rate for T-stage ranging between 0.84 and 0.87 [ 18 ] The rule-based approach is thought to be the easiest way to accurately determine the oncological stage, as TNM is already a rule-based system. When, for instance, only machine learning (ML) strategies for staging were used, apart from the issue of correctly finding the specific concepts, the algorithm also needs to extract the set of rules of each concept from the training data, which requires a very large amount of data.…”

Year 2021: COVID-19, Information Extraction and BERTization among the Hottest Topics in Medical Natural Language Processing

“…We can observe that, in 2021, researchers mainly concentrated on studying English-language data. Indeed, compared to previous years, a fewer number of languages were covered: Chinese [3][4][5][6][7][8][9][10], Dutch [11], French [12,13], Italian [14][15][16], Japanese [17], Korean [18,19], Norwegian [20], and Spanish . Besides, except for Chinese, there were also very few works done for the languages represented in publications.…”