Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Mar 30;54(1):e2035300.
doi: 10.25100/cm.v54i1.5300. eCollection 2023 Jan-Mar.

Automated extraction of information from free text of Spanish oncology pathology reports

Diana Marcela Mendoza-Urbano  1 , Johan Felipe Garcia  2 , Juan Sebastian Moreno  2   3 , Juan Carlos Bravo-Ocaña  4 , Alvaro José Riascos  2   3   5 , Angela Zambrano Harvey  6 , Sergio I Prada  7   8
Affiliations

Automated extraction of information from free text of Spanish oncology pathology reports

Diana Marcela Mendoza-Urbano et al. Colomb Med (Cali). .

Abstract
in English, Spanish

Background: Pathology reports are stored as unstructured, ungrammatical, fragmented, and abbreviated free text with linguistic variability among pathologists. For this reason, tumor information extraction requires a significant human effort. Recording data in an efficient and high-quality format is essential in implementing and establishing a hospital-based-cancer registry.

Objective: This study aimed to describe implementing a natural language processing algorithm for oncology pathology reports.

Methods: An algorithm was developed to process oncology pathology reports in Spanish to extract 20 medical descriptors. The approach is based on the successive coincidence of regular expressions.

Results: The validation was performed with 140 pathological reports. The topography identification was performed manually by humans and the algorithm in all reports. The human identified morphology in 138 reports and by the algorithm in 137. The average fuzzy matching score was 68.3 for Topography and 89.5 for Morphology.

Conclusions: A preliminary algorithm validation against human extraction was performed over a small set of reports with satisfactory results. This shows that a regular-expression approach can accurately and precisely extract multiple specimen attributes from free-text Spanish pathology reports. Additionally, we developed a website to facilitate collaborative validation at a larger scale which may be helpful for future research on the subject.

Introducción: Los reportes de patología están almacenados como texto libre sin estructura, gramática, fragmentados o abreviados, con variabilidad lingüística entre patólogos. Por esta razón, la extracción de información de tumores requiere un esfuerzo humano significativo. Almacenar información en un formato eficiente y de alta calidad es esencial para implementar y establecer un registro hospitalario de cáncer.

Objetivo: Este estudio busca describir la implementación de un algoritmo de Procesamiento de Lenguaje Natural para reportes de patología oncológica.

Métodos: Desarrollamos un algoritmo para procesar reportes de patología oncológica en Español, con el objetivo de extraer 20 descriptores médicos. El abordaje se basa en la coincidencia sucesiva de expresiones regulares.

Resultados: La validación se hizo con 140 reportes de patología. La identificación topográfica se realizó por humanos y por el algoritmo en todos los reportes. La morfología fue identificada por humanos en 138 reportes y por el algoritmo en 137. El valor de coincidencias parciales (fuzzy matches) promedio fue de 68.3 para Topografía y 89.5 para Morfología.

Conclusiones: Se hizo una validación preliminar del algoritmo contra extracción humana sobre un pequeño grupo de reportes, con resultados satisfactorios. Esto muestra que múltiples atributos del espécimen pueden ser extraídos de manera precisa de texto libre de reportes de patología en Español, usando un abordaje de expresiones regulares. Adicionalmente, desarrollamos una página web para facilitar la validación colaborativa a gran escala, lo que puede ser beneficioso para futuras investigaciones en el tema.

Keywords: National Program of Cancer Registries; algorithm; artificial intelligence; cancer pathology reports; data science; ontology learning; regular expressions.

PubMed Disclaimer

Conflict of interest statement

Conflict of interests: authors declare no conflict of interest.

Figures

Figure 1
Figure 1. Algorithm: the figure shows the process followed to identify and retrieve the relevant characteristics of the oncology pathology report. The algorithm is feed with three types of data: microscopic, macroscopic and diagnosis data. It then follows a four step process in which the data is sorted (step 1), characteristics are identified inside the text (step 2) and finally, they are retrieved (step 3) and parsed or tokenized into grammatical parts (step 4).
Figure 2
Figure 2. Confusion matrices between human and algorithmic extraction for the nonapplicable (NA), non-reported (NR) and reported (R) values in the special descriptors. The fill colour indicates the contribution of each entry to the f-score.
Figura 1
Figura 1. Algoritmo: la figura muestra el proceso aplicado para identificar y recuperar las características relevantes del reporte de patología oncológico. El algoritmo se alimenta de tres tipos de datos: microscópico, macroscópico y datos de diagnóstico. Luego, sigue un proceso de cuatro pasos en el que los datos se sortean (paso 1), luego se identifican las características en el texto (paso 2) para finalmente ser traídos (paso 3) y analizados o “monetizados” en partes gramaticales (paso 4).
Figura 2
Figura 2. Matrices de confusión entre extracción humana y algorítmica para los valores No aplicable (NA), no reportado (NR) y reportado (R) en los descriptores especiales. El color lleno indica la contribución de cada entrada al puntaje f.
See this image and copyright information in PMC

References

    1. Ruiz A, Facio Á. Hospital-based cancer registry A tool for patient care, management and quality. A focus on its use for quality assessment. Rev Oncol. 2004;6(2):104–113. doi: 10.1007/BF02710038. - DOI
    1. Kreimeyer K, Foster M, Pandey A, Arya N, Halford G, Jones SF. Natural language processing systems for capturing and standardizing unstructured clinical information A systematic review. J Biomed Inform. 2017;73:14–29. doi: 10.1016/j.jbi.2017.07.012. - DOI - PMC - PubMed
    1. Alawad M, Gao S, Qiu JX, Yoon HJ, Blair Christian J, Penberthy L. Automatic extraction of cancer registry reportable information from free-text pathology reports using multitask convolutional neural networks. J Am Med Informatics Assoc. 2020;27(1):89–98. doi: 10.1093/jamia/ocz153. - DOI - PMC - PubMed
    1. Nadkarni PM, Ohno-Machado L, Chapman WW. Natural language processing an introduction. J Am Med Inform Assoc. 2011;18(5):544–551. doi: 10.1136/amiajnl-2011-000464. - DOI - PMC - PubMed
    1. Meystre S, Savova G, Kipper-Schuler KC, Hurdle JF. Extracting information from textual documents in the electronic health record: a review of recent research. Yearb Med Inf. 2007:128–144. - PubMed

LinkOut - more resources