PATHOLOGICA 2023;115:318-324;

DOI: 10.32074/1591-951X-952

Original Article

Natural Language Processing to extract

SNOMED-CT codes from pathological reports

Giorgio Cazzaniga1, Albino Eccher2, Enrico Munari3, Stefano Marletta3, Emanuela Bonoldi4,
Vincenzo Della Mea5, Moris Cadei6, Marta Sbaraglia7, Angela Guerriero7, Angelo Paolo Dei Tos7,
Fabio Pagni1, Vincenzo L’Imperio1
1
Department of Medicine and Surgery, Pathology, IRCCS Fondazione San Gerardo dei Tintori, University of
Milano-Bicocca, Italy; 2 Section of Pathology, Department of Medical and Surgical Sciences for Children and
Adults, University of Modena and Reggio Emilia, University Hospital of Modena, Modena, Italy; 3 Department
of Diagnostic and Public Health, Section of Pathology, University of Verona, Verona, Italy; 4 Unit of Surgical
Pathology and Cytogenetics, ASST Grande Ospedale Metropolitano Niguarda, Milan, Italy; 5 Department of
Mathematics, Computer Science and Physics, University of Udine, Udine, Italy; 6 Pathology Unit, ASST Spedali
Civili di Brescia, Brescia, Italy; 7 Surgical Pathology and Cytopathology Unit, Department of Medicine-DIMED,
University of Padua School of Medicine, Padua, Italy

Summary
Objective. The use of standardized structured reports (SSR) and suitable terminologies
like SNOMED-CT can enhance data retrieval and analysis, fostering large-scale studies
and collaboration. However, the still large prevalence of narrative reports in our laboratories
warrants alternative and automated labeling approaches. In this project, natural language
processing (NLP) methods were used to associate SNOMED-CT codes to structured and
unstructured reports from an Italian Digital Pathology Department.
Methods. Two NLP-based automatic coding systems (support vector machine, SVM, and
long-short term memory, LSTM) were trained and applied to a series of narrative reports.
Results. The 1163 cases were tested with both algorithms, showing good performances
Received November 16, 2023 in terms of accuracy, precision, recall, and F1 score, with SVM showing slightly better
Accepted: November 17, 2023 performances as compared to LSTM (0.84, 0.87, 0.83, 0.82 vs 0.83, 0.85, 0.83, 0.82,
respectively). The integration of an explainability allowed identification of terms and groups
Correspondence of words of importance, enabling fine-tuning, balancing semantic meaning and model per-
Vincenzo L’Imperio formance.
Department of Medicine and Surgery, Univer- Conclusions. AI tools allow the automatic SNOMED-CT labeling of the pathology
sity of Milano-Bicocca archives, providing a retrospective fix to the large lack of organization of narrative reports.
IRCCS Fondazione San Gerardo dei Tintori,
via Pergolesi 33, 20900, Monza (Italy) Key words: SNOMED-CT, digital pathology, natural language processing, laboratory
E-mail: [Link]@[Link] information system

How to cite this article: Cazzaniga G, Eccher A,

Munari E, et al. Natural Language Processing to
extract SNOMED-CT codes from pathological
Introduction
reports. Pathologica 2023;115:318-324.
[Link] The progressive digitization of pathology laboratories represents a val-
uable opportunity to promote automation and standardization of the di-
© Copyright by Società Italiana di Anatomia Pato-
agnostic process. This can be perceived through the progressive intro-
logica e Citopatologia Diagnostica, Divisione Itali-
ana della International Academy of Pathology duction of artificial intelligence (AI) algorithms for the most disparate
tasks 1, even in highly sub-specialized fields 2-5, showcasing the incred-
OPEN ACCESS ible potentialities of these new technologies in reducing interobserver
This is an open access journal distributed in accordance
variability and optimizing time-consuming tasks. On the other hand, the
with the CC-BY-NC-ND (Creative Commons Attribution- modifications of the pathology workflow towards a fully tracked and con-
NonCommercial-NoDerivatives 4.0 International) license: the trolled system through the implementation of innovative instruments 6,7
work can be used by mentioning the author and the license,
but only for non-commercial purposes and only in the original and increasing attention to the sample archives 8,9 is significantly twisting
version. For further information: [Link] our routine practice, leading international societies to promote guide-
org/licenses/by-nc-nd/4.0/[Link]
lines and recommendations to guide in this intricate transition 10. The
NLP FOR SNOMED-CT LABELLING OF REPORTS 319

Figure 1. The multiple benefits of labeling diagnosis. Standardization and Structured Data: labeled diagnoses can be con-
sistently categorized, making it easier to analyze and compare data across different cases and institutions. Data Retrieval and
Analysis: with structured data, specific diagnoses can be quickly searched and extracted, allowing pathologists and research-
ers to access relevant cases more easily. Enhanced Research and Collaboration: researchers can perform retrospective stud-
ies by extracting cases with specific diagnosis from digital archives, enabling large-scale analyses and exploration of various
research questions. Automated Data Processing: by having structured labels, it becomes possible to develop algorithms and
machine learning models that can automatically process pathology data. Digital Archives and Data Mining: digital archives
can be organized and indexed based on diagnoses, enabling efficient retrieval and utilization of archived cases. Translation of
Labels to Images: labeled diagnoses can be translated to images, opening avenues for weakly supervised machine learning
projects, even without explicit pixel-level annotations.

introduction of standardized structured reports (SSR) tially integrable within our laboratory information sys-
represents another milestone of the digital pathology tems (LIS) 16.
transition in the precision medicine era, either full 11 or The identification of robust supports for data mining
partial 12,13,14. Waiting for the progressive adoption of and curation may be the turning point for building da-
SSR in our routine practice, the application of natural ta lakes projects in pathology, as a test possibly ex-
language processing (NLP) methods for the automat- pandable to many other hospital departments. Here
ed labeling of retrospective series with Systematized
we investigate whether an automated NLP approach
Nomenclature of Medicine Clinical Term (SNOMED-
for the SNOMED-CT code labeling on a prospective
CT) is a promising alternative to untap the goldmine
of the digital pathology archives (Fig. 1). The retrieval cohort of structured and unstructured reports from an
of this information holds the potential for conducting Italian Digital Pathology Department with two different
high-quality analyses and research, aiming to en- AI algorithms can optimize the processing of highly
hance standards of care and diagnosis, ultimately specialized languages by developing a scalable, light-
leading to improved patient outcomes 15, being poten- weight, and efficient model.
320 G. Cazzaniga et al.

Materials and methods Explainability

Explainability was incorporated into the models using
the Local Interpretable Model-agnostic Explanations
Data extraction and filtering
(LIME) library. This approach allowed the identifica-
Narrative and structured reports, along with their cor- tion of the most relevant terms associated with each
responding SNOMED-3 codes, were extracted from SNOMED-CT code category. Before the analysis with
the LIS Athena (Version 4.3.0, Dedalus, Florence, Ita- LIME library, words such as “eventual”, “useful”, “evalu-
ly) at the Pathology Department of Fondazione IRCCS ation”, “confirm”, “appears”, etc. were therefore gradually
San Gerardo dei Tintori, University of Milano-Bicocca, eliminated during the preprocessing phase to achieve
Monza, Italy for cases diagnosed from January 2023 a smoother workflow and avoid classification errors
to November 2023. The dataset underwent a series of stemming from words that are semantically important
filtering steps aimed to clean it from cases lacking a but not medically/diagnostically relevant. By examin-
SNOMED code and those containing multiple lesions ing the explanations provided by LIME, insights were
within the same diagnosis text resulting in multiple gained into the features and patterns that influenced
coding. Specifically, only diagnosis (D) and morphol- the model’s predictions.
ogy (M) codes were retained and appropriately trans-
lated to SNOMED-CT codes.

Ad hoc ontology Results

Ad hoc ontology was constructed based on the Inter-
national Classification of Diseases SNOMED-CT. In Cases
instances where a report contained multiple morpho- A total of 1163 reports from 2023 were retrieved and
logical codes, we selected the one that indicated more used to test the two different NLP algorithms. The
advanced disease progression or held greater clinical translation of SNOMED-3 codes to SNOMED CT re-
significance as the primary representation. Converse- sulted in a total of 46 ‘Morphologic abnormality’ codes,
ly, cases where multiple lesions with equivalent poten- 1 ‘Disorder’ code, 2 ‘Finding’ codes and 1 ‘Quantitative’
tial for malignancy or not linked by a single pathoge- code. The most frequent codes were associated with
netic process were excluded from the analysis. dermatopathology and gastroenterology, as well as
more generic codes such as “Normal tissue morphol-
Preprocessing ogy” or “chronic inflammation”, reflecting the case fre-
The dataset underwent a series of preprocessing quencies in the Department (Supplementary Tab. I).
steps to prepare it for testing on the models trained The restriction of the dataset to the most 50 used
as demonstrated in the following repository: https:// codes allowed us to work on codes where we had a
[Link]/Gizmopath/SVM-LSTM-models-for-auto- substantial sample size still being able to cover 81%
mated-SNOMED-CT-labeling. This process included of the initial dataset. The choice to narrow down the
retaining the 50 most frequent codes along with their dataset to the top 50 cases was made to strike a bal-
corresponding diagnoses. Outliers, such as unusual- ance between maintaining a high dataset coverage
ly lengthy diagnosis texts exceeding 750 characters, percentage and ensuring consistent accuracy results.
were removed. Additionally, class balancing tech- Expanding the dataset to 70 or 100 cases would have
niques were implemented to address potential biases increased the coverage from 81% to 86% and just un-
in the data, ensuring equal representation of different der 90%, respectively (Fig. 2).
classes. Stopwords, special characters, and punctua-
tion marks were removed and text was converted into NLP models performances
lowercase. Italian lemmatization and tokenization pro- The comparison of the performances obtained with
cesses were also performed. LSTM and SVM models on the prospective 2023 co-
hort test set is reported in Table I. Although not reach-
Natural language processing models performances ing statistical significance, the SVM model showed
Two different models, long short-term memory (LSTM) similar or even better results as compared to the
and support vector machine (SVM), were tested for LSTM one in terms of accuracy, precision, recall and
their ability to assign the correct SNOMED-CT code F1 score. The code assigned with worst performance
to the relative diagnostic report. The models’ perfor- was M-09410 (in SNOMED-CT: ‘Negative for residual
mances were evaluated keeping track of accuracy, tumor (Finding)’, F1 score: 0.26). This is related to a
precision, recall and F1 score as metrics, providing common difficulty in NLP when dealing with negations,
insights into their predictive capabilities. which needs to be treated with specific techniques.
NLP FOR SNOMED-CT LABELLING OF REPORTS 321

Figure 2. The restriction of the dataset to the most 50 used codes allowed us to work on codes where we had a substantial
sample size still being able to cover 81% of the initial dataset. The original number of unique codes in our dataset was 1251,
representing about 1.4% of the sum of all morphological abnormality (M-) and disease (D-) SNOMED-3 codes.

Explainability AI for SNOMED-keywords reasons for the missed ones as demonstrated in the
correspondence examples reported in Figure 3. In particular, with-
in different semantic domains (e.g. breast, skin and
The application of explainability algorithms for the de-
general pathology) the NLP algorithms demonstrated
tection of human interpretable features within the nar- fluctuant accuracy in predicting the correct SNOMED-
rative report certified the genuinity of the SNOMED- CT based on the availability of sufficient keywords for
CT prediction by the NLP approach, highlighting the the correct assignment. As a result, cases of atypi-
cal fibroadenomatoid proliferations of the breast were
correctly (Atypia, suspicious for malignancy) or in-
Table I. Comparison of metrics achievement between LSTM correctly (Benign fibroadenoma) classified due to the
and SVM models. SVM outperformed LSTM achieving bet- presence/absence of concurrent keywords such as
ter results. LSTM, long short-term memory; SVM, support “atypical” or “unknown malignant potential”. The same
vector machine; SD, standard deviation; 95%CI, 95% con- was noted for in situ squamous lesions of the skin,
fidence interval. where the term “Bowen” complicated the assignment
LSTM SVM p-value task, and in inflammatory conditions where terms like
Accuracy 0.83 0.84 “acute” or “chronic” were key for the correct SNOMED-
Precision 0.85 (0.21) 0.87 (0.22) 0.64 CT assignment.
mean (SD) [0.19-1] [0.20-1]
[95%CI]
Recall 0.83 (0.26) 0.83 (0.21) 1
mean (SD) [0.15 -1] [0.21-1] Discussion
[95%CI]
F1-Score 0.82 (0.26) 0.82 (0.23) 1 One of the major challenges that our discipline is
mean (SD) [0.15-1] [0.21-1] facing within the digital pathology revolution is linked
[95%CI] to the archiving phase, and specifically for reports,
322 G. Cazzaniga et al.

Figure 3. Examples of correct (left column) and missed (right column) SNOMED-CT prediction in different domains (breast,
skin and surgical/general pathology are reported here). Each box reports the most probable SNOMED-CT code attributable to
the case by the NLP model, the predicted code and the ground truth, the most relevant keywords used by the model to predict
the code and examples of the narrative text with most relevant keywords highlighted in progressively dark colors.

whose standardized and systematic organization information extraction process. Finally, the complexi-
would allow efficient retrieval and storage of patient ty that our reports are reaching through the integra-
information 17. The progressive prospective adoption tion of molecular and genomic big data incorporation
of SSR, eventually associated with NLP able of ex- represent a further source of information overload,
tracting granular information from structured and requiring a strong standardization effort (e.g. with
narrative reports, will probably solve this issue when health informatics standards as CDA or FHIR) ensur-
adequately integrated within our LIS and widely em- ing that the data are computer identifiable, retrievable
ployed in our departments. However, while SSR are and processable. In this direction, the Italian Society
largely available and applicable in the oncological of Pathology (SIAPEC) is working on the creation of
setting 18,19, their application in other pathology fields a standard dataset for reporting, moving from free-
(e.g. inflammatory conditions) is still delayed, due to text narrative reports (level 1 according to the Ontario
the higher variability and complexity of these reports scale)22 to synoptic reports (level 3) to fully structured
and to the lack of standardization 18 20,21. Moreover, the reporting, which should include discrete data embed-
extraction of specific features from structured reports ded in LIS and structured messaging/data exchange
through NLP can be computationally demanding. To standards (level 6), culminating in the creation of
overcome this challenge, restricting the research to common and shared data lake 23. Meanwhile, our ar-
smaller disease groups through appropriate coding chives are already hosting a large quantity of narra-
(e.g. SNOMED-CT) can prospectively facilitate the tive reports whose retrieval is often inaccurate or com-
NLP FOR SNOMED-CT LABELLING OF REPORTS 323

pletely lacking due to their intrinsic “analog” nature. and reliable future dataset on which these algorithms
In this setting, the employment of SNOMED-CT can can be fine-tuned 23.
represent a solution for labeling cases to help in their
retrospective retrieval and organization 24,25. Although
limited by the partial adoption of SNOMED-CT among Conclusions
departments and its need for progressive updates
(e.g. for molecular and omic terms) 26,27, leveraging The application of AI tools allows the automatic
on these codes can potentially help us in re-organiz- SNOMED-CT labeling of the pathology archives, pro-
ing the retrospective pathology archives 8,28,29. In this viding a retrospective fix to the large lack of organiza-
direction, the application of rule-based approaches tion of narrative reports. This rescue solution can pave
have already been proposed to overcome the intrinsic the way a the combined NLP and SSR prospective
limits (time-consuming process, prone to human error approach through adequate LIS implementation for
and interobserver variability) of manual SNOMED- improving patients’ care.
CT coding 30-32. The present study demonstrates that
the application of SVM and LSTM AI tools can be a Acknowledgements
valid alternative to manual rule engineering, showing We gratefully acknowledge Giuseppe Barletta for his
proficiency in capturing complex patterns that sim- contributions to this article, including the creation of
pler techniques may overlook, improving automation the SVM model and valuable IT support.
in the setting of NLP, and providing a temporary fix
while SSR are more widely adopted in our depart- Conflicts of interest
ments. Moreover, interpretability analysis highlighted Authors do not have any conflict of interest to disclose.
the impact of specific words of the pathology report
on the final decision. In particular, this visual analysis Funding
demonstrated the need for additional keywords that The present work has been funded by the Italian Min-
may contribute to the final assignment of a SNOMED- istry of the University MUR Dipartimenti di Eccellenza
CT code together with the “defining” words of particu- 2023-2027 (l. 232/2016, art. 1, commi 314 - 337).
lar entities. In this setting, emblematic is the example
of breast pathology domain, where the coexistence of Authors’ contributions
uncertainty keywords (e.g. “atypical” or “unknown ma- GC defined the study design, performed the retro-
lignant potential”), along with specific terms indicating spective research and data extraction, as well as the
the disease (e.g. “fibroadenomatous”), were essential elaboration of the computational pipeline for the study;
to avoid misassignment of the code (benign, fibroad- APDT, MS, AG, AE, VDM, EM, MC, SM and EB pro-
enoma), allowing the correct identification of the case vided counseling as experts in the nomenclature and
(Atypia, suspicious for malignancy). Finally, the pres- SNOMED field; FP and VL performed the supervision
ence of narrative/descriptive reports can be a source of the work revising critically the manuscript before
of variability in the final code assignment, as docu- the approval by all the authors; FP provided the fund-
mented by the example reported in acute vs chronic ing acquisition and administrative support. All authors
inflammation and atrophy. The proposed NLP-based were involved in writing the paper and had final ap-
methods can thus represent a valid supporting tool for proval of the submitted and published versions.
the retrospective labeling of narrative reports, waiting
for a more comprehensive implementation of tandem Ethical consideration
SSR/NLP approach in our LIS. However, the further The research was conducted ethically, with all study
expansion of alternative language/vocabularies 33, the
procedures being performed in accordance with the
fine-tuning of SVM to account variations across lan-
requirements of the World Medical Association’s Dec-
guages 34, the extension to more “rare” codes that goes
laration of Helsinki.
beyond the 50 tested here can help to overcome the
limits of generalizability with rarer diagnosis, improv-
ing this tool for the actual employment by researchers References
and pathologists 35. Challenges faced by SNOMED 1
Caputo A, L’Imperio V, Merolla F, et al. The slow-paced digi-
CT in its widespread use, such as redundant codes, tal evolution of pathology: lights and shadows from a mul-
tifaceted board. Pathologica 2023;115:127-136. [Link]
manual errors, and difficulties in creating user-friendly org/10.32074/1591-951X-868
labeling modules, also impact the dataset’s reliabili- 2
Cazzaniga G, Rossi M, Eccher A, et al. Time for a full digital ap-
ty.36 The future implementation of automatic coding proach in nephropathology: a systematic review of current arti-
through LIS-integrated SSR promises a more robust ficial intelligence applications and future directions. J Nephrol
324 G. Cazzaniga et al.

Published online September 28, 2023. [Link] dence-based datasets. Virchows Arch 2016;468:51-59. [Link]
s40620-023-01775-w org/10.1007/s00428-015-1834-4
3
Cazzaniga G, Bolognesi MM, Stefania MD, et al. Congo Red 19
Hewer E. The Oncologist’s Guide to Synoptic Reporting: A Primer.
Staining in Digital Pathology: The Streamlined Pipeline for Oncology 2020;98:396-402. [Link]
Amyloid Detection Through Congo Red Fluorescence Digital 20
Leh S, Dendooven A. Systematic reporting of medical kidney bi-
Analysis. Lab Invest 2023;103:100243. [Link]
opsies. Clin Kidney J 2021;15:21-30. [Link]
labinv.2023.100243
sfab140
4
L’Imperio V, Wulczyn E, Plass M, et al. Pathologist Validation of 21
Langford CR, Goldinger MH, Treanor D, et al. Improved pathol-
a Machine Learning-Derived Feature for Colon Cancer Risk
Stratification. JAMA Netw Open 2023;6:e2254891. [Link] ogy reporting in NAFLD/NASH for clinical trials. J Clin Pathol
org/10.1001/jamanetworkopen.2022.54891 2022;75:73-75.
22
5
Marletta S, L’Imperio V, Eccher A, et al. Artificial intelligence-based Srigley JR, McGowan T, Maclean A, et al. Standardized synoptic
tools applied to pathological diagnosis of microbiological diseas- cancer pathology reporting: a population-based approach. J Surg
es. Pathol Res Pract 2023;243:154362. [Link] Oncol 2009;99:517-524. [Link]
prp.2023.154362. 23
Renshaw AA, Mena-Allauca M, Gould EW, et al. Synoptic Report-
6
L’Imperio V, Gibilisco F, Fraggetta F. What is Essential is (No More) ing: Evidence-Based Review and Future Directions. CO Clin Can-
Invisible to the Eyes: The Introduction of BlocDoc in the Digital cer Inform 2018;2. [Link]
Pathology Workflow. J Pathol Inform 2021;12:32. [Link] 24
Paskal W, Paskal AM, Dębski T, Gryziak M, Jaworowski J. Aspects
org/10.4103/jpi.jpi_35_21 of Modern Biobank Activity - Comprehensive Review. Pathol Oncol
7
Munari E, Scarpa A, Cima L, et al. Cutting-edge technol- Res 2018;24:771-785. [Link]
ogy and automation in the pathology laboratory. Virchows Arch 25
Ali M, Evans H, Whitney P, Minhas F, Snead DRJ. Using Sys-
Published online November 6, 2023. [Link] temised Nomenclature of Medicine (SNOMED) codes to select
s00428-023-03637-z
digital pathology whole slide images for long-term archiving. J Clin
8
Eccher A, Dei Tos AP, Scarpa A, et al. Cost analysis of archives Pathol 2023;76:349-352. [Link]
in the pathology laboratories: from safety to management. J Clin 26
Ceusters W. SNOMED CT revisions and coded data repositories:
Pathol. Published online August 2, 2023. [Link]
when to upgrade? AMIA Annu Symp Proc 2011;2011:197-206.
jcp-2023-209035
27
9
Eccher A, Scarpa A, Dei Tos AP. Impact of a centralized archive Jiang X. Intelligent Classification Method of Archive Data
for pathology laboratories on the health system. Pathol Res Pract Based on Multigranular Semantics. Comput Intell Neurosci
2023;245:154488. [Link] 2022;2022:7559523. [Link]
28
10
Fraggetta F, L’Imperio V, Ameisen D, et al. Best Practice Recom- Richesson RL, Andrews JE, Krischer JP. Use of SNOMED CT to
mendations for the Implementation of a Digital Pathology Work- represent clinical research data: a semantic characterization of
flow in the Anatomic Pathology Laboratory by the European Soci- data items on case report forms in vasculitis research. J Am Med
ety of Digital and Integrative Pathology (ESDIP). Diagnostics (Ba- Inform Assoc 2006;13:536-546. [Link]
sel) 2021;11:2167. [Link] M2093
29
11
Fraggetta F, Caputo A, Guglielmino R, et al. A Survival Guide Cornet R, de Keizer N. Forty years of SNOMED: a literature re-
for the Rapid Transition to a Fully Digital Workflow: The “Calta- view. BMC Med Inform Decis Mak 2008;8 Suppl 1(Suppl 1):S2.
girone Example.” Diagnostics (Basel) 2021;11:1916. [Link] [Link]
org/10.3390/diagnostics11101916 30
Carter KJ, Rinehart S, Kessler E, et al. Quality assurance
12
L’Imperio V, Casati G, Cazzaniga G, et al. Improvements in digi- in anatomic pathology: automated SNOMED coding. J Am
tal pathology equipment for renal biopsies: updating the standard Med Inform Assoc 1996;3:270-272. [Link]
model. J Nephrol. Published online February 14, 2023. [Link] jamia.1996.96413134
org/10.1007/s40620-023-01568-1 31
García-Rojo M, Daniel C, Laurinavicius A. SNOMED CT in pathol-
13
L’Imperio V, Brambilla V, Cazzaniga G, Ferrario F, Nebuloni M, ogy. Stud Health Technol Inform 2012;179:123-140.
Pagni F. Digital pathology for the routine diagnosis of renal dis- 32
Skeppstedt M, Kvist M, Dalianis H. Rule-based Entity Recognition
eases: a standard model. J Nephrol 2021;34:681-688. [Link]
org/10.1007/s40620-020-00805-1. and Coverage of SNOMED CT in Swedish Clinical Text. In: LREC.;
14
2012:1250-1257.
Pallua JD, Brunner A, Zelger B, Schirmer M, Haybaeck J. The fu- 33
ture of pathology is digital. Pathol Res Pract 2020;216:153040. Millar J. The Need for a Global Language - SNOMED CT Introduc-
[Link] tion. Stud Health Technol Inform. 2016;225:683-685.
34
15
Snoek JAA, Nagtegaal ID, Siesling S, van den Broek E, van Sloo- Schulz S, Hammer L, Nik DH, Kreuzthaler M. Localising the Clini-
ten HJ, Hugen N. The impact of standardized structured reporting cal Terminology SNOMED CT by Semi-automated Creation of a
of pathology reports for breast cancer care. Breast 2022;66:178- German Interface Vocabulary. MULTILINGUALBIO. Published on-
182. [Link] line 2020. Accessed June 4, 2023. [Link]
16
Soysal E, Warner JL, Wang J, et al. Developing Customizable org/paper/92fa8856652d1a7cf2d4781da45e4829fd0a2fd6
35
Cancer Information Extraction Modules for Pathology Reports Amin M, Dhir R. Data Representation, Coding, and Communica-
Using CLAMP. Stud Health Technol Inform 2019;264:1041-1045. tion Standards. Surg Pathol Clin 2015;8:[Link] M, Dhir R.
[Link] Data Representation, Coding, and Communication Standards.
17
Gutman DA, Khalilia M, Lee S, et al. The Digital Slide Archive: Surg Pathol Clin 2015;8:109-121.
A Software Platform for Management, Integration, and Analysis 36
Deeken-Draisey A, Ritchie A, Yang GY, et al. Current Proce-
of Histology for Cancer Research. Cancer Res 2017;77:e75-e78. dural Terminology Coding for Surgical Pathology: A Review and
[Link] One Academic Center’s Experience With Pathologist-Verified
18
Ellis DW, Srigley J. Does standardised structured reporting con- Coding. Arch Pathol Lab Med 2018;142:1524-1532. [Link]
tribute to quality in diagnostic pathology? The importance of evi- org/10.5858/arpa.2017-0190-RA