BACKGROUND: In oncology, electronic health records contain textual key information for the diagnosis, staging, and treatment planning of patients with cancer. However, text data processing requires a lot of time and effort, which limits the utilization of these data. Recent advances in natural language processing (NLP) technology, including large language models, can be applied to cancer research. Particularly, extracting the information required for the pathological stage from surgical pathology reports can be utilized to update cancer staging according to the latest cancer staging guidelines.
OBJECTIVE: This study has two main objectives. The first objective is to evaluate the performance of extracting information from text-based surgical pathology reports and determining pathological stages based on the extracted information using fine-tuned generative language models (GLMs) for patients with lung cancer. The second objective is to determine the feasibility of utilizing relatively small GLMs for information extraction in a resource-constrained computing environment.
METHODS: Lung cancer surgical pathology reports were collected from the Common Data Model database of Seoul National University Bundang Hospital (SNUBH), a tertiary hospital in Korea. We selected 42 descriptors necessary for tumor-node (TN) classification based on these reports and created a gold standard with validation by two clinical experts. The pathology reports and gold standard were used to generate prompt-response pairs for training and evaluating GLMs which then were used to extract information required for staging from pathology reports.
RESULTS: We evaluated the information extraction performance of six trained models as well as their performance in TN classification using the extracted information. The Deductive Mistral-7B model, which was pre-trained with the deductive dataset, showed the best performance overall, with an exact match ratio of 92.24% in the information extraction problem and an accuracy of 0.9876 (predicting T and N classification concurrently) in classification.
CONCLUSIONS: This study demonstrated that training GLMs with deductive datasets can improve information extraction performance, and GLMs with a relatively small number of parameters at approximately seven billion can achieve high performance in this problem. The proposed GLM-based information extraction method is expected to be useful in clinical decision-making support, lung cancer staging and research.

暂无摘要。

This paper describes the development of Health Level Seven Fast Healthcare Interoperability Resource (FHIR) profiles for pathology reports integrated with whole slide images and clinical data to create a pathology research database. A report template was designed to collect structured reports, enabling pathologists to select structured terms based on a checklist, allowing for the standardization of terms used to describe tumor features. We gathered and analyzed 190 non-small-cell lung cancer pathology reports in free text format, which were then structured by mapping the itemized vocabulary to FHIR observation resources, using international standard terminologies, such as the International Classification of Diseases, LOINC, and SNOMED CT. The resulting FHIR profiles were published as an implementation guide, which includes 25 profiles for essential data elements, value sets, and structured definitions for integrating clinical data and pathology images associated with the pathology report. These profiles enable the exchange of structured data between systems and facilitate the integration of pathology data into electronic health records, which can improve the quality of care for patients with cancer.

Deep learning applied to whole-slide histopathology images (WSIs) has the potential to enhance precision oncology and alleviate the workload of experts. However, developing these models necessitates large amounts of data with ground truth labels, which can be both time-consuming and expensive to obtain. Pathology reports are typically unstructured or poorly structured texts, and efforts to implement structured reporting templates have been unsuccessful, as these efforts lead to perceived extra workload. In this study, we hypothesised that large language models (LLMs), such as the generative pre-trained transformer 4 (GPT-4), can extract structured data from unstructured plain language reports using a zero-shot approach without requiring any re-training. We tested this hypothesis by utilising GPT-4 to extract information from histopathological reports, focusing on two extensive sets of pathology reports for colorectal cancer and glioblastoma. We found a high concordance between LLM-generated structured data and human-generated structured data. Consequently, LLMs could potentially be employed routinely to extract ground truth data for machine learning from unstructured pathology reports in the future. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

UNASSIGNED: To evaluate the diagnostic mismatch (discrepancy) of pathology reports in consulted specimens referred for second opinion.
UNASSIGNED: This cross-sectional study was conducted at a major cancer center, Omid Hospital. In this study, 350 primary pathology reports and 350 reviewed pathology reports were extracted from the archives of Omid Hospital from 2011 to 2020 and assessed in terms of the extent of discrepancy, by two pathologists and one oncologist. The required data for each sample were entered into a checklist and then statistically analyzed. Cases with the same diagnosis on both reports were assigned to the matched group and the rest were assigned to the minor or major mismatch (discrepancy) group. Minor mismatches included changes in diagnosis that did not lead to changes in treatment (may lead to changes in prognosis or provide additional information to the oncologist) and major mismatches included changes in diagnosis leading to changes in treatment or remedies.
UNASSIGNED: Two hundred seven cases (59.1%) out of three hundred fifty cases had concordant results between the diagnosis of the first pathologist and the reviewing pathologist. In one hundred forty-three cases (40.9%) mismatch (discrepancy) was observed, including eighty- two cases (23.4%) with minor mismatches (discrepancy) and sixty-one cases (17.4%) with major mismatches (discrepancy). In the major mismatch group, fifteen cases (4.3%) changed from malignant to benign, eighteen cases (5.1%) changed from benign to malignant, two cases (0.6%) changed from one stage to another stage of Disease and twenty-six cases (7.4%) had changes in the type of malignancy. In this study, it was found that there was no significant relationship between anatomical areas of sampling and diagnostic mismatch (p = 0.254). The study also found that the rate of diagnostic mismatch in specimens obtained by resection or excisional biopsy was greater than that of small biopsies (eighty cases (22.8%) and sixty-two cases (17.7%, respectively)). There was no significant relationship in this regard (p = 0.077).
UNASSIGNED: Compared to most similar studies, the present study reported the highest discrepancy between the diagnosis of the first pathologist and the reviewing pathologist (40.9%).

Introduction Colorectal cancer is the fifth most common cancer in the world. For loco-regionally confined disease surgery is the definitive treatment. An adequate surgical pathology report is mandatory for the selection of adjuvant therapy. The objective of this study is to analyze whether adequate information is provided or not in the surgical pathology reports of colorectal carcinoma as according to College of American Pathologists (CAP) guidelines. Method This is a cross-sectional study carried out in the Department of Clinical Oncology, Jinnah Postgraduate Medical Center (JPMC) Karachi, tertiary care hospital in Pakistan. The duration of the study was from February 2020 to January 2021. A total of 153 surgical pathology reports issued by 11 different hospital-based laboratories after definitive surgery was assessed to look at its concordance rate with the checklist adapted from the CAP guidelines. Results Out of 153 surgical pathology reports, clinical information was provided in 72.5% of reports. Details of tumor extension were present in 88.2%, tumor margin in 75%, surgical procedure in 79%, and tumor deposits in 39.2% of reports. Macroscopic details including tumor perforation and evaluation of mesorectum were documented in 51.6% and 53.5% of the reports respectively. Details regarding perineural invasion along with lymphovascular invasion were present in 81.6% and 93% of the reports, respectively. The treatment effect was documented in only 25% of reports and regional lymph node status has been described in 85% of reports. Parameters described in all surgical pathology reports were: tumor site, tumor type, histologic type, and histologic grade. The pathological stage of the disease was documented in 91.5% of the reports. Conclusion This study concluded that surgical pathology reports of the majority of pathology laboratories were not fully adhered to the checklist provided by the CAP guidelines. This will affect post-operative management along with the prediction of disease prognosis.

Adequate reporting of pathological findings is essential for optimal patient management and to perform high-quality research. The aim of this study was to assess the completeness of pathology reports of gallbladder cancer (GBC) at the nationwide level to assess guideline adherence and make recommendations for improvement. A retrospective population-based cohort of GBC patients diagnosed in the Netherlands from 2000 to 2019 was collected using data from the Dutch Cancer Registry and the nationwide network and registry of histology. Pathology reports were scored on the presence and content of essential and optional items according to the Dutch consensus-based guideline on biliary tract cancer. By histopathological review of cases, we compared findings with the conclusion of the corresponding pathology report. All pathology reports (n = 849) had a narrative, nonstructured format. Overall completeness was low. Information on key prognostic factors, such as tumor side (hepatic vs. serosal), status of cystic duct and liver surgical margins and venous and perineural invasion, was frequently lacking (80%, 23%, 59%, 74% and 74% missing, respectively). Whereas certain items were often missing from the report, they could be retrospectively detected in a substantial proportion of cases during pathology review (n = 738). In conclusion, significant improvements could be made in the reporting of GBC in the Netherlands. Synoptic reporting could greatly enhance the completeness of reports, as already demonstrated for tumor types.

The Gleason score is an important grading factor of prostate cancer. Gleason scores can be extracted from pathology report texts using regular expressions, but previously developed programmes have targeted only relatively simple Gleason score expressions. We developed a programme capable of extracting also complex expressions. The programme is relatively easy to adapt to other languages and datasets.
We developed and evaluated our regular expression-based programme using manually processed pathology reports of prostate cancer cases diagnosed in Finland in 2016-2017. Both simple and complex Gleason score expressions were targeted. We measured the performance of our programme using recall, precision, and the F1. The proportion of complex Gleason score expressions was estimated as the complement of the recall when only addition expressions (e.g. \"Gleason 3 + 4\") were targeted.
The detection of values (scores and score components) is based on mandatory keywords before or after the value. The programme favours precision over recall by primarily allowing for lists of optional expressions between keyword-value pairs and only secondarily allowing for arbitrary expressions. The programme is straightforward to adapt to new datasets by modifying the lists of mandatory and optional expressions. The full and addition-only programmes had 92% (95% CI: [90%, 95%]) and 65% ([61%, 70%]) recall and high precision (98% [97%, 99%] and 100% [99%, 100%]), respectively. The estimated proportion of complex Gleason score expressions was 100-65 = 35%.
Even complex Gleason score expressions can be extracted with high recall and precision using regular expressions. We recommend implementing automated Gleason score extraction where possible by adapting our validated programme.

UNASSIGNED: Biopsy plays a crucial role in definitive diagnosis of lesions and consequently, appropriate treatment of them. Clinicians should correctly do the biopsy in accordance to the existing principles and guidelines to prevent adverse effects on the pathologist\'s diagnosis. This study aimed to determine the frequency and reasons for not providing definitive histopathological diagnosis of the biopsy samples belong to the laboratory of the Department of Oral and Maxillofacial Pathology, Faculty of Dentistry, Hamadan University of Medical Sciences.
UNASSIGNED: Archival reports belong to 2006-2016 period of the related laboratory were studied to determine the reports with no definitive histopathological diagnosis.
UNASSIGNED: Out of 1018 archived reports; 90 reports (8.84%) had no definitive diagnosis. The most common reasons found were incompatibility between the clinical/radiographical diagnosis and histopathological findings for 42 cases (46.66%), absence of adequate information about the clinical/radiographical findings for 17 cases (18.88%) and inappropriate quality of samples for 13 cases (14.44%), respectively.
UNASSIGNED: The reasons for not providing definitive histopathological diagnosis of the biopsy samples in present study indicated that preparation, assessment and diagnosis of microscopic slide by pathologists do not separate from the clinician performance.

In perihilar cholangiocarcinoma (PHC), interpretation of the resection specimen is challenging for pathologists and clinicians alike. Thorough and correct reporting is necessary for reliable interpretation of residual disease status. The aim of this study is to assess completeness of PHC pathology reports in a single center and assess what hampers interpretation of pathology reports by clinicians. Pathology reports of patients resected for PHC at a single expert tertiary center drafted between 2000 and 2018 were assessed. Reports were assessed regarding completeness, according to the guideline of the International Collaboration on Cancer Reporting (ICCR). A total of 146 reports were assessed. Prognostic tumor characteristics such as vasoinvasive growth and perineural growth were missing in 30/146 (34%) and 22/146 (15%), respectively. One or more planes were missing in 94/146 (64%) of the reports, with the periductal dissection plane missing in 51/145 (35%). Residual disease could be re-classified from R0 to R1 in 22 patients (15%). Reasons for R1 in these patients were the presence of a positive periductal dissection plane (n = 2), < 1-mm margin at the periductal dissection plane (n = 11), or liver parenchyma (n = 9). Completeness of reports improved significantly when drafted by an expert HPB pathologist. This study demonstrates that pathology reporting of PHC is challenging. Reports are frequently incomplete and often do not incorporate assessment of all resection planes and the dissection plane. The periductal dissection plane is frequently overlooked, but is a major cause of residual disease.