Objectives: To guide language model (LM) selection by comparing finetuning vs. zero-shot use, generic pretraining vs. domain-adjacent vs. further domain-specific pretraining, and bidirectional language models (BiLMs) such as BERT vs. unidirectional LMs (LLMs) for clinical classification. Materials and Methods: We evaluated BiLMs (RoBERTa, PathologyBERT, Gatortron) and LLM (Mistral nemo instruct 12B) on three British Columbia Cancer Registry (BCCR) pathology classification tasks varying in difficulty/data size. We assessed zero-shot vs. finetuned BiLMs, zero-shot LLM, and further BCCR-specific pretraining using macro-average F1 scores. Results: Finetuned BiLMs outperformed zero-shot BiLMs and zero-shot LLM. The zero-shot LLM outperformed zero-shot BiLMs but was consistently outperformed by finetuned BiLMs. Domain-adjacent BiLMs generally outperformed generic BiLMs after finetuning. Further domain-specific pretraining boosted complex/low-data task performance, with otherwise modest gains. Conclusions: For specialized classification, finetuning BiLMs is crucial, often surpassing zero-shot LLMs. Domain-adjacent pretrained models are recommended. Further domain-specific pretraining provides significant performance boosts, especially for complex/low-data scenarios. BiLMs remain relevant, offering strong performance/resource balance for targeted clinical tasks.

PURPOSE: Population-based cancer registries (PBCRs) collect data on all new cancer diagnoses in a defined population. Data are sourced from pathology reports, and the PBCRs rely on manual and rule-based solutions. This study presents a state-of-the-art natural language processing (NLP) pipeline, built by fine-tuning pretrained language models (LMs). The pipeline is deployed at the British Columbia Cancer Registry (BCCR) to detect reportable tumors from a population-based feed of electronic pathology. METHODS: We fine-tune two publicly available LMs, GatorTron and BlueBERT, which are pretrained on clinical text. Fine-tuning is done using BCCR's pathology reports. For the final decision making, we combine both models' output using an OR approach. The fine-tuning data set consisted of 40,000 reports from the diagnosis year of 2021, and the test data sets consisted of 10,000 reports from the diagnosis year 2021, 20,000 reports from diagnosis year 2022, and 400 reports from diagnosis year 2023. RESULTS: The retrospective evaluation of our proposed approach showed boosted reportable accuracy, maintaining the true reportable threshold of 98%. CONCLUSION: Disadvantages of rule-based NLP in cancer surveillance include manual effort in rule design and sensitivity to language change. Deep learning approaches demonstrate superior performance in classification. PBCRs distinguish reportability status of incoming electronic cancer pathology reports. Deep learning methods provide significant advantages over rule-based NLP.

PURPOSE: Breast cancer relapses are rarely collected by cancer registries because of logistical and financial constraints. Hence, we investigated natural language processing (NLP), enhanced with state-of-the-art deep learning transformer tools and large language models, to automate relapse identification in the text of computed tomography (CT) reports. METHODS: We analyzed follow-up CT reports from patients diagnosed with breast cancer between January 1, 2005, and December 31, 2014. The reports were curated and annotated for the presence or absence of local, regional, and distant breast cancer relapses. We performed 10-fold cross-validation to evaluate models identifying different types of relapses in CT reports. Model performance was assessed with classification metrics, reported with 95% confidence intervals. RESULTS: In our data set of 1,445 CT reports, 799 (55.3%) described any relapse, 72 (5.0%) local relapses, 97 (6.7%) regional relapses, and 743 (51.4%) distant relapses. The any-relapse model achieved an accuracy of 89.6% (87.8-91.1), with a sensitivity of 93.2% (91.4-94.9) and a specificity of 84.2% (80.9-87.1). The local relapse model achieved an accuracy of 94.6% (93.3-95.7), a sensitivity of 44.4% (32.8-56.3), and a specificity of 97.2% (96.2-98.0). The regional relapse model showed an accuracy of 93.6% (92.3-94.9), a sensitivity of 70.1% (60.0-79.1), and a specificity of 95.3% (94.2-96.5). Finally, the distant relapse model demonstrated an accuracy of 88.1% (86.2-89.7), a sensitivity of 91.8% (89.9-93.8), and a specificity of 83.7% (80.5-86.4). CONCLUSION: We developed NLP models to identify local, regional, and distant breast cancer relapses from CT reports. Automating the identification of breast cancer relapses can enhance data collection about patient outcomes.

High dose rate (HDR) brachytherapy is increasingly adopted for dose escalation in prostate cancer treatment. We report the clinical efficacy and toxicity of HDR prostate brachytherapy combined with external beam radiotherapy (EBRT) and evaluate the predictability of the biochemical definition of cure of 4-year PSA ≤0.2 ng/mL for failure free survival (FFS). A single centre retrospective study was conducted, including all patients with high-tier intermediate risk and high-risk prostate cancer treated with HDR brachytherapy combined with EBRT from 2011 to 2019. Patient and prostate cancer characteristics, treatment, clinical endpoints, and follow up were collected. Total 319 patients were analyzed. The median age was 68 with median follow up of 77.1 months. Total 142 had high-tier intermediate and 177 had high-risk disease. Brachytherapy doses were initially 20 Gy/2 fractions, and subsequently 15 Gy/1 fraction. All patients received 46 Gy/23 fractions of EBRT. Overall survival at 5 and 9 years was 92.2% and 77.0%, respectively. Failure-free survival (FFS) was 86.0% at 5 years and 76.1% at 9 years. PSA ≤ 0.2 ng/mL at 4 years was seen in 79.3% of patients and was associated with FFS of 94.1% at 9 years. Grade 3 urethral stricture, hematuria, or proctitis occurred in 2.8%, 0%, and 0%, respectively. HDR brachytherapy in addition to EBRT is effective treatment for unfavourable localized prostate cancer with a very acceptable toxicity profile. The biochemical definition of cure of PSA < 0.2 ng/mL at 4 years was predictive for FFS at 9 years.

PURPOSE: Many Natural Language Processing (NLP) methods achieve greater performance when the input text is preprocessed to remove extraneous or unnecessary text. A technique known as text segmentation can facilitate this step by isolating key sections from a document. Give that transformer models—such as Bidirectional Encoder Representations from Transformers (BERT)—have demonstrated state-of-the-art performance on many NLP tasks, it is desirable to leverage such models for segmentation. However, transformer models are typically limited to only 512 input tokens and are not well suited for lengthy documents such as cancer pathology reports. The Longformer is a modified transformer model designed to intake longer documents while retaining the positive characteristics of standard transformers. This study presents a Longformer model fine-tuned for cancer pathology report segmentation. METHODS: We fine-tuned a Longformer Question-Answer (QA) model on 504 manually annotated pathology reports to isolate sections such as diagnosis, addenda, and clinical history. We compared baseline methods including regular expressions (regex) and BERT QA. However, those methods may fail to correctly identify section boundaries. Model performance was evaluated using sequence recall, precision, and F1 score. RESULTS: Final test results were obtained on a hold-out test set of 304 cancer pathology reports. We report sequence F1 scores for the following sections: diagnosis (0.77), addenda (0.48), clinical history (0.89), and overall (0.68). CONCLUSION: We present a fine-tuned Longformer model to isolate key sections from cancer pathology reports for downstream analyses. Our model performs segmentation with greater accuracy.