Natural Language Processing to Identify Abnormal Breast, Lung, and Cervical Cancer Screening Test Results from Unstructured Reports to Support Timely Follow-up

Bootstrapping BI-RADS classification using large language models and transformers in breast magnetic resonance imaging reports

Breast cancer is one of the most common malignancies among women globally. Magnetic resonance imaging (MRI), as the final non-invasive diagnostic tool before biopsy, provides detailed free-text reports that support clinical decision-making. Therefore, the effective utilization of the information in MRI reports to make reliable decisions is crucial for patient care. This study proposes a novel method for BI-RADS classification using breast MRI reports. Large language models are employed to transform free-text reports into structured reports. Specifically, missing category information (MCI) that is absent in the free-text reports is supplemented by assigning default values to the missing categories in the structured reports. To ensure data privacy, a locally deployed Qwen-Chat model is employed. Furthermore, to enhance the domain-specific adaptability, a knowledge-driven prompt is designed. The Qwen-7B-Chat model is fine-tuned specifically for structuring breast MRI reports. To prevent information loss and enable comprehensive learning of all report details, a fusion strategy is introduced, combining free-text and structured reports to train the classification model. Experimental results show that the proposed BI-RADS classification method outperforms existing report classification methods across multiple evaluation metrics. Furthermore, an external test set from a different hospital is used to validate the robustness of the proposed approach. The proposed structured method surpasses GPT-4o in terms of performance. Ablation experiments confirm that the knowledge-driven prompt, MCI, and the fusion strategy are crucial to the model's performance.

A Multilevel Primary Care Intervention to Improve Follow-Up of Overdue Abnormal Cancer Screening Test Results: A Cluster Randomized Clinical Trial

Importance

Realizing the benefits of cancer screening requires testing of eligible individuals and processes to ensure follow-up of abnormal results.

Objective

To test interventions to improve timely follow-up of overdue abnormal breast, cervical, colorectal, and lung cancer screening results.

Design, Setting, and Participants

Pragmatic, cluster randomized clinical trial conducted at 44 primary care practices within 3 health networks in the US enrolling patients with at least 1 abnormal cancer screening test result not yet followed up between August 24, 2020, and December 13, 2021.

Intervention

Automated algorithms developed using data from electronic health records (EHRs) recommended follow-up actions and times for abnormal screening results. Primary care practices were randomized in a 1:1:1:1 ratio to (1) usual care, (2) EHR reminders, (3) EHR reminders and outreach (a patient letter was sent at week 2 and a phone call at week 4), or (4) EHR reminders, outreach, and navigation (a patient letter was sent at week 2 and a navigator outreach phone call at week 4). Patients, physicians, and practices were unblinded to treatment assignment.

Main Outcomes and Measures

The primary outcome was completion of recommended follow-up within 120 days of study enrollment. The secondary outcomes included completion of recommended follow-up within 240 days of enrollment and completion of recommended follow-up within 120 days and 240 days for specific cancer types and levels of risk.

Results

Among 11 980 patients (median age, 60 years [IQR, 52-69 years]; 64.8% were women; 83.3% were White; and 15.4% were insured through Medicaid) with an abnormal cancer screening test result for colorectal cancer (8245 patients [69%]), cervical cancer (2596 patients [22%]), breast cancer (1005 patients [8%]), or lung cancer (134 patients [1%]) and abnormal test results categorized as low risk (6082 patients [51%]), medium risk (3712 patients [31%]), or high risk (2186 patients [18%]), the adjusted proportion who completed recommended follow-up within 120 days was 31.4% in the EHR reminders, outreach, and navigation group (n = 3455), 31.0% in the EHR reminders and outreach group (n = 2569), 22.7% in the EHR reminders group (n = 3254), and 22.9% in the usual care group (n = 2702) (adjusted absolute difference for comparison of EHR reminders, outreach, and navigation group vs usual care, 8.5% [95% CI, 4.8%-12.0%], P < .001). The secondary outcomes showed similar results for completion of recommended follow-up within 240 days and by subgroups for cancer type and level of risk for the abnormal screening result.

Conclusions and Relevance

A multilevel primary care intervention that included EHR reminders and patient outreach with or without patient navigation improved timely follow-up of overdue abnormal cancer screening test results for breast, cervical, colorectal, and lung cancer.

Trial Registration

ClinicalTrials.gov Identifier: NCT03979495.

A scoping review of natural language processing of radiology reports in breast cancer

Various natural language processing (NLP) algorithms have been applied in the literature to analyze radiology reports pertaining to the diagnosis and subsequent care of cancer patients. Applications of this technology include cohort selection for clinical trials, population of large-scale data registries, and quality improvement in radiology workflows including mammography screening. This scoping review is the first to examine such applications in the specific context of breast cancer. Out of 210 identified articles initially, 44 met our inclusion criteria for this review. Extracted data elements included both clinical and technical details of studies that developed or evaluated NLP algorithms applied to free-text radiology reports of breast cancer. Our review illustrates an emphasis on applications in diagnostic and screening processes over treatment or therapeutic applications and describes growth in deep learning and transfer learning approaches in recent years, although rule-based approaches continue to be useful. Furthermore, we observe increased efforts in code and software sharing but not with data sharing.