Photo Credit: magicmine

The following is a summary of “Use of Natural Language Processing to Retrospectively Identify and Improve Follow-Up of Adrenal Incidentalomas Based on Patient and Nodule Characteristics,” published in the January 2025 issue of Surgery by Veazey et al.

Adrenal gland incidentalomas (AGIs) are adrenal lesions discovered incidentally during imaging studies conducted for reasons unrelated to adrenal pathology. The prevalence of AGIs is estimated to be about 5% in all imaging scans and up to 10% in abdominal imaging for patients over 65 years of age. While most AGIs are benign, certain imaging characteristics, such as lesion size, density, heterogeneity, and growth patterns, can indicate a malignant potential. Furthermore, functional lesions, which involve the excessive secretion of adrenal hormones, are present in up to 30% of cases. Given the risks associated with malignancy and hormonal hypersecretion, adherence to established guidelines for follow-up imaging and biochemical studies is essential to ensure significant diagnoses are not overlooked, as these could have serious implications for patient health.

Despite the relatively high prevalence of AGIs, particularly among individuals with obesity, hypertension, and diabetes, many lesions are not adequately flagged for follow-up. Previous research indicates that while recommendations for follow-up have increased, actual completion rates remain low, with less than 33% of patients undergoing imaging and only 20% undergoing biochemical evaluation. The generalizability of these findings to non-metropolitan areas and different healthcare contexts remains uncertain, as does the methodology employed by prior studies to identify and report AGIs.

Natural Language Processing (NLP), a computational technique for analyzing and understanding human language, has gained traction in healthcare for enhancing patient care, research, and administrative processes. In clinical practice, NLP algorithms can analyze unstructured medical records, such as clinical notes and patient histories, to support decision-making and aid in disease diagnosis and treatment planning. Recognizing that incidental findings in radiology reports may often be overlooked by busy providers, this study assessed the feasibility of employing an NLP algorithm to identify AGIs in radiology reports. Following identification, the patient records were reviewed to determine the cohort undergoing guideline-adherent follow-up imaging and biochemical evaluations. It was hypothesized that adherence rates would be low, highlighting the need for improved detection and follow-up protocols.

Source: sciencedirect.com/science/article/abs/pii/S0002961025000212