Thyroid cancer, recent advances in diagnosis and therapy.

Over the past several decades, the approach to the diagnosis and management of patients with follicular cell-derived thyroid cancer has evolved based on improved classification of patients better matching clinical outcomes, as well as advances in imaging, laboratory, molecular technologies and knowledge. While thyroid surgery, radioactive iodine therapy and TSH suppression remain the mainstays of treatment, this expansion of knowledge has enabled de-escalation of therapy for individuals diagnosed with low-risk well-differentiated thyroid cancer; better definition of treatment choices for patients with more aggressive disease; and improved ability to optimize treatments for patients with persistent and/or progressive disease. Most recently, the advancement of knowledge regarding the molecular aspects of thyroid cancer has improved thyroid cancer diagnosis and has enabled individualized therapeutic options for selected patients with the most aggressive forms of the disease. Guidelines from multiple societies across the world reflect these changes, which focus on taking a more individualized approach to clinical management. In this review, we discuss the current more personalized approach to patients with follicular cell-derived thyroid cancer and point toward areas of future research still needed in the field.

Identification, isolation and characterization of human LGR5-positive colon adenoma cells.

The intestine is maintained by stem cells located at the base of crypts and distinguished by the expression of LGR5. Genetically engineered mouse models have provided a wealth of information about intestinal stem cells, whereas less is known about human intestinal stem cells owing to difficulty detecting and isolating these cells. We established an organoid repository from patient-derived adenomas, adenocarcinomas and normal colon, which we analyzed for variants in 71 colorectal cancer (CRC)-associated genes. Normal and neoplastic colon tissue organoids were analyzed by immunohistochemistry and fluorescent-activated cell sorting for LGR5. LGR5-positive cells were isolated from four adenoma organoid lines and were subjected to RNA sequencing. We found that LGR5 expression in the epithelium and stroma was associated with tumor stage, and by integrating functional experiments with LGR5-sorted cell RNA sequencing data from adenoma and normal organoids, we found correlations between LGR5 and CRC-specific genes, including dickkopf WNT signaling pathway inhibitor 4 (DKK4) and SPARC-related modular calcium binding 2 (SMOC2). Collectively, this work provides resources, methods and new markers to isolate and study stem cells in human tissue homeostasis and carcinogenesis.

In vitro patterning of pluripotent stem cell-derived intestine recapitulates in vivo human development.

The intestine plays a central role in digestion, nutrient absorption and metabolism, with individual regions of the intestine having distinct functional roles. Many examples of region-specific gene expression in the adult intestine are known, but how intestinal regional identity is established during development is a largely unresolved issue. Here, we have identified several genes that are expressed in a region-specific manner in the developing human intestine. Using human embryonic stem cell-derived intestinal organoids, we demonstrate that the duration of exposure to active FGF and WNT signaling controls regional identity. Short-term exposure to FGF4 and CHIR99021 (a GSK3ß inhibitor that stabilizes ß-catenin) resulted in organoids with gene expression patterns similar to developing human duodenum, whereas longer exposure resulted in organoids similar to ileum. When region-specific organoids were transplanted into immunocompromised mice, duodenum-like organoids and ileum-like organoids retained their regional identity, demonstrating that regional identity of organoids is stable after initial patterning occurs. This work provides insights into the mechanisms that control regional specification of the developing human intestine and provides new tools for basic and translational research.

COVID-19 Pandemic and Surgical Oncology: Preserving the Academic Mission.

BACKGROUND: The global pandemic of respiratory disease cause by the novel human coronavirus (SARS-CoV-2) has caused untold suffering, loss of life and upheaval in society. The pandemic has lead to massive redirection of health care resources to treat the surge of COVID-19 patients, and enforcement of social distancing to reduce the rate of transmission. METHODS: Editorial Board members provided observations of the implications of the pandemic on academic surgical oncology. RESULTS: Delivery of health care to other populations including cancer patients has been significantly disrupted. The implications both short term and long term threaten preservation of the academic mission in medicine at large, and certainly in the field of surgical oncology. CONCLUSIONS: The effects on surgical oncology training, research and clinical trials are major.

Thyroid Ultrasound Reports: Will the Thyroid Imaging, Reporting, and Data System Improve Natural Language Processing Capture of Critical Thyroid Nodule Features?

BACKGROUND: Critical thyroid nodule features are contained in unstructured ultrasound (US) reports. The Thyroid Imaging, Reporting, and Data System (TI-RADS) uses five key features to risk stratify nodules and recommend appropriate intervention. This study aims to analyze the quality of US reporting and the potential benefit of Natural Language Processing (NLP) systems in efficiently capturing TI-RADS features from text reports. MATERIALS AND METHOD: This retrospective study used free-text thyroid US reports from an academic center (A) and community hospital (B). Physicians created "gold standard" annotations by manually extracting TI-RADS features and clinical recommendations from reports to determine how often they were included. Similar annotations were created using an automated NLP system and compared with the gold standard. RESULTS: Two hundred eighty-two reports contained 409 nodules at least 1-cm in maximum diameter. The gold standard identified three nodules (0.7%) which contained enough information to calculate a complete TI-RADS score. Shape was described most often (92.7% of nodules), whereas margins were described least often (11%). A median number of two TI-RADS features are reported per nodule. The NLP system was significantly less accurate than the gold standard in capturing echogenicity (27.5%) and margins (58.9%). One hundred eight nodule reports (26.4%) included clinical management recommendations, which were included more often at site A than B (33.9 versus 17%, P < 0.05). CONCLUSIONS: These results suggest a gap between current US reporting styles and those needed to implement TI-RADS and achieve NLP accuracy. Synoptic reporting should prompt more complete thyroid US reporting, improved recommendations for intervention, and better NLP performance.

Outcomes after completion thyroidectomy versus total thyroidectomy for differentiated thyroid cancer: A single-center experience.

BACKGROUND: Thyroid cancer diagnoses are often discovered after diagnostic thyroid lobectomy. Completion thyroidectomy (CT) may be indicated for intermediate or high-risk tumors to facilitate surveillance and/or adjuvant treatment. The completeness of thyroid resection and the safety of CT compared to total thyroidectomy (TT) is unclear. We assessed outcomes after TT or CT to determine completeness of resection and risk of complications. METHODS: Patients undergoing TT or CT between 2000 and 2018 were retrospectively reviewed. Pathology, unstimulated thyroglobulin (uTg), parathyroid hormone (PTH), rates of hematoma, and recurrent laryngeal nerve (RLN) injury were compared. RESULTS: Differentiated thyroid cancer (DTC) was identified in 954 patients undergoing TT and 142 patients undergoing CT. Postoperative uTg at 6 months was not different between TT and CT, 0.2 vs 0.2 ng/mL, P = .37. Transient hypoparathyroidism with immediate postoperative PTH less than 10 was more common after TT, 14.3 vs 6.0% (P = .009). No differences were noted regarding postoperative hematoma, transient RLN injury, permanent hypoparathyroidism, and permanent RLN injury. CONCLUSIONS: If CT is required for DTC, a complete resection, as assessed by postoperative uTg, can be achieved. Furthermore, CT is significantly less likely to result in transient hypoparathyroidism and poses no additional risk of RLN injury, hematoma, or permanent hypoparathyroidism.

Hormonal Evaluation of Incidental Adrenal Masses: The Exception, Not the Rule.

BACKGROUND: Incidental adrenal masses (IAMs) occur in approximately 4% of patients undergoing abdominal CT scans for any indication. Hormonal evaluation is recommended for all IAMs. The purpose of this study was to identify the rate of IAMs in a screening population and to determine the adequacy of endocrine evaluation of newly identified IAMs based on established guidelines. METHODS: This was a retrospective analysis of 6913 patients undergoing a non-contrast screening CT colonography at a single academic medical center between June 2004 and July 2012. RESULTS: The prevalence of IAMs in this asymptomatic screening population was 2.1% (n = 148). Of those patients, 8.8% (n = 11) underwent some form of hormonal evaluation and only 6.4% (n = 8) patients had a "complete" workup. Cortisol, metanephrines, and an aldosterone-renin ratio were evaluated in 8.0%, 7.2%, and 4.0% of patients, respectively. Of the patients (n = 11) who underwent hormonal evaluation, 27.3% had functional masses and 36.4% underwent surgery. Of those who did not have hormonal evaluation, 42.1% (n = 48) had comorbidities that should have prompted hormonal evaluation based on established guidelines. Hormonal evaluation was not performed in 89.4% of patients with hypertension and 21.1% of patients with diabetes. 88.9% of patients on three or more antihypertensive medications did not undergo any hormonal evaluation. CONCLUSIONS: Compliance with IAM workup guidelines is poor, which may result in missed diagnosis of functional adrenal masses. Establishment of a robust protocol and education on appropriate workup for IAMs is necessary for adequate hormonal evaluation.

Notch1 and Notch2 receptors regulate mouse and human gastric antral epithelial cell homoeostasis.

OBJECTIVE: We tested the ability of Notch pathway receptors Notch1 and Notch2 to regulate stem and epithelial cell homoeostasis in mouse and human gastric antral tissue. DESIGN: Mice were treated with the pan-Notch inhibitor dibenzazepine (DBZ) or inhibitory antibodies targeting Notch1 and/or Notch2. Epithelial proliferation, apoptosis and cellular differentiation were measured by histological and molecular approaches. Organoids were established from mouse and human antral glands; growth and differentiation were measured after treatment with Notch inhibitors. RESULTS: Notch1 and Notch2 are the predominant Notch receptors expressed in mouse and human antral tissue and organoid cultures. Combined inhibition of Notch1 and Notch2 in adult mice led to decreased epithelial cell proliferation, including reduced proliferation of LGR5 stem cells, and increased apoptosis, similar to the response to global Notch inhibition with DBZ. Less pronounced effects were observed after inhibition of individual receptors. Notch pathway inhibition with DBZ or combined inhibition of Notch1 and Notch2 led to increased differentiation of all gastric antral lineages, with remodelling of cells to express secretory products normally associated with other regions of the GI tract, including intestine. Analysis of mouse and human organoids showed that Notch signalling through Notch1 and Notch2 is intrinsic to the epithelium and required for organoid growth. CONCLUSIONS: Notch signalling is required to maintain gastric antral stem cells. Notch1 and Notch2 are the primary Notch receptors regulating epithelial cell homoeostasis in mouse and human stomach.

Organoid Models of Human Gastrointestinal Development and Disease.

We have greatly advanced our ability to grow a diverse range of tissue-derived and pluripotent stem cell-derived gastrointestinal (GI) tissues in vitro. These systems, broadly referred to as organoids, have allowed the field to move away from the often nonphysiological, transformed cell lines that have been used for decades in GI research. Organoids are derived from primary tissues and have the capacity for long-term growth. They contain varying levels of cellular complexity and physiological similarity to native organ systems. We review the latest discoveries from studies of tissue-derived and pluripotent stem cell-derived intestinal, gastric, esophageal, liver, and pancreatic organoids. These studies have provided important insights into GI development, tissue homeostasis, and disease and might be used to develop personalized medicines.

Potential effects of molecular testing of indeterminate thyroid nodule fine needle aspiration biopsy on thyroidectomy volume.

BACKGROUND: Increasing utilization of genetic expression profiling (GEP) for thyroid nodules with indeterminate fine needle aspiration (FNA) results will potentially decrease the number of patients requiring diagnostic thyroidectomy. This study sought to determine the potential effects of GEP for indeterminate thyroid FNA results on thyroidectomy volume. METHODS: A retrospective review of thyroidectomy procedures performed over 1 year at the University of Michigan in the endocrine surgery division evaluated the indications for thyroidectomy, FNA Bethesda classification, and final surgical pathology to determine how application of GEP on indeterminate FNA results would affect decision for surgery and subsequent thyroidectomy volume. RESULTS: During the study period, 358 thyroidectomies were performed. The indication for procedure included: FNA findings, n = 122; symptomatic multinodular goiter, n = 85; nodule >4 cm, n = 30; Graves', n = 26; other, n = 95. FNA was performed in 231 patients. Bethesda classification included: benign, n = 69; malignant, n = 55; follicular lesion of undetermined significance, n = 59; follicular neoplasm, n = 20; suspicious for malignancy, n = 16; nondiagnostic, n = 12. If standard GEP was performed for all indeterminate FNA results, it would have influenced the decision for surgery in 68 (19 %) patients. Assuming 38 % of indeterminate FNA specimens will have benign results on genetic profiling, 27 patients would not have undergone thyroidectomy, translating into a 7.2 % decrease in overall thyroidectomy volume over a year. CONCLUSIONS: In an academic endocrine surgery program, the most common indication for thyroidectomy is an FNA result; however, standard application of GEP for all indeterminate thyroid FNAs would result in a minimal reduction in overall thyroidectomy volume.

Heterogeneous parathyroid near-infrared autofluorescence patterns are associated with single adenomas in primary hyperparathyroidism.

BACKGROUND: Primary adenoma (PA) and multi-gland hyperplasia (MGH) account for 85% and 15% of primary hyperparathyroidism (PHPT) cases, respectively. Near-infrared autofluorescence (NIRAF) enhances intraoperative parathyroid identification. We hypothesized that PA would display a more heterogeneous NIRAF pattern compared to MGH. METHODS: Patients undergoing surgery for sporadic PHPT were categorized based on the presence of PA or MGH. To quantify heterogeneity, we utilized ratios of (1) mean parathyroid gland (PG) NIRAF over background NIRAF (mean ratio), (2) minimum and (3) maximum PG NIRAF over mean PG NIRAF (minimum and maximum ratios). Additionally, a heterogeneity score was quantified using mean ratio (mean PG NIRAF over background NIRAF), and overall NIRAF (mean NIRAF of eight random 15 × 15 pixel areas). A point was assigned to ratios <0.8 or >1.2. Images were quantified by ImageJ software. Mann-Whitney test was performed for all comparisons. RESULTS: Of 78 patients, 63 had a single PA and 15 had MGH, totaling 102 PGs. There was no difference between their mean ratios. PA had a lower minimum ratio compared to that of MGH (0.86 ± 0.01 vs. 0.93 ± 0.01, p = 0.001) and a brighter maximum ratio (1.21 ± 0.02 vs. 1.12 ± 0.01, p = 0.0008). PA also scored higher on their heterogeneity scores compared to MGH (1.27 ± 0.23 vs. 0.33 ± 0.15, p = 0.001). CONCLUSION: Single parathyroid adenomas display a more heterogeneous autofluorescence pattern compared to that of multi-gland hyperplasia. Intraoperative characterization of PGs by real-time NIR imaging patterns may be a beneficial adjunct during parathyroid surgery.

Bioengineered in vitro three-dimensional tumor models in endocrine cancers.

Abstract: Endocrine tumors are a heterogeneous cluster of malignancies that originate from cells that can secrete hormones. Examples include, but are not limited to, thyroid cancer, adrenocortical carcinoma, and neuroendocrine tumors. Many endocrine tumors are relatively slow to proliferate, and as such, they often do not respond well to common antiproliferative chemotherapies. Therefore, increasing attention has been given to targeted therapies and immunotherapies in these diseases. However, in contrast to other cancers, many endocrine tumors are relatively rare, and as a result, less is understood about their biology, including specific targets for intervention. Our limited understanding of such tumors is in part due to a limitation in model systems that accurately recapitulate and enable mechanistic exploration of these tumors. While mouse models and 2D cell cultures exist for some endocrine tumors, these models often may not accurately model nuances of human endocrine tumors. Mice differ from human endocrine physiology and 2D cell cultures fail to recapitulate the heterogeneity and 3D architectures of in vivo tumors. To complement these traditional cancer models, bioengineered 3D tumor models, such as organoids and tumor-on-a-chip systems, have advanced rapidly in the past decade. However, these technologies have only recently been applied to most endocrine tumors. In this review we provide descriptions of these platforms, focusing on thyroid, adrenal, and neuroendocrine tumors and how they have been and are being applied in the context of endocrine tumors.

Clinical characteristics and outcomes of adrenal hemorrhage.

BACKGROUND: Although uncommon, adrenal hemorrhage has multiple etiologies. Because clinical characteristics, management, and outcomes of patients with adrenal hemorrhage are inadequately described, we examined the underlying etiology, need for intervention, evolution of imaging characteristics, and adequacy of subsequent evaluation. METHODS: We performed a retrospective review of patients diagnosed with adrenal hemorrhage (radiologist-confirmed density consistent with hemorrhage on computed tomography) from 2005 to 2021 at a university-based institution. Demographic characteristics, hemorrhage etiology, and subsequent follow-up were analyzed. RESULTS: Of 193 adrenal hemorrhage patients, the mean age was 49.2 ± 18.3 years, and 35% were female. Clinical presentations included trauma (47%), abdominal or flank pain (28%), incidental findings on imaging acquired for other reasons (12%), postoperative complication (8%), or shock (3%). Hemorrhage outside of the gland was present in 62% of patients. Unilateral hemorrhage was more frequent (93%) than bilateral (7%). A total of 12% of patients had nodules, but only 70% of these were identified on initial imaging, and only 43% had hormonal evaluation. Of 7 patients who had adrenalectomy or biopsy, pathology was either benign (57%) or nonadrenal malignancy (43%). No adrenocortical carcinomas were identified. Follow-up imaging was performed in 56% of patients and revealed decreased, stable, resolved, or increased adrenal hemorrhage size in 39%, 19%, 30%, and 12% of patients, respectively. CONCLUSION: Adrenal hemorrhage is secondary to multiple etiologies, most commonly trauma. In the setting of adrenal hemorrhage, many adrenal nodules were not identified on initial imaging. Only a minority of patients with nodules underwent "complete" biochemical evaluation. Follow-up imaging may improve the identification of underlying nodules needing hormonal evaluation.

A 3D adrenocortical carcinoma tumor platform for preclinical modeling of drug response and matrix metalloproteinase activity.

Adrenocortical carcinoma (ACC) has a poor prognosis, and no new drugs have been identified in decades. The absence of drug development can partly be attributed to a lack of preclinical models. Both animal models and 2D cell cultures of ACC fail to accurately mimic the disease, as animal physiology is inherently different than humans, and 2D cultures fail to represent the crucial 3D architecture. Organoids and other small 3D in vitro models of tissues or tumors can model certain complexities of human in vivo biology; however, this technology has largely yet to be applied to ACC. In this study, we describe the generation of 3D tumor constructs from an established ACC cell line, NCI-H295R. NCI-H295R cells were encapsulated to generate 3D ACC constructs. Tumor constructs were assessed for biomarker expression, viability, proliferation, and cortisol production. In addition, matrix metalloproteinase (MMP) functionality was assessed directly using fluorogenic MMP-sensitive biosensors and through infusion of NCI-H295R cells into a metastasis-on-a-chip microfluidic device platform. ACC tumor constructs showed expression of biomarkers associated with ACC, including SF-1, Melan A, and inhibin α. Treatment of ACC tumor constructs with chemotherapeutics demonstrated decreased drug sensitivity compared to 2D cell culture. Since most tumor cells migrate through tissue using MMPs to break down extracellular matrix, we validated the utility of ACC tumor constructs by integrating fluorogenic MMP-sensitive peptide biosensors within the tumor constructs. Lastly, in our metastasis-on-a-chip device, NCI-H295R cells successfully engrafted in a downstream lung cell line-based construct, but invasion distance into the lung construct was decreased by MMP inhibition. These studies, which would not be possible using 2D cell cultures, demonstrated that NCI-H295R cells secreted active MMPs that are used for invasion in 3D. This work represents the first evidence of a 3D tumor constructs platform for ACC that can be deployed for future mechanistic studies as well as development of new targets for intervention and therapies.

Transformation by Tribbles homolog 2 (Trib2) requires both the Trib2 kinase domain and COP1 binding.

Tribbles homolog 2 (Trib2) is a pseudokinase that induces acute myelogenous leukemia (AML) in mice and is highly expressed in a subset of human AML. Trib2 has 3 distinct regions, a proline-rich N-terminus, a serine/threonine kinase homology domain, and a C-terminal constitutive photomorphogenesis 1 (COP1)-binding domain. We performed a structure-function analysis of Trib2 using in vitro and in vivo assays. The N-terminus was not required for Trib2-induced AML. Deletion or mutation of the COP1-binding site abrogated the ability of Trib2 to degrade CCAAT/enhancer-binding protein-α (C/EBP-α), block granulocytic differentiation, and to induce AML in vivo. Furthermore, COP1 knockdown inhibited the ability of Trib2 to degrade C/EBP-α, showing that it is important for mediating Trib2 activity. We also show that the Trib2 kinase domain is essential for its function. Trib2 contains variant catalytic loop sequences, compared with conventional kinases, that we show are necessary for Trib2 activity. The kinase domain mutants bind, but cannot efficiently degrade, C/EBP-α. Together, our data demonstrate that Trib2 can bind both COP1 and C/EBP-α, leading to degradation of C/EBP-α. Identification of the functional regions of Trib2 that are essential to its oncogenic role provides the basis for developing inhibitors that will block Trib functions in cancer.

Differential ability of Tribbles family members to promote degradation of C/EBPalpha and induce acute myelogenous leukemia.

Trib1, Trib2, and Trib3 are mammalian homologs of Tribbles, an evolutionarily conserved Drosophila protein family that mediates protein degradation. Tribbles proteins function as adapters to recruit E3 ubiquitin ligases and enhance ubiquitylation of the target protein to promote its degradation. Increased Trib1 and Trib2 mRNA expression occurs in human myeloid leukemia and induces acute myeloid leukemia in mice, whereas Trib3 has not been associated with leukemia. Given the high degree of structural conservation among Tribbles family members, we directly compared the 3 mammalian Tribbles in hematopoietic cells by reconstituting mice with hematopoietic stem cells retrovirally expressing these proteins. All mice receiving Trib1 or Trib2 transduced hematopoietic stem cells developed acute myeloid leukemia, whereas Trib3 mice did not. Our previous data indicated that Trib2-mediated degradation of the transcription factor, CCAAT/enhancer-binding protein-alpha (C/EBPalpha), is important for leukemogenesis. Similar to Trib2, Trib1 induced C/EBPalpha degradation and inhibited its function. In contrast, Trib3 failed to inactivate or promote efficient degradation of C/EBPalpha. These data reveal that the 3 Tribbles homologs differ in their ability to promote degradation of C/EBPalpha, which account for their differential ability to induce leukemia.

Smaller parathyroids have higher near-infrared autofluorescence intensity in hyperparathyroidism.

BACKGROUND: Intraoperative parathyroid gland identification can be challenging. Parathyroid glands have an intrinsic autofluorescence when excited by wavelengths in the near-infrared region. Studies using near-infrared cameras to detect parathyroid gland near-infrared autofluorescence have suggested improved identification. The pathologic parathyroid glands in primary hyperparathyroidism have variable near-infrared autofluorescence intensity, but how this correlates with different characteristics of hyperparathyroidism is unknown. Our objective was to correlate the fluorescent intensity of excited glands with clinical variables to enhance a surgeon's ability to identify parathyroid glands. METHODS: The data on patients undergoing surgery for primary hyperparathyroidism were collected. The images were collected intraoperatively with a handheld near-infrared device and analyzed. The data consisted of the ratio of mean parathyroid gland near-infrared autofluorescence over background (white gauze) near-infrared autofluorescence. The variables assessed for correlation with autofluorescence intensity were gland volume and weight, preoperative serum calcium and parathyroid hormone, age, body mass index, and sex. The images were quantified by Image J software (National Institutes of Health, Bethesda, MD). The lasso regression was analyzed by R version 4.1.3 to calculate adjusted P values (R Foundation for Statistical Computing, Vienna, Austria). RESULTS: From 2017 to 2021, 131 patients with primary hyperparathyroidism underwent parathyroidectomies of 151 parathyroid glands. The mean near-infrared autofluorescence intensity of parathyroid glands had a negative correlation with weight with lighter glands fluorescing more (P = .019) and a positive correlation with age with glands from older patients fluorescing more (P = .013). There were no significant correlations with preoperative serum calcium and parathyroid hormone, body mass index, and sex (P > .05). CONCLUSION: In patients with primary hyperparathyroidism, we found that autofluorescence intensity correlated with parathyroid gland weight and patient age. This suggested that near-infrared camera use may be particularly helpful in identifying smaller adenomas and in older patients.

Near-infrared Autofluorescence Features of Parathyroid Carcinoma.

Introduction: Parathyroid carcinoma is very rare, and intraoperative definitive diagnosis can be elusive with currently available diagnostics. Near-infrared (NIR) autofluorescence is an emerging tool that identifies parathyroid glands in real time. It is not known whether NIR autofluorescence can detect parathyroid carcinoma intraoperatively. Methods: Patients with preoperative suspicion for parathyroid carcinoma were identified from ongoing studies examining parathyroid autofluorescence with a NIR camera and probe. Specimens from these patients were examined intraoperatively to determine their autofluorescence patterns. Results: Three patients with suspected parathyroid carcinoma were identified preoperatively. Intraoperative NIR autofluorescence imaging showed a relative lack of autofluorescence for all cases, in contrast to parathyroid adenomas and normal parathyroid glands, which typically exhibit significant autofluorescence. Final pathology confirmed parathyroid carcinoma in all cases. Conclusion: Parathyroid carcinoma can be difficult to confirm prior to final pathology review. Our 3 cases suggest that absence of NIR autofluorescence may suggest the likelihood of parathyroid carcinoma, but more studies are needed to investigate this experience.

Ambiguous and Incomplete: Natural Language Processing Reveals Problematic Reporting Styles in Thyroid Ultrasound Reports.

OBJECTIVE: Natural language processing (NLP) systems convert unstructured text into analyzable data. Here, we describe the performance measures of NLP to capture granular details on nodules from thyroid ultrasound (US) reports and reveal critical issues with reporting language. METHODS: We iteratively developed NLP tools using clinical Text Analysis and Knowledge Extraction System (cTAKES) and thyroid US reports from 2007 to 2013. We incorporated nine nodule features for NLP extraction. Next, we evaluated the precision, recall, and accuracy of our NLP tools using a separate set of US reports from an academic medical center (A) and a regional health care system (B) during the same period. Two physicians manually annotated each test-set report. A third physician then adjudicated discrepancies. The adjudicated "gold standard" was then used to evaluate NLP performance on the test-set. RESULTS: A total of 243 thyroid US reports contained 6,405 data elements. Inter-annotator agreement for all elements was 91.3%. Compared with the gold standard, overall recall of the NLP tool was 90%. NLP recall for thyroid lobe or isthmus characteristics was: laterality 96% and size 95%. NLP accuracy for nodule characteristics was: laterality 92%, size 92%, calcifications 76%, vascularity 65%, echogenicity 62%, contents 76%, and borders 40%. NLP recall for presence or absence of lymphadenopathy was 61%. Reporting style accounted for 18% errors. For example, the word "heterogeneous" interchangeably referred to nodule contents or echogenicity. While nodule dimensions and laterality were often described, US reports only described contents, echogenicity, vascularity, calcifications, borders, and lymphadenopathy, 46, 41, 17, 15, 9, and 41% of the time, respectively. Most nodule characteristics were equally likely to be described at hospital A compared with hospital B. CONCLUSIONS: NLP can automate extraction of critical information from thyroid US reports. However, ambiguous and incomplete reporting language hinders performance of NLP systems regardless of institutional setting. Standardized or synoptic thyroid US reports could improve NLP performance.