Combining Image Similarity and Predictive Artificial Intelligence Models to Decrease Subjectivity in Thyroid Nodule Diagnosis and Improve Malignancy Prediction.

OBJECTIVES: To evaluate the efficacy of combining predictive artificial intelligence (AI) and image similarity model to risk stratify thyroid nodules, using retrospective external validation study. METHODS: Two datasets were used to determine efficacy of the AI application. One was Stanford dataset ultrasound images of 192 nodules between April 2017 and May 2018 and the second was private practice consisting of 118 thyroid nodule images between January 2018 and December 2023. The nodules had definitive diagnosis by cytology or surgical pathology. The AI application was used to predict the diagnosis and American College of Radiology Thyroid Imaging and Data System (ACR TI-RADS) score. RESULTS: In the Stanford dataset, the AI application predicted malignancies with sensitivity of 1.0 and specificity of 0.55. Positive predictive value (PPV) was 0.18 and negative predictive value (NPV) was 1.0. The Area Under the Curve - Receiver Operating Characteristic was 0.78. ACR TI-RADS based clinical recommendation had a polychoric correlation of 0.67. In the private dataset, the AI application predicted malignancies with sensitivity of 0.91 and specificity of 0.95. PPV was 0.8 and NPV was 0.98. The area under the curve - receiver operating characteristic was 0.93 and accuracy was 0.94. ACR TI-RADS based score had a polychoric correlation of 0.94. CONCLUSION: The AI application showed good performance for sensitivity and NPV between the two datasets and demonstrated potential for 61.5% reduction in the need for fine needle aspiration and strong correlation to ACR TI-RADS. However, PPV was variable between the datasets possibly from variability in image selection and prevalence of malignancy. If implemented widely and consistently among various clinical settings, this could lead to decreased patient burden associated with an invasive procedure and possibly to decreased health care spending.

Effects of Ultrasound-Guided Laser Ablation Therapy on Symptomatic Benign Thyroid Nodules, Using EchoLaser - Results of a Pilot Study in the United States.

OBJECTIVE: To evalate the efficacy of single ultrasound-guided laser ablation treatment on symptomatic benign thyroid nodules in outpatient setting, we conducted a single-center, prospective, nonrandomized, open label interventional study. METHODS: This study included adults with normal thyroid function and symptomatic thyroid nodules measuring >29 mm and ≤60 mm with <20% cystic area by ultrasound. Laser ablation energy was calculated based on device guidelines. A single nodule was treated under local anesthesia. Each application session lasted between 400 and 600 seconds under a laser power output of 3 W. RESULTS: Ten subjects underwent the procedure and 9 subjects completed 12-month follow-up. Baseline thyroid nodule volumes ranged from 5.51 to 28.22 mL and the total energy delivered ranged from 1477.00 to 7206.50 J. The average nodule volume reduction at 12 months after procedure for 9 patients was 71.087%. Nodules were visible from >1 m away at baseline for all 10 subjects and not visible in 8 of 10 at 3 months and 6 of 9 subjects at 12 months. Most subjects reported that they would likely undergo the procedure again. No significant adverse events were reported. CONCLUSION: This study demonstrated the feasibility of performing laser ablation in a clinical setting in the United States. Reduction in nodule volume and symptoms were observed, and the procedure was tolerated well by the subjects.

Factors Affecting the Job Satisfaction of U.S. Adult Endocrinologists: A Nationwide Survey.

OBJECTIVE: Despite the increased demand and worsening burnout among U.S. endocrinologists, there is a paucity of data on job satisfaction and associated factors. This study examines the factors associated with job satisfaction among a nationally representative sample of U.S. endocrinologists. METHODS: We conducted a cross-sectional survey of 1700 U.S. adult endocrinologists on the Facebook group "Endocrinologists." The survey was conducted over 4 weeks using an anonymous online questionnaire. The 45-question survey assessed job and salary satisfaction scores on a 5-point Likert scale along with multiple job-related variables. Univariate and multivariate analyses were conducted to identify the factors affecting job satisfaction. RESULTS: Out of 1700, 654 adult endocrinologists (504 women and 139 men) completed the survey. The mean job satisfaction score was 3.72 ± 0.86, with 67.5% having high job satisfaction. Comparatively, 339 (52.1%) had high salary satisfaction. There was a statistically significant relationship between the job and salary satisfaction scores (P < .01). Factors significantly associated with the job satisfaction score (P < .05) included the practice region, gender, number of medical assistants per endocrinologist, self-performance of thyroid ultrasound, and number of patients in the hospital per week. Multivariate analysis showed that full-time employment, along with high salary satisfaction, seeing fewer new patients per day, performing thyroid ultrasounds, and fewer patients in the hospital were associated with the highest job satisfaction. CONCLUSION: This study found about one-third of endocrinologists to have lower job satisfaction and identified multiple modifiable factors associated with endocrinologists' job satisfaction. Interventions focused on these potentially modifiable factors may improve job satisfaction among U.S. endocrinologists.

Predicting Forage Quality of Warm-Season Legumes by Near Infrared Spectroscopy Coupled with Machine Learning Techniques.

Warm-season legumes have been receiving increased attention as forage resources in the southern United States and other countries. However, the near infrared spectroscopy (NIRS) technique has not been widely explored for predicting the forage quality of many of these legumes. The objective of this research was to assess the performance of NIRS in predicting the forage quality parameters of five warm-season legumes-guar (Cyamopsis tetragonoloba), tepary bean (Phaseolus acutifolius), pigeon pea (Cajanus cajan), soybean (Glycine max), and mothbean (Vigna aconitifolia)-using three machine learning techniques: partial least square (PLS), support vector machine (SVM), and Gaussian processes (GP). Additionally, the efficacy of global models in predicting forage quality was investigated. A set of 70 forage samples was used to develop species-based models for concentrations of crude protein (CP), acid detergent fiber (ADF), neutral detergent fiber (NDF), and in vitro true digestibility (IVTD) of guar and tepary bean forages, and CP and IVTD in pigeon pea and soybean. All species-based models were tested through 10-fold cross-validations, followed by external validations using 20 samples of each species. The global models for CP and IVTD of warm-season legumes were developed using a set of 150 random samples, including 30 samples for each of the five species. The global models were tested through 10-fold cross-validation, and external validation using five individual sets of 20 samples each for different legume species. Among techniques, PLS consistently performed best at calibrating (R2c = 0.94-0.98) all forage quality parameters in both species-based and global models. The SVM provided the most accurate predictions for guar and soybean crops, and global models, and both SVM and PLS performed better for tepary bean and pigeon pea forages. The global modeling approach that developed a single model for all five crops yielded sufficient accuracy (R2cv/R2v = 0.92-0.99) in predicting CP of the different legumes. However, the accuracy of predictions of in vitro true digestibility (IVTD) for the different legumes was variable (R2cv/R2v = 0.42-0.98). Machine learning algorithms like SVM could help develop robust NIRS-based models for predicting forage quality with a relatively small number of samples, and thus needs further attention in different NIRS based applications.

Evaluation of 30% DAS-444Ø6-6 soybean meal in a subchronic rat toxicity study.

Event DAS-444Ø6-6 soybean is genetically modified (GM) to provide tolerance to 2,4-diclorophenoxyacetic acid (2,4-D), glyphosate, and glufosinate herbicides through expression of the AAD-12, 2mEPSPS, and PAT proteins, respectively. DAS-444Ø6-6 soybeans were evaluated for safety in subchronic rat feeding studies. The results from two previous subchronic rat feeding studies evaluating diets formulated with 20% inclusion of DAS-444Ø6-6 soybean meal (the latter also containing DAS-444Ø6-6 derived hulls and oil) did not show any treatment-related adverse effects. In 2017, to comply with recent guidance from EFSA, a third 90-day rat feeding study was conducted with Sprague-Dawley rats (16/sex/group) with diets formulated either with 15% or 30% w/w of toasted DAS-444Ø6-6 soybean meal. DAS-444Ø6-6 soybean hulls and oil were also added to the transgenic test diets at 1% or 2% w/w and 1.35% or 2.7%, respectively, for the low- and high-dose groups. No toxicologically significant effects were observed under the conditions of this study.

No treatment-related effects with aryloxyalkanoate dioxygenase-12 in three 28-day mouse toxicity studies.

The aryloxyalkanoate dioxygenase-12 (AAD-12) protein is expressed in genetically modified soybean events DAS-68416-4 and DAS-444Ø6-6. Expression of the AAD-12 protein in soybeans confers tolerance to the herbicide 2,4-dichlorophenoxyacetic acid (2,4-D) providing an additional herbicide choice to farmers. This enzyme acts by catalyzing the degradation of 2,4-D into herbicidally inactive metabolites. To meet evolving interpretation of regulations in the European Union, three separate 28-day repeat-dose oral mouse studies were conducted at increasing doses of up to 1100 mg AAD-12 protein/kg bw/day. No treatment-related effects were seen in any of these three studies.

Pronamide: Human relevance of liver-mediated rat leydig cell tumors.

Dietary exposure to pronamide resulted in higher incidences of Leydig cell tumors (LCT) at 1000ppm in a 2-year cancer bioassay, but there were no testes effects at 40 or 200ppm, and no testes effects at 12-months at any concentration. A 90-day mode-of-action (MoA) study was conducted at concentrations of 0, 200, 1000 and 2000ppm. Standard parameters and stereological and proliferation analyses of LCs, targeted testis and liver gene expression, in vitro metabolism of testosterone by liver microsomes, and quantification of serum hormones and testosterone metabolites were evaluated. Increased testosterone metabolism due to increases in hepatic microsomal activity, alterations in serum hormone levels, and other data suggest that LCTs were mediated through a perturbation of the HPG-axis. Data suggest that this occurs after a threshold of exposure is reached, indicating a nonlinear/threshold dose-response. Pronamide-induced rat LCTs mediated by alterations to the HPG-axis have low relevance to humans due to quantitative differences in sensitivity between rats and humans to LCTs. Pronamide displayed no genotoxicity or direct endocrine effects. A margin of exposure approach for risk assessment and derivation of the chronic reference dose based on a point of departure of 200ppm is most appropriate and protective of human health.

An integrated approach for prospectively investigating a mode-of-action for rodent liver effects.

Registration of new plant protection products (e.g., herbicide, insecticide, or fungicide) requires comprehensive mammalian toxicity evaluation including carcinogenicity studies in two species. The outcome of the carcinogenicity testing has a significant bearing on the overall human health risk assessment of the substance and, consequently, approved uses for different crops across geographies. In order to understand the relevance of a specific tumor finding to human health, a systematic, transparent, and hypothesis-driven mode of action (MoA) investigation is, appropriately, an expectation by the regulatory agencies. Here, we describe a novel approach of prospectively generating the MoA data by implementing additional end points to the standard guideline toxicity studies with sulfoxaflor, a molecule in development. This proactive MoA approach results in a more robust integration of molecular with apical end points while minimizing animal use. Sulfoxaflor, a molecule targeting sap-feeding insects, induced liver effects (increased liver weight due to hepatocellular hypertrophy) in an initial palatability probe study for selecting doses for subsequent repeat-dose dietary studies. This finding triggered the inclusion of dose-response investigations of the potential key events for rodent liver carcinogenesis, concurrent with the hazard assessment studies. As predicted, sulfoxaflor induced liver tumors in rats and mice in the bioassays. The MoA data available by the time of the carcinogenicity finding supported the conclusion that the carcinogenic potential of sulfoxaflor was due to CAR/PXR nuclear receptor activation with subsequent hepatocellular proliferation. This MoA was not considered to be relevant to humans as sulfoxaflor is unlikely to induce hepatocellular proliferation in humans and therefore would not be a human liver carcinogen.

Occurrence of spontaneous amphophilic-vacuolar renal tubule tumors in Sprague-Dawley rats from subchronic toxicity studies.

The low background incidence of tumors in rodents from subchronic toxicity studies makes it difficult to assess their relevance, especially when present only in treated animals. This report investigates the occurrence of renal tubule tumors (RTTs), specifically the amphophilic-vacuolar (AV) phenotypic variant, in young Sprague-Dawley (SD) rats from a survey of laboratories conducting subchronic toxicity studies spanning a period of 10 years (2002-2012). This survey establishes a general profile of tumor occurrence; it does not estimate overall incidence or prevalence. AV tumors are spontaneous, nontreatment-related tumors of familial origin, and morphologically distinct from conventional RTTs induced by exposure to renal carcinogens. They are composed of distinct lobules of large, round to polyhedral cells with vacuolated amphophilic to eosinophilic cytoplasm and prominent nucleoli. Data from five collaborating laboratories, representing 37 qualifying studies, are presented. In total, 58 renal tubule neoplasms were recorded in this data set. The AV tumor variant was reported more commonly than the conventional RTT (n = 45 and 13, respectively), and it was recorded in both experimental (n = 32) and control (n = 13) groups. AV tumors occurred more often in females (n = 34) than in males (n = 11); conventional RTTs were recorded more often in males (n = 9) than in females (n = 4). AV tumors often occurred in more than one rat within the same study (up to 7) and were documented to occur in rats as young as 7 to 10 weeks of age. Results from this survey indicate that AV tumors are being reported more commonly in recent years; the majority (n = 33) were reported in studies commencing since 2009. In conclusion, this study reaffirms that AV tumors are spontaneous, nontreatment-related lesions, and suggests that they may be more common than conventional RTTs in young SD rats. The authors propose that AV tumors be recorded separately from conventional RTTs in order to clearly distinguish these two renal tubule neoplasms from one another and allow for appropriate interpretation of a compound's potential carcinogenic effect in the kidney.

Anthrax lethal factor activates K(+) channels to induce IL-1ß secretion in macrophages.

Anthrax lethal toxin (LeTx) is a virulence factor of Bacilillus anthracis that is a bivalent toxin, containing lethal factor (LF) and protective Ag proteins, which causes cytotoxicity and altered macrophage function. LeTx exposure results in early K(+) efflux from macrophages associated with caspase-1 activation and increased IL-1ß release. The mechanism of this toxin-induced K(+) efflux is unknown. The goals of the current study were to determine whether LeTx-induced K(+) efflux from macrophages is mediated by toxin effects on specific K(+) channels and whether altered K(+)-channel activity is involved in LeTx-induced IL-1ß release. Exposure of macrophages to LeTx induced a significant increase in the activities of two types of K(+) channels that have been identified in mouse macrophages: Ba(2+)-sensitive inwardly rectifying K(+) (Kir) channels and 4-aminopyridine-sensitive outwardly rectifying voltage-gated K(+) (Kv) channels. LeTx enhancement of both Kir and Kv required the proteolytic activity of LF, because exposure of macrophages to a mutant LF-protein (LF(E687C)) combined with protective Ag protein had no effect on the currents. Furthermore, blocking Kir and Kv channels significantly decreased LeTx-induced release of IL-1ß. In addition, retroviral transduction of macrophages with wild-type Kir enhanced LeTx-induced release of IL-1ß, whereas transduction of dominant-negative Kir blocked LeTx-induced release of IL-1ß. Activation of caspase-1 was not required for LeTx-induced activation of either of the K(+) channels. These data indicate that a major mechanism through which LeTx stimulates macrophages to release IL-1ß involves an LF-protease effect that enhances Kir and Kv channel function during toxin stimulation.

An evaluation of 2,4-dichlorophenoxyacetic acid in the Amphibian Metamorphosis Assay and the Fish Short-Term Reproduction Assay.

2,4-Dichlorophenoxyacetic acid (2,4-D) was evaluated in both the Amphibian Metamorphosis Assay (AMA) and the Fish Short Term Reproduction Assay (FSTRA). In the AMA, tadpoles were exposed to mean measured 2,4-D concentrations of 0 (water control), 0.273, 3.24, 38.0 and 113 mg acid equivalents (ae)/L for either seven or 21 days. In the FSTRA, fathead minnows were exposed to mean measured 2,4-D concentrations of 0 (water control), 0.245, 3.14, 34.0, and 96.5 mg ae/L for 21 days. The respective concentrations of 2,4-D were not overtly toxic to either Xenopus laevis tadpoles or fathead minnows (Pimephales promelas). In the AMA, there were no signs of either advanced or delayed development, asynchronous development, or significant histopathological effects of the thyroid gland among 2,4-D exposed tadpoles evaluated on either day seven or day 21 of the exposure. Therefore, following the AMA decision logic, 2,4-D is considered "likely thyroid inactive" in the AMA with a No Observable Effect Concentration (NOEC) of 113 mg ae 2,4-D/L. In the FSTRA, there were no significant differences between control and 2,4-D exposed fish in regard to fertility, wet weight, length, gonado-somatic indices, tubercle scores, or blood plasma concentrations of vitellogenin. Furthermore, there were no treatment-related histopathologic changes in the testes or ovaries in any 2,4-D exposed group. The only significant effect was a decrease in fecundity among fish exposed to 96.5 mg ae 2,4-D/L. The cause of the reduced fecundity at the highest concentration of 2,4-D tested in the assay was most likely due to a generalized stress response in the fish, and not due to a specific endocrine mode of action of 2,4-D. Based on fish reproduction, the NOEC in the FSTRA was 34.0 mg ae 2,4-D/L.

Artificial Intelligence for Evaluation of Thyroid Nodules: A Primer.

Background: Artificial intelligence (AI) is broadly defined as the ability of machines to apply human-like reasoning to problem solving. Recent years have seen a rapid growth of AI in many disciplines. This review will focus on AI applications in the assessment of thyroid nodules. Summary: AI encompasses two related computational techniques: machine learning, in which computers learn by observing data provided by humans, and deep learning, which employs neural networks that mimic brain structure and function to analyze data. Some experts believe the way AI systems reach a conclusion should be transparent, or explainable, while others disagree. Most AI platforms in thyroid disease have focused on malignancy risk stratification of nodules. To date, four have been approved by the United States Food and Drug Administration. While the results of validation studies have been mixed, there is ample evidence that AI can exceed the performance of some humans, particularly physicians with less experience. AI has also been applied to assessment of lymph nodes and cytopathology specimens. Conclusions: Adoption of AI in thyroid disease will require vendors to demonstrate that their software works as intended, is readily usable in real-world settings, and is cost effective. AI platforms that perform best in head-to-head comparisons will dominate and spur wider adoption.

Exploring a global interpretation mechanism for deep learning networks when predicting sepsis.

The purpose of this study is to identify additional clinical features for sepsis detection through the use of a novel mechanism for interpreting black-box machine learning models trained and to provide a suitable evaluation for the mechanism. We use the publicly available dataset from the 2019 PhysioNet Challenge. It has around 40,000 Intensive Care Unit (ICU) patients with 40 physiological variables. Using Long Short-Term Memory (LSTM) as the representative black-box machine learning model, we adapted the Multi-set Classifier to globally interpret the black-box model for concepts it learned about sepsis. To identify relevant features, the result is compared against: (i) features used by a computational sepsis expert, (ii) clinical features from clinical collaborators, (iii) academic features from literature, and (iv) significant features from statistical hypothesis testing. Random Forest was found to be the computational sepsis expert because it had high accuracies for solving both the detection and early detection, and a high degree of overlap with clinical and literature features. Using the proposed interpretation mechanism and the dataset, we identified 17 features that the LSTM used for sepsis classification, 11 of which overlaps with the top 20 features from the Random Forest model, 10 with academic features and 5 with clinical features. Clinical opinion suggests, 3 LSTM features have strong correlation with some clinical features that were not identified by the mechanism. We also found that age, chloride ion concentration, pH and oxygen saturation should be investigated further for connection with developing sepsis. Interpretation mechanisms can bolster the incorporation of state-of-the-art machine learning models into clinical decision support systems, and might help clinicians to address the issue of early sepsis detection. The promising results from this study warrants further investigation into creation of new and improvement of existing interpretation mechanisms for black-box models, and into clinical features that are currently not used in clinical assessment of sepsis.

Acute and 28-day repeated dose toxicology studies in mice with aryloxyalkanoate dioxygenase (AAD-1) protein expressed in 2,4-D tolerant DAS-40278-9 maize.

DAS-40278-9 maize (corn) plants have been genetically modified by the insertion of the aad-1 gene (aryloxyalkanoate dioxygenase), which confers tolerance to 2,4-dichlorophenoxyacetic acid (2,4-D) and aryloxyphenoxypropionate (AOPP) acetyl coenzyme A carboxylase (ACCase) inhibitors ("fop" herbicides) to enable the effective use of these herbicides on maize. The aad-1 gene, derived from Sphingobium herbicidovorans, encodes the aryloxyalkanoate dioxygenase (AAD-1) enzyme. As part of the safety assessment of the AAD-1 protein expressed in maize, acute and repeated dose mammalian toxicology studies were conducted. AAD-1 protein (heterologously produced) was orally administered to mice at a dose of 2000mg/kg, and no acute lethality or adverse effects were observed. Similarly, no adverse effects were observed in mice in a 28-day repeated-dose dietary toxicity study that incorporated the AAD-1 protein into diets at concentrations up to 1000-fold greater than the highest estimate of human exposure to maize. These results support the conclusion that the AAD-1 protein, as expressed in biotechnology derived DAS-40278-9 maize, represents a negligible risk to human health.

External validation of AIBx, an artificial intelligence model for risk stratification, in thyroid nodules.

Background: Artificial intelligence algorithms could be used to risk-stratify thyroid nodules and may reduce the subjectivity of ultrasonography. One such algorithm is AIBx which has shown good performance. However, external validation is crucial prior to clinical implementation. Materials and methods: Patients harboring thyroid nodules 1-4 cm in size, undergoing thyroid surgery from 2014 to 2016 in a single institution, were included. A histological diagnosis was obtained in all cases. Medullary thyroid cancer, metastasis from other cancers, thyroid lymphomas, and purely cystic nodules were excluded. Retrospectively, transverse ultrasound images of the nodules were analyzed by AIBx, and the results were compared with histopathology and Thyroid Imaging Reporting and Data System (TIRADS), calculated by experienced physicians. Results: Out of 329 patients, 257 nodules from 209 individuals met the eligibility criteria. Fifty-one nodules (20%) were malignant. AIBx had a negative predictive value (NPV) of 89.2%. Sensitivity, specificity, and positive predictive values (PPV) were 78.4, 44.2, and 25.8%, respectively. Considering both TIRADS 4 and TIRADS 5 nodules as malignant lesions resulted in an NPV of 93.0%, while PPV and specificity were only 22.4 and 19.4%, respectively. By combining AIBx with TIRADS, no malignant nodules were overlooked. Conclusion: When applied to ultrasound images obtained in a different setting than used for training, AIBx had comparable NPVs to TIRADS. AIBx performed even better when combined with TIRADS, thus reducing false negative assessments. These data support the concept of AIBx for thyroid nodules, and this tool may help less experienced operators by reducing the subjectivity inherent to thyroid ultrasound interpretation.

Evaluation of the amphibian metamorphosis assay: exposure to the goitrogen methimazole and the endogenous thyroid hormone L-thyroxine.

The U.S. Environmental Protection Agency (U.S. EPA) has included an amphibian metamorphosis assay (AMA) to detect thyroid active chemicals in Tier 1 testing of their endocrine screening program. To understand the variability, specificity, and reliability of the key endpoints of this assay, two exposure studies with Xenopus laevis tadpoles were conducted with two known thyroid-active compounds, namely, methimazole or L-thyroxine, for a total of 21 d. In addition, various increased-flow-rate treatments were included in the exposures to evaluate the effects of physical stress on metamorphic development. The endpoints examined in the exposures were wet weight, snout-vent length, hind-limb length, developmental stage, and thyroid and gonadal histopathology. As expected, the results indicated that both methimazole and L-thyroxine were thyroid active in the AMA, hind-limb length and thyroid histopathology being the most sensitive endpoints of thyroid activity. Tadpoles that were exposed to the various physical stressors in these experiments showed no signs of altered metamorphic development, and exposure to the thyroid-active compounds had no effect on the developing gonad of X. laevis. Taken together, these results support the use of the AMA as a Tier 1 endocrine screen for detection of potential thyroid pathway activity; however, the lack of a true negative response (no-effect) during the validation process prevents a full evaluation of this assay's specificity at this time.

AIBx, Artificial Intelligence Model to Risk Stratify Thyroid Nodules.

Background: Current classification systems for thyroid nodules are very subjective. Artificial intelligence (AI) algorithms have been used to decrease subjectivity in medical image interpretation. One out of 2 women over the age of 50 years may have a thyroid nodule and at present the only way to exclude malignancy is through invasive procedures for those that are suspicious on ultrasonography. Hence, there exists a need for noninvasive objective classification of thyroid nodules. Some cancers have benign appearance on ultrasonogram. Hence, we decided to create an image similarity algorithm rather than image classification algorithm. Materials and Methods: Ultrasound images of thyroid nodules from patients who underwent either biopsy or thyroid surgery from February 2012 to February 2017 in our institution were used to create AI models. Nodules were excluded if there was no definitive diagnosis of it being benign or malignant. A total of 482 nodules met the inclusion criteria and all available images from these nodules were used to create the AI models. Later, these AI models were used to test 103 thyroid nodules that underwent biopsy or surgery from March 2017 to July 2018. Results: Negative predictive value (NPV) of the image similarity model was 93.2%. Sensitivity, specificity, positive predictive value (PPV), and accuracy of the model were 87.8%, 78.5%, 65.9%, and 81.5%, respectively. Conclusions: When compared with published results of ultrasound thyroid cancer risk stratification systems, our image similarity model had comparable NPV with better sensitivity, specificity, and PPV. By using image similarity AI models, we can decrease subjectivity and decrease the number of unnecessary biopsies. Using image similarity AI model, we were able to create an explainable AI model that increases physician's confidence in the predictions.

Use of artificial intelligence and machine learning for estimating malignancy risk of thyroid nodules.

PURPOSE OF REVIEW: Current methods for thyroid nodule risk stratification are subjective, and artificial intelligence algorithms have been used to overcome this shortcoming. In this review, we summarize recent developments in the application of artificial intelligence algorithms for estimating the risks of malignancy in a thyroid nodule. RECENT FINDINGS: Artificial intelligence have been used to predict malignancy in thyroid nodules using ultrasound images, cytopathology images, and molecular markers. Recent clinical trials have shown that artificial intelligence model's performance matched that of experienced radiologists and pathologists. Explainable artificial intelligence models are being developed to avoid the black box problem. Risk stratification algorithms using artificial intelligence for thyroid nodules are now commercially available in many countries. SUMMARY: Artificial intelligence models could become a useful tool in a thyroidolgist's armamentarium as a decision support tool. Increased adoption of this emerging technology will depend upon increased awareness of the potential benefits and pitfalls in using artificial intelligence.

Clinical performance of multiplatform mutation panel and microRNA risk classifier in indeterminate thyroid nodules.

INTRODUCTION: We evaluated the clinical performance of an expanded mutation panel in combination with microRNA classification (MPTX) for the management of indeterminate thyroid nodules. MATERIALS AND METHODS: MPTX included testing of fine-needle aspirates from multiple centers with a combination of ThyGeNEXT mutation panel for strong and weak driver oncogenic changes and ThyraMIR microRNA risk classifier (both from Interpace Diagnostics; Pittsburgh, PA). MPTX test status (positive or negative) and MPTX clinical risk classifications (low, moderate, or high risk) were determined blind to patient outcomes. Surgical pathology and clinical follow-up records of patients from multiple centers were used to determine patient outcomes. MPTX performance was assessed by Kaplan Meier analysis for cancer-free survival of patients, with risk of malignancy determined by hazard ratio (HR). RESULTS: Our study included 140 patients with AUS/FLUS or FN/SFN nodules, of which 13% had malignancy. MPTX negative test status and MPTX low risk results conferred a high probability (94%) that patients would remain cancer-free. MPTX positive test status (HR 11.2, P < 0.001) and MPTX moderate-risk results (HR 8.5, P = 0.001) were significant risk factors for malignancy, each conferring a 53% probability of malignancy. MPTX high-risk results elevated risk of malignancy even more so, conferring a 70% probability of malignancy (HR 38.5, P < 0.001). CONCLUSIONS: MPTX test status accurately stratifies patients for risk of malignancy. Further classification using MPTX clinical risk categories enhances utility by accurately identifying patients at low, moderate, or high risk of malignancy at the low rate of malignancy encountered when clinically managing patients with indeterminate thyroid nodules.