Generating a multimodal artificial intelligence model to differentiate benign and malignant follicular neoplasms of the thyroid: A proof-of-concept study.

BACKGROUND: Machine learning has been increasingly used to develop algorithms that can improve medical diagnostics and prognostication and has shown promise in improving the classification of thyroid ultrasound images. This proof-of-concept study aims to develop a multimodal machine-learning model to classify follicular carcinoma from adenoma. METHODS: This is a retrospective study of patients with follicular adenoma or carcinoma at a single institution between 2010 and 2022. Demographics, imaging, and perioperative variables were collected. The region of interest was annotated on ultrasound and used to perform radiomics analysis. Imaging features and clinical variables were then used to create a random forest classifier to predict malignancy. Leave-one-out cross-validation was conducted to evaluate classifier performance using the area under the receiver operating characteristic curve. RESULTS: Patients with follicular adenomas (n = 7) and carcinomas (n = 11) with complete imaging and perioperative data were included. A total of 910 features were extracted from each image. The t-distributed stochastic neighbor embedding method reduced the dimension to 2 primary represented components. The random forest classifier achieved an area under the receiver operating characteristic curve of 0.76 (clinical only), 0.29 (image only), and 0.79 (multimodal data). CONCLUSION: Our multimodal machine learning model demonstrates promising results in classifying follicular carcinoma from adenoma. This approach can potentially be applied in future studies to generate models for preoperative differentiation of follicular thyroid neoplasms.

How Effective is the Use of Molecular Testing in Preoperative Decision Making for Management of Indeterminate Thyroid Nodules?

INTRODUCTION: We performed Thyroseq v2 molecular testing on indeterminate thyroid nodules and evaluated whether they underwent a management change from the standard of thyroid lobectomy. METHODS: We conducted a retrospective analysis of all indeterminate thyroid nodules that underwent Thyroseq v2 molecular testing from 2014 to 2019 at a large academic center. Pathology was reviewed by thyroid cytopathologists. Thyroseq results were reported benign (malignancy probability less than 10%) or suspicious (malignancy probability greater than 30%). The primary endpoint was a management change from a diagnostic lobectomy. RESULTS: A total of 142 nodules were included: 113 (80%) Bethesda III and 29 (20%) Bethesda IV. Seventy-three nodules underwent surgical management and 69 did not. We noted a change in management in 64% (91/142) of nodules. Patients who underwent a change in management to no surgery had a significantly higher rate of benign Thyroseq result than those without a change (75.8% vs. 49.0%, p = 0.001). On logistic regression analysis, a benign Thyroseq result was a positive independent predictor of a change to no surgery (OR 3.87, 95% CI 1.69-8.89). Nodule size, multiple nodules, compressive symptoms, and history of hypothyroidism were not significant. Of the 91 patients who underwent a management change, 71% (65/91) did not undergo surgery. On follow-up (average 985 ± 615 days), 12% (8/65) of those nodules were growing or developed suspicious features requiring surgery. CONCLUSIONS: Molecular testing helped avoid surgery in almost half our population with indeterminate thyroid nodules, and benign results may help avoid surgery in asymptomatic patients with indeterminate thyroid nodules.

The Role of Surgeon-Performed Office and Preincision Ultrasounds in Localization of Parathyroid Adenomas in Primary Hyperparathyroidism.

OBJECTIVE: We studied the use of surgeon-performed office ultrasound (OU) and preincision ultrasound (PIU) in preoperatively localizing parathyroid adenomas in primary hyperparathyroidism (PHPT). METHODS: A retrospective chart review was performed for patients with PHPT who underwent parathyroidectomy between 2013 and 2015. The results of OU and PIU were recorded and compared with the final surgical pathology. RESULTS: Of 348 patients with PHPT, 285 (81.9%) had single-lesion disease, 49 (14.1%) had double-lesion disease, and 14 (4.0%) had multigland disease with 3 or more lesions. For single-lesion disease, the overall sensitivity and specificity of OU to correctly lateralize the lesion were 64.2% and 91.2%, while those of PIU were 89.4% and 93.6%, respectively. The sensitivity and specificity of PIU were comparable to those of 4-dimensional computed tomography (87.1% and 90.7%, respectively) and 99mTc-sestamibi scintigraphy (70.4% and 95.9%, respectively). While the majority of PIU cases were preceded by other imaging studies, the accuracy in localizing lesions was not largely affected by the presence of prior computed tomography and/or 99mTc-sestamibi scintigraphy, as opposed to ultrasounds only. For detecting the presence of multigland disease, the sensitivity and specificity of OU were 26% and 92.2%, while those of PIU were 64.3% and 94.7%, respectively. CONCLUSION: Surgeon-performed OU and PIU are valuable tools in preoperatively localizing the parathyroid adenoma in single-lesion disease, while their utility may be limited for double-lesion or multigland disease. PIU in particular yields high accuracy in detecting parathyroid lesions in combination with other imaging modalities.

Transoral Endoscopic Thyroidectomy Vestibular Approach: A Single-institution Experience of the First 50 Cases.

Background. Transoral endoscopic thyroidectomy vestibular approach (TOETVA) has evolved from a novel procedure to a cosmetically appealing alternative to the traditional open thyroidectomy. TOETVA is limited to only high-volume centers with the hope to expand and demonstrate the safety, reproducibility, and application for the correct patient with thyroid disease. We present our experience with the first 50 TOETVA procedures performed at our institution. Methods. From September 2016 to June 2018, we performed 50 transoral endoscopic thyroidectomies via the vestibular approach for a variety of thyroid pathologies. Outcomes were analyzed from a prospectively maintained database. Patients were treated for a different range of thyroid pathologies. Results. A total of 12 (24%) patients underwent total thyroidectomy, whereas 38 (76%) patients underwent a subtotal thyroidectomy or thyroid lobectomy. The mean surgical time was 149 ± 40.0 (90-256) minutes for lobectomy and 217.6 ± 33.3 (175-276) minutes for total thyroidectomy. Of our 50 patients, 16 (32%) had postoperative transient lower lip numbness with an average time to recovery of 23.8 (.43-48) weeks, with 1 (2%) patient having persistent, but improving, lower lip numbness beyond the 6-month follow-up. There were 13 (26%) patients with transient chin numbness with an average time to recovery of 15.7 (2-48) weeks. Two (4%) patients had transient recurrent laryngeal nerve (RLN) injury with hoarseness, whereas 1 (2%) patient had permanent injury. Conclusion. The TOETVA is a safe and reproducible procedure. For selected patients, this technique is a viable alternative to conventional thyroidectomy.

Post-thyroidectomy emergency room visits and readmissions: Assessment from the Collaborative Endocrine Surgery Quality Improvement Program (CESQIP).

BACKGROUND: This study analyzed independent factors associated with post-thyroidectomy Emergency Room (ER) visits and Hospital Readmissions (HR). METHODS: This is a retrospective review from the CESQIP registry of 8381 thyroidectomy patients by 173 surgeons at 46 institutions. A total of 7142 ER visits and 7265 HR were analyzed. Multivariable logistic regression analysis was performed to determine the risk factors for an ER visit or HR. RESULTS: Within 30-days of surgery, rates of all ER visits were 3.4% (n = 250) and all HR were 2.3% (n = 170). Hypocalcemia was the reason for 21.9% of ER encounters and 36.4% of HR. BMI >40 kg/m2 was a risk factor for both ER visit (OR1.86) and HR (OR1.94). Surgical duration >3 h (OR2.63), and transection of recurrent laryngeal nerve (OR4.58) were risk factors for HR. CONCLUSIONS: Strategies to decrease hypocalcemia and improve perioperative care of patients with BMI >40 kg/m2 may improve post-thyroidectomy outcome.

Clinical performance of a next-generation sequencing assay (ThyroSeq v2) in the evaluation of indeterminate thyroid nodules.

BACKGROUND: Molecular testing with the Thyroseq v2 next generation sequencing panel ("Thyroseq") is used to estimate the risk of cancer in indeterminate thyroid nodules. METHODS: We analyzed 156 indeterminate thyroid nodules evaluated with Thyroseq, across 3 institutions. Thyroseq data and surgical pathology were matched via pathologic re-review. A result was considered Thyroseq positive if molecular alterations were annotated on the report with malignancy probability >30%. Performance characteristics were estimated using Bayes theorem. RESULTS: The Thyroseq-negative call rate was 65% (102/156). On surgical pathology, 16% (10/63) of nodules were malignant. The positive predictive value of a Thyroseq-positive result was 22% (8/37; if 2 noninvasive follicular thyroid neoplasm with papillary-like nuclear features are counted as malignant, 27%, 10/37). There was 1 false-negative result (negative predictive value 96%, 22/23). The most common mutation was NRAS (19/37) with a positive predictive value of 7% (1/15). The positive predictive value of all RAS mutations (HRAS, KRAS, NRAS) was 9% (2/22). The second most common mutation, BRAF V600E, had positive predictive value of 100% (3/3). CONCLUSION: We report an external analysis of Thyroseq performance in the evaluation of indeterminate thyroid nodules. These data indicate that Thyroseq is likely to offer high negative predictive value but low positive predictive value. Many genetic alterations appear to be nonspecific for malignancy, and positive results should be interpreted with caution. These findings have implications for the management of indeterminate thyroid nodules profiled with Thyroseq.