Cytomolecular Classification of Thyroid Nodules Using Fine-Needle Washes Aspiration Biopsies

Abstract

Fine-needle aspiration biopsies (FNA) represent the gold standard to exclude the malignant nature of thyroid nodules. After cytomorphology, 20–30% of cases are deemed “indeterminate for malignancy” and undergo surgery. However, after thyroidectomy, 70–80% of these nodules are benign. The identification of tools for improving FNA’s diagnostic performances is explored by matrix-assisted laser-desorption ionization mass spectrometry imaging (MALDI-MSI). A clinical study was conducted in order to build a classification model for the characterization of thyroid nodules on a large cohort of 240 samples, showing that MALDI-MSI can be effective in separating areas with benign/malignant cells. The model had optimal performances in the internal validation set (n = 70), with 100.0% (95% CI = 83.2–100.0%) sensitivity and 96.0% (95% CI = 86.3–99.5%) specificity. The external validation (n = 170) showed a specificity of 82.9% (95% CI = 74.3–89.5%) and a sensitivity of 43.1% (95% CI = 30.9–56.0%). The performance of the model was hampered in the presence of poor and/or noisy spectra. Consequently, restricting the evaluation to the subset of FNAs with adequate cellularity, sensitivity improved up to 76.5% (95% CI = 58.8–89.3). Results also suggest the putative role of MALDI-MSI in routine clinical triage, with a three levels diagnostic classification that accounts for an indeterminate gray zone of nodules requiring a strict follow-up.

Intraoperative Identification of Thyroid and Parathyroid Tissues During Human Endocrine Surgery Using the MasSpec Pen

“…Interestingly, the probability value obtained for the misclassified analysis (0.52) was close to the prediction cutoff value of 0.50, compared to the second and third correctly classified analyses (0.21 and 0.09). This observation prompted us to explore implementing a prediction gray zone, a concept previously explored in machine learning and mass spectrometry . Essentially, prediction probability values within 0.2 of the cutoff (0.30-0.70) were considered of lower confidence and not classified, while probabilities values outside of that range are considered of higher confidence and classified.…”

“…This observation prompted us to explore implementing a prediction gray zone, a concept previously explored in machine learning and mass spectrometry. 29,38,39 Essentially, prediction probability values within 0.2 of the cutoff (0.30-0.70) were considered of lower confi-dence and not classified, while probabilities values outside of that range are considered of higher confidence and classified. When comparing the probability values for the test set (Figure 3B), 26 of the 184 analyses fell within the gray zone, with 11 of those (42%) misclassified.…”

The journey towards clinical adoption of MALDI‐MS‐based imaging proteomics: from current challenges to future expectations

“…Despite the FNA thyroid samples used in the validation set (n = 170) included samples with inadequate cellularity, the approach showed a specificity of 82.9% and a sensitivity of 43.1%, whilst when the analysis was focused only on a subset of FNAs with adequate cellularity, sensitivity increased to 76.5%. More recently, Capitoli et al [75] suggested a novel workflow, which included MALDI-MSI in the clinical routine, to be used in cases of indeterminate diagnosis. In particular, the authors suggested a three-level diagnostic classification for indeterminate nodules in the pixel-by-pixel approach based on the percentage of malignant and benign pixels present in the whole sample: a number of malignant pixels lower than 7 suggest a benign sample and a ultrasound follow-up at 12 months; a number of malignant pixels higher than 16.7% suggest a malignant sample, and thus a thyroidectomy, whilst in instances where the number of malignant pixels was between 7.0% and 16.7% (defined as a grey zone), this identified nodules which will require a strict ultrasound follow-up, and, eventually associated with a repeat biopsy [75].…”

The application of high-throughput proteomics in cytopathology

“…To complement the morphologic evaluation of FNA in the evaluation of thyroid lesions, especially the ones with indeterminate interpretations, a few studies utilized in situ proteomics, more specifically the MALDI-mass spectrometry imaging (MSI) technique [48][49][50][51][52][53]. For instance, MALDI-MSI distinguished benign thyroid lesions from papillary thyroid carcinomas (PTCs) and correctly triaged indeterminate FNA lesions as either benign or malignant [51], while it also distinguished Hashimoto thyroiditis from hyperplastic nodules and PTC in another study [50].…”