Emily Rachel Christison-Lagay

Context: Differentiated thyroid carcinoma (DTC) is rare in children, comprising 2% to 4% of pediatric cancers. Because of low case volumes at individual institutions, comprehensive data on patient and tumor characteristics and treatment outcomes are limited. The Child and Adolescent Thyroid Consortium was established in 2019 to address this gap and to assess impact of changes in practice on outcomes. Objective: This inaugural Child and Adolescent Thyroid Consortium study analyzes the landscape of pediatric DTC and predictors of invasive disease. Design: International, multicenter retrospective analysis of children (<age 19 years) with DTC diagnosed from 2010 to 2019. Setting: Five high-volume pediatric centers. Results: The cohort included 715 children (78.3% female) diagnosed with DTC at median age 15.2 years (range: 3.2-18.9). There were 673 (94%) patients with papillary thyroid carcinoma (PTC) and 42 (6%) had follicular thyroid carcinoma. Among patients with PTC, stage 1(M0) disease was present in 562 individuals (83.5%; 222 N0, 161 N1a, and 179 N1b). Stage 2 (M1) disease was present in 111 (16.5%) children with PTC and 1/42 (2.4%) with follicular thyroid carcinoma. Males and children aged <10 years exhibited more advanced T, N, and M stages. Any extrathyroidal extension correlated with increased nodal and pulmonary metastases. Oncogenic fusions, compared with BRAFV600E, were associated with more advanced N and M stages. Deintensification of therapy since 2015 was not associated with changes in disease outcomes. Conclusion: Invasive DTC is associated with age <10 years, male sex, tumors driven by oncogenic fusions, and extrathyroidal extension. Despite less aggressive therapies in recent years, disease outcomes have not worsened. Recognition of these associations may guide care and prognostication.

Overview of the transformer architecture used in this paper.

Comparing the size, training time, and inference time differences between the SAMPLER model and the Transformer based model.

Benchmarking various foundation models for histology feature extraction and the downstream task of ERMS vs. ARMS.

Performance comparison of CLAM vs SAMPLER-STQ end to end pipelines on holdout institution predictions.

Metrics for SAMPLER on the two stage Ewing sarcoma pipeline.

Pediatric sarcomas present diagnostic challenges due to their rarity and diverse subtypes, often requiring specialized pathology expertise and costly genetic tests. To overcome these barriers, we developed a computational pipeline leveraging deep learning methods to accurately classify pediatric sarcoma subtypes from digitized histology slides. To ensure classifier generalizability and minimize center-specific artifacts, a dataset comprising 867 whole-slide images (WSI) from three medical centers and the Children’s Oncology Group was collected and harmonized. Multiple convolutional neural network and vision transformer (ViT) architectures were systematically evaluated as feature extractors for SAMPLER-based WSI representations, and input parameters, such as tile size combinations and resolutions, were tested and optimized. The analysis showed that advanced ViT foundation models (UNI and CONCH) significantly outperformed earlier approaches, and incorporating multiscale features enhanced classification accuracy. The optimized models achieved high performance, distinguishing rhabdomyosarcoma (RMS) from non-RMS soft-tissue sarcomas (NRSTS) with an AUC of 0.969 and differentiating RMS subtypes (alveolar vs. embryonal) with an AUC of 0.961. Additionally, a two-stage pipeline effectively identified scarce Ewing sarcoma images from other NRSTS (AUC = 0.929). Compared with conventional transformer-encoder architectures used for WSI representations, these SAMPLER-based classifiers were three orders of magnitude faster to train, despite operating entirely without a graphical processing unit. This study highlights that digital histopathology paired with rigorous image harmonization provides a powerful solution for pediatric sarcoma classification.Significance:An approach pairing a multi-institutional dataset with a published pipeline extends imaging-based diagnostic capabilities and creates the potential for accurate, rapid cancer diagnoses across resource-limited and remote settings.This article is part of a special series: Driving Cancer Discoveries with Computational Research, Data Science, and Machine Learning/AI.

Results for models trained to classify fusion status in RMS.

Zero-shot classification of adult sarcoma and the task of RMS vs. NRSTS prediction.