A deep learning-based system trained for gastrointestinal stromal tumor screening can identify multiple types of soft tissue tumors.

Researchers

Journal

Modalities

Models

Abstract

The accuracy and timeliness of the pathological diagnosis of soft tissue tumors (STTs) critically affect treatment decision and patient prognosis. Thus, it is crucial to make a preliminary judgement on whether the tumor is benign or malignant with hematoxylin-and-eosin (H&E)-stained images. Here, we present a deep learning-based system, STT-BOX, with only H&E images for malignant STTs identification from benign STTs with histopathological similarity. STT-BOX assumed gastrointestinal stromal tumor (GIST) as a baseline for malignant STT evaluation, and distinguished GIST from leiomyoma and Schwannoma with 100% AUC in patients from three hospitals, which achieved higher accuracy than the interpretation of experienced pathologists. Particularly, this system performed well on six common types of malignant STTs from TCGA dataset, accurately highlighting the malignant mass lesion. Moreover, without any fine-tuning, STT-BOX was capable to distinguish ovarian malignant sex-cord stromal tumors. Our study includes mesenchymal tumors originated from the digestive system, bone and soft tissues, and reproductive system, where the high accuracy of migration verification may reveal the morphological similarity of the nine types of malignant tumors. Further evaluation in a pan-STT setting would be potential and prospective, obviating the overuse of immunohistochemistry and molecular tests, and providing a practical basis for clinical treatment selection in a timely manner.Copyright © 2023. Published by Elsevier Inc.