Publications by authors named "Zuxuan Zhao"
Article Synopsis
- Breast cancer is the most common type of cancer, and intraoperative frozen sections help determine if lymph node dissection is needed during surgery.
- A deep-learning network called the RRCART model was proposed to identify metastases in frozen sections, analyzing 2362 samples to find and categorize errors.
- The study found 101 incorrect identifications, with the highest error rates linked to normal lymph node structure and micrometastasis, leading to a framework that addresses causes of inaccuracies in the artificial intelligence model.
Clinicopathologic characteristics of thymic clear cell carcinoma: a case report with literature review.
J Cardiothorac Surg
February 2023
Background: Thymic clear cell carcinoma is a rare mediastinal neoplasm, with only 25 reported cases to date. We report a case of a 45-year-old man with thymic clear cell carcinoma. We think imaging and laboratory tests may be helpful for differential diagnosis.
View Article and Find Full Text PDFArticle Synopsis
- In China, frozen section diagnoses are critical for determining breast cancer metastasis in sentinel lymph nodes, but deep neural networks (DNNs) can misdiagnose, impacting patient treatment.
- This study evaluates a new method called relative risk classification and regression tree (RRCART) to better identify misdiagnosed whole-slide images (WSIs) from a DNN and recommends review by pathologists.
- Results from testing 2,362 WSIs show that while the RRCART method achieved high accuracy rates, it also revealed a significantly higher misdiagnosis rate for the DNN compared to human pathologists, indicating the need for pathologist review of low-accuracy cases.