Time: October 15, 2026

Venue: Lanzhou, Gansu Province, China

Working Language: English

Objectives:

The core objective of this training is not only to empower participants with technical capacity, but also to systematically impart the full-chain technologies of “radiomics feature engineering–deep learning model development–clinical validation and deployment.” This will enable participants to ｍａｓｔｅｒ the independently develop cross-disease benign and malignant differentiation models, achieve local adaptation, and apply technical solutions such as few-shot learning in resource-limited scenarios., thereby reducing reliance on high-performance computing  and large-scale labeled data. In addition, the program aims to promote the establishment of a “Medical Imaging AI Collaboration Alliance” across Asia, Africa, and Latin America to support standardized data sharing and multi-center clinical validation, accelerating technology implementation.

Programs:

This training builds a full-chain medical imaging AI system of “data–algorithm–clinical.”

Module One: Data Standardization (DICOM parsing, multimodal registration), Radiomics feature engineering, and Advances in few-shot and Federated Learning (FATE Framework) Technologies; 

Module 2: Based on the LIDC-IDRI /TCIA datasets, taking pulmonary nodules (3D CNN+ radiomics), breast tumors (multi-sequence fusion), and bone tumors (Enneking staging) as examples, the interpretability of the model was verified through Grad-CAM/SHAP; 

Module 3: TensorRT quantization, ONNX cross-platform deployment and Jetson edge computing development, building the FHIR standard federated learning platform; 

Module 4: Follow the WHO ethical guidelines, use ROC/Delong analysis to mitigate bias, and guide the development of AI tools for tuberculous spondylitis (MD Anderson/ Oxford Expert Review). The entire course adopts a three-dimensional linkage of “theory–code–hardware”， forming a closed loop of data annotation–model training–clinical deployment, achieving a breakthrough in the clinical transformation of AI diagnosis for multiple diseases.

Organizer: Gansu Provincial Hospital

Address: No. 204, Donggang West Road, Lanzhou City, Gansu Province, China

Postcode: 730000

Coordinator: Sheng Zhou

Tel: +86-931-8282002

Fax: +86-931-8282002

E-mail: lzzs@sina.com