Multiscale Time-Sharing Elastography Algorithms and Transfer Learning of Clinicopathological Features of Uterine...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Computational Pathology

Background:

• Intelligent medical systems face challenges in complex object recognition, large-scale imaging, and real-time diagnosis.
• Existing systems exhibit poor real-time computing, inefficient data storage, and low lesion recognition rates, particularly in cervical cancer pathology.

Purpose of the Study:

• To propose an efficient and reliable medical intelligent computing system for cervical cancer pathology analysis.
• To develop novel algorithms addressing challenges in multi-scale imaging and transfer learning for complex medical data.

Main Methods:

• Designed a multi-scale time-sharing elastic imaging algorithm considering data dimensions and imaging errors.
• Developed a transfer learning algorithm for clinical and pathological features of cervical cancer.
• Established a medical intelligent computing system for pathology analysis and calculation.

Main Results:

• The proposed algorithms demonstrated superior performance in cervical cancer pathological imaging and scoring compared to single-scale Retinex (SSR).
• Enhanced feature extraction and lesion recognition capabilities were observed.
• The system execution efficiency significantly outperformed the comparison algorithm (SSR-TL).

Conclusions:

• The developed medical intelligent computing system and algorithms offer a high-efficiency and reliable solution for cervical cancer pathology.
• The multi-scale imaging and transfer learning framework effectively addresses limitations in current intelligent medical diagnosis systems.