Clinical Imaging

I. Multimodal Cancer Molecular Subtyping Study

This project employs multimodal data for tasks such as predicting cancer molecular subtypes and sub-classifications, by deeply exploring both the common and specific features of tumors across various modalities to elucidate the intrinsic mechanisms of cancer.

II. Clinical CT Imaging Study for Gastric Cancer Diagnosis

Vascular invasion, referring to cancer cells infiltrating blood vessels and lymphatic vessels, is a critical indicator affecting the treatment and survival prognosis of gastric cancer. Traditionally determined through invasive biopsy followed by pathological examination, developing non-invasive methods to ascertain this is immensely valuable. The objective of this study is to investigate annotated tumor cycle CT images using computer-aided diagnostic techniques, aiming to predict the presence of vascular invasion and lymph node metastasis in patients' gastric regions, thereby facilitating more optimal treatment strategies.

III. Clinical CT Imaging Study for Lung Cancer Diagnosis

Lung cancer, a malignant tumor originating from lung tissue, has a high incidence and mortality rate among cancers, posing a severe threat to human health. With diverse histological types including squamous cell carcinoma, adenocarcinoma, large cell carcinoma, and small cell carcinoma, each with distinct etiology, progression patterns, and clinical manifestations significantly impacting patient treatment and outcomes, early detection and effective intervention can greatly enhance survival rates. This project aims to introduce a deep learning-based CT imaging diagnosis method for lung cancer, utilizing image classification technology to classify clinical CT data, assisting radiologists in clinical practice and enhancing workflow efficiency.

IV. Raman Spectroscopy Breast Cancer Diagnosis Research

Breast cancer, one of the cancers with high mortality rates globally and particularly prevalent among women, necessitates effective early diagnosis to alleviate patients' psychological and economic burdens. Non-invasive Raman spectroscopy has been widely adopted for this purpose. Through Raman analysis of samples from patients (frozen sections) and cells (single-cell suspensions), this study analyzes frequency shifts in the obtained Raman spectra to study the structure and properties of the analyzed materials. By examining the positions of Raman shift peaks, it determines whether the nature of the material has changed, indicating the presence of disease. With a dataset of spectra and employing neural networks for feature extraction, the goal is to improve early breast cancer diagnosis accuracy and reduce patient burden.

V. Multi-Omics Research

Omics technologies hold enormous potential in cancer research, offering unparalleled opportunities to depict cancer biology at various pathological and molecular levels. Common omics fields include genomics, transcriptomics, and proteomics. Genomics investigates an organism’s complete genetic information; transcriptomics examines the composition and changes in RNA within cells under specific conditions, reflecting gene expression levels; proteomics studies the expression, modification, and interactions of all proteins in organisms, cells, or tissues. This project seeks to integrate and analyze data from genomics, transcriptomics, and proteomics to comprehensively reveal information about cancer’s onset mechanisms, diagnostics, prognostic assessment, and therapy selection, contributing further to precision medicine in oncology.