Deep learning-based hyperspectral technique identifies metastatic lymph nodes in oral squamous cell carcinoma-A pilot...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Metastatic lymph nodes are critical indicators of cancer spread.
• Accurate detection of cancer cells in lymph nodes is essential for effective treatment planning.
• Current methods for lymph node metastasis detection can be time-consuming and subjective.

Purpose of the Study:

• To develop a hyperspectral imaging (HSI) and deep learning system for detecting cancer cells in metastatic lymph nodes.
• To differentiate between cancer cells, lymphocytes, and normal tissue using spectral analysis.
• To improve the accuracy and efficiency of diagnosing oral squamous cell carcinoma (OSCC) metastasis.

Main Methods:

• Collected continuous sections of metastatic lymph nodes from 45 OSCC patients.
• Developed an improved ResUNet algorithm for deep learning analysis.
• Analyzed spectral curve differences between various cell types and tissues using HSI.

Main Results:

• Distinguished cancer cells, lymphocytes, and erythrocytes in metastatic lymph nodes with 87.30% overall accuracy and 85.46% average accuracy.
• Achieved an average intersection over union (IOU) of 0.6253 and accuracy of 0.7692 in recognizing cancerous areas.
• Demonstrated the capability of HSI and deep learning to identify tumor tissue within lymph nodes.

Conclusions:

• Deep learning-based hyperspectral techniques can accurately identify tumor tissue in OSCC metastatic lymph nodes.
• The developed system offers high accuracy in pathological diagnosis, improving work efficiency and reducing workload.
• These preliminary findings highlight the potential of HSI and AI in cancer diagnostics, warranting further investigation with larger sample sizes.