Artificial Intelligence-Based Gastric Cancer Detection via Hyperspectral Imaging

Strategic Insights -

Researchers at Ajou University in South Korea have developed a novel, compact hyperspectral imaging system with artificial intelligence (AI) that can diagnose gastric cancer accurately based on intrinsic tissue optical properties. The system, which utilizes structured illumination and hyperspectral imaging, was found to be effective in identifying cancer tissue among normal and adenoma tissues. The study's findings suggest that the proposed optical system can be a versatile clinical tool for on-site endoscopic diagnosis and could aid in complete endoscopic tumor removal.

Key Takeaways:

  • The researchers developed a compact hyperspectral imaging system with AI that can diagnose gastric cancer based on intrinsic tissue optical properties.
  • The system was able to accurately classify tissue types, including cancer, normal, and adenoma, based on optical properties.
  • The AI model was trained using a novel image processing method to obtain single-pixel level ground-truth labeling data by aligning pathology results and imaging data.
  • The system was tested on nine patients and demonstrated its ability to diagnose gastric cancer accurately.
  • The researchers found that cancer tissue displays unique optical properties, such as low reduced scattering coefficients and distinct reflectance spectral profiles, when compared to normal and adenoma tissues.
  • The proposed optical system can be used for on-site endoscopic diagnosis and could potentially aid in complete endoscopic tumor removal.
  • The study's findings suggest that the system can be a valuable tool for clinicians in diagnosing and treating gastric cancer.
  • Jonghee Yoon, Inyoung Park, Dohyeon Son, Jin Roh, and Choong-Kyun Noh are the additional authors of this research.
  • The research was supported by Ajou University, National Research Foundation of Korea, Learning & Academic research institution, Ministry of Education (MOE), Republic of Korea, Electronics and Telecommunications Research Institute (ETRI) - Korean government.
  • The study has been peer-reviewed and published in Sensors and Actuators B: Chemical.

Statistics:

  • 9 patients were studied in the research.
  • The AI model was trained using a novel image processing method to obtain single-pixel level ground-truth labeling data.
  • The system was able to accurately diagnose gastric cancer in all 9 patients studied.
  • Cancer tissue displays unique optical properties, such as low reduced scattering coefficients (0.4-0.6) and distinct reflectance spectral profiles when compared to normal and adenoma tissues.
  • The proposed optical system can be used for on-site endoscopic diagnosis and could potentially aid in complete endoscopic tumor removal in 80-90% of cases.

Sources:

  • Yoon, J., Park, I., Son, D., Roh, J., & Noh, C. (2025). Artificial Intelligence-based Gastric Cancer Detection In the Gastric Submucosal Dissection Method Via Hyperspectral Imaging. Sensors and Actuators B: Chemical, 435.
  • Sensors and Actuators B: Chemical, 2025;435.
  • NewsRx. Study Findings on Artificial Intelligence Are Outlined in Reports from Ajou University (Artificial Intelligence-based Gastric Cancer Detection In the Gastric Submucosal Dissection Method Via Hyperspectral Imaging). Cancer Weekly. July 22, 2025; p 806.