Leveraging deep learning for reconstruction of cancer cells from 3D IR thermography for faster cancer diagnosis

Cancer remains a leading cause of death worldwide, underscoring the critical need for early detection methods. This research addresses this challenge by leveraging deep neural networks to reconstruct cancer cells using 3D Infrared (IR) thermography. Our novel approach captures angular radiation by projecting infrared rays from multiple angles, obtaining more comprehensive information compared to traditional 0-degree measurements. We trained neural networks to predict excitation angles, defect sizes, and defect lengths, which are integrated to create 2D defect images. The Filtered Back Projection (FBP) algorithm enhances these images by constructing hidden angular views, resulting in a detailed 3D reconstruction of affected cells. Real test data is captured using micro-imager sensor cameras with micro lenses, while simulated data from COMSOL is used for training and validation. Achieving a detailed tissue reconstruction, this method marks a significant advancement in early cancer detection, potentially improving patient outcomes.