Experimental Insights Into Data Augmentation Techniques for Deep Learning‐Based Multimode Fiber Imaging: Limitations and Success

Multimode fiber (MMF) imaging using deep learning has high potential to produce minimally invasive endoscopes. Nevertheless, it relies on large, real‐world medical data, whose availability is limited by privacy concerns. Although data augmentation has been extensively studied in various other deep learning tasks, it has not been explored for MMF imaging. This work provides the first experimental and computational study on the efficacy and limitations of augmentation techniques in this field. We demonstrate that standard image transformations and conditional generative adversarial‐based synthetic speckle generation fail to improve reconstruction quality in our experimental settings, as they neglect the modal interference that results in speckle formation. To address this, we introduce a physical data augmentation method in which only organ images are digitally transformed, while their corresponding speckles are experimentally acquired via fiber. This approach preserves the physics of light‐fiber interaction and not only improves reconstruction fidelity in relatively less experimental time but also makes the model robust to rotational changes in images. It enhances the reconstruction structural similarity index measure by up to 22.81%, forming a viable system for reliable MMF imaging under limited data conditions.