Using a deep neural network (DNN) that imitates the way the brain processes images, researchers reconstructed images transmitted over multimode optical fibers at distances of up to 1 km. In the same way the human brain learns to recognize and categorize images and symbols, the DNN was trained to recognize certain images (in this case, handwritten digits) until it was able to recognize other images that were from the same category. A speckle pattern from an image transmitted through a multimode fiber passes through the hidden layers of a deep neural network and is reproduced as the number 3. Courtesy of Demetri Psaltis, Swiss Federal Institute of Technology Lausanne. A team from the Swiss Federal Institute of Technology trained its DNN to recognize images of numbers from the speckle pattern that was created when the numbers were transmitted through a multimode optical fiber. To do so, it used a database containing 20,000 samples of handwritten numbers from 0 through 9. Researchers selected 16,000 samples to be used as training data, and set aside 2000 to validate the training and another 2000 to test the validated system. The team used a laser to illuminate each digit and sent the light beam through a multimode optical fiber with about 4500 channels to a camera at the end of the fiber. A computer algorithm measured how the intensity of the output light varied across the captured image. The researchers collected a series of examples for each digit. Although the speckle patterns collected for each digit looked the same to the human eye, the DNN was able to discern differences and recognize patterns of intensity associated with each digit. Testing with the set-aside images showed that the algorithm achieved 97.6 percent accuracy for images transmitted through a 0.1-meter-long fiber and 90 percent accuracy through a 1-km length of fiber. Better recognition accuracy was obtained when the DNN was trained to first reconstruct the input and then classify based on the recovered image. Researcher Navid Borhani said that the machine-learning approach is much simpler than other methods used to reconstruct images passed through optical fibers, which require making a holographic measurement of the output. The DNN was also able to cope with distortions caused by environmental disturbances to the fiber, such as temperature fluctuations. The technique could be used to boost the amount of information carried over fiber optic telecommunication networks and to improve imaging for medical diagnoses. Doctors could use ultrathin fiber probes to collect images of the tracts and arteries inside the human body without needing complex holographic recorders or worrying about movement. “Slight movements because of breathing or circulation can distort the images transmitted through a multimode fiber,” said professor Demetri Psaltis, a team leader. Telecommunication signals often have to travel through many kilometers of fiber and can suffer distortions. Deep neural networks are a promising solution for correcting noise caused by these distortions.; Psaltis and his team plan to test the technique with biological samples. The team hopes to conduct a series of studies using different categories of images to explore the possibilities and the limits of its DNN-based technique. The research was published in Optica, a publication of OSA, The Optical Society (doi:10.1364/OPTICA.5.000960).