Handheld optical device, and method of operation thereof, for capturing non-mydriatic retinal images from a patient using a smartphone having a camera, an inertial sensor and an electronic data processor, said device comprising a housing for coupling to the smartphone and an optical system for capturing retina images through said camera, said method comprising, for aligning the camera with the retina visual axis, the electronic data processor carrying out the steps of: capturing an image of the retina； using an image classification method configured by pre-training on previously captured images to detect whether the camera is lower or higher than the retina visual axis； if the camera was detected as being lower or higher, then inputting the value of the Y-axis component of the inertial sensor； depending on these, instructing the user to translationally move up or down the device and/or instructing the user to tilt up or down the device.
