First off, if we cut through the usual dismal quality of scientific reporting, what they made is a photodetector, not an image sensor. It detects single events rather than capture an image. The sensitivity of the detector is not the same as quantum efficiency. The sensitivity they mention here includes a "photogain" by virtue of the detector operating more or less as a light-controlled amplifier. It takes electrical input energy and simply amplifies it based on incident light. That can create a flow of many more electrons than incident photons. The same thing can possibly be also done by introducing a gain in the conventional image sensor electronics too, but having this photogain right inside the sensor should theoretically lead to better noise performance. So we would expect the paper to quantify the noise characteristics, but it is woefully sparse on the noise details - which leads me to suspect this is yet another "invention" that is never going to see the light of day.