Lunch Lecture PDEng

Intelligent Microscope

Artificial Intelligence (AI) technologies evolve rapidly and successfully solve tasks such as object detection and natural language processing. AI gives software users the ability of a novel, natural interaction with applications. For example, users can talk to a software system to explain their needs and get direct response, rather than use a mouse or a keyboard to ex-tract and analyze information. Companies that develop modern software systems conduct research on how AI can be used and integrated into their software stack. This work describes how a prototype of an intelligent microscope (IM) was created. The IM is a software artifact. It is an AI-web system that can un-derstand predefined user voice-commands, control an electron microscope, detect objects on microscope images, and high-light specific objects requested by a user. The main motivation for the project was to explore and show possible ways of User-IM interaction. The IM use cases and software design are described in this report. The IM is capable of detecting and highlighting specific types of objects, such as cells and mitochondria. Also, it is able to interpret information about detected objects and show which of the objects are big, small, or relatively close to each other. These use cases provide examples of how the IM can be used. Even though these use cases are unique they do not limit the IM to mitochondrion and cell detection only, because one of the main implemented IM requirements is extendibility with other use cases in the future. The results of the project show how microscope control can be simplified using conversational interface, how image analysis can be improved using object segmentation, and how information about microscope image context can be interpreted using intelligent microscope knowledge database.

Note: After subscription showing up is mandatory. You can unsubscribe untill the 8th of may 23:59.