Search & dAIgnose" - Toward intelligent diagnostics with adaptive spectroscopic imaging

Artificial intelligence is gradually establashing a foundational role in health care, with applications in diagnostics, intraoperative assitance and personalized medicine. It is only a matter of time before instruments are largely operated by artificially intelligent controllers.

The current project aims to bridge that gap by introducing a novel type of instrument that can both non-invasively, and without direct contact, extract detailed spectroscopic information from tissue.

By using structured illumination through adaptive optics with a push-broom hyperspectral camera, these components will operate autonomously via the control of an intelligent algorithm.

The algorithm will only recieve an instruction to search for a particular spectral fingerprint and will thereafter iteratively control the illumination and detection patterns and seek to minimize the discrepancy between the target and what is measured.

The Hyperspectral Intelliscope essentially enables the physical manifestation of an optimization algorithm that can quickly search a tissue volume of several cubic centimeters for a particular spectral feature.

The novelty in the proposed project lies in the technical design and the type of tissue interrogation it enables, as well as the implementation of the optimization algorithm.

Being able to search a volume of tissue for a deviating molecular feature has a clear impact on diagnostics, but its general application to intelligent instrumentation is also of important value.