Quantum-Optimized Reconstruction of Ketone Dynamics from Fabry-Pérot Optical Breath Measurements
By Aarjav, Mohan
Introduction: Diabetes management depends on measurements that are often invasive, intermittent, or expensive. Acetone in exhaled breath is linked to ketone metabolism and can be measured optically, but a breath spectrum is not itself a metabolic state: the clinically meaningful quantities—blood β-hydroxybutyrate (BHB), acetoacetate, acetone, and their rates of change—must be inferred from noisy, indirect observations. Prior cavity-enhanced systems have detected breath acetone near 1670~nm, establishing a realistic optical starting point. Earlier QLab students also demonstrated the feasibility and practical limitations of student-built breath-acetone spectroscopy.
We propose a coupled computational and experimental project. Mohan will extend a verified slow-fast ketone model and classical physics-informed neural network (PINN) with a small variational quantum layer, then use quantum Fisher information (QFI) to quantify which model parameters and optical features are most identifiable. Aarjav will design and fabricate a Fabry-Pérot interferometer, select light-emitting diodes and a near-infrared spectrometer, and characterize its resolution, stability, and acetone sensitivity. QFI-guided priorities will inform wavelength and component choices; measured instrument noise and spectral response will then be returned to the inverse model. The project will test whether this feedback loop improves reconstruction accuracy and uncertainty relative to matched classical baselines, without presuming a quantum advantage or requiring a large number of qubits.
Intellectual Merit: The project connects mechanistic multiscale modeling, hybrid quantum-classical learning, quantum information geometry, and a real optical instrument in one identifiable inverse problem. Beyond investigating quantum modeling applications and optical sensing, this project could result in a practically useful tool for monitoring a disease impacting hundreds of millions of people worldwide.
Broader Impact: A low-cost, noninvasive method for tracking ketone dynamics could ultimately support earlier warning of dangerous metabolic changes. In particular, a byproduct of our research could involve a tool for predicting diabetic ketoacidosis (a dangerous complication of diabetes caused by insufficient insulin and rising ketones) without requiring skin insertion. With that said, this project serves as a feasibility study, not a diagnostic device, and any faculty breath study will proceed only with required approval and informed consent.
See our current project proposal in the second downloadable file.