Real-time multi-exposure laser speckle contrast imaging of skin microcirculatory perfusion / Martin Hultman.

• The microcirculation, the blood flow in the smallest blood vessels in the body, has a vital function as this is where oxygen and nutrients diffuses from the blood to to the surrounding cells. An important quantity is the tissue perfusion, a measure of the microcirculation’s capacity to provide oxygen and nutrients to the cells. Laser speckle contrast imaging (LSCI) is a non-invasive optical technique that captures images of the microcirculatory perfusion by analysing the local contrast in the laser speckle pattern that forms when tissue is illuminated by a laser. LSCI has seen extensive use in clinical research due to the easy and cheap measurement setup, and high spatial and temporal resolution. Despite this, clinical acceptance and routine use remains low. Some of the drawbacks of the technique is a limitation to relative measurements in arbitrary units, as well as high susceptibility to measurement noise and confounding properties of the tissue. This makes comparisons difficult, especially between patients. An extension of LSCI called multi-exposure laser speckle contrast imaging (MELSCI) was proposed to deal with some of these issues, although the more complicated data acquisition and models prevented real-time use. MELSCI has in-stead been used exclusively as an offline technique where data is post-processed, and the clinical use has been non-existent. Furthermore, existing models for LSCI and MELSCI are designed for tissues where individual vessels are visible, such as the surface of the brain or on the retina. For measurements in the diffuse regime, such as on skin tissue, these models are no longer physiologically accurate, resulting in incorrect perfusion estimates. This thesis presents a MELSCI-based perfusion imaging instrument that is simultaneously fast and physiologically accurate for measurements of skin. There are three main parts to this work; development of a real-time MELSCI system, development of perfusion models for skin, and demonstration of the system in a clinical feasibility study. A real-time MELSCI instrument was developed based on a high-speed CMOS camera tightly integrated with algorithms in a field programmable gate array (FPGA). The algorithm was based on synthetic multi-exposure, where a set of 64 individual 1-ms images were digitally added to create multi-exposure images at 1, 2, 4, 8, 16, 32, and 64 ms. The resulting multi-exposure data was demonstrated to have high quality and less susceptibility to measurement noise than previous models. The instrument enabled continuous acquisition and analysis of MELSCI data in real-time at 15.6 frames per second, sufficiently fast to capture the temporal dynamics of the skin perfusion. To enable real-time estimation of accurate and physiologically relevant perfusion from the MELSCI data, two artificial neural networks were trained on synthetic data from a mathematical model of skin. The first estimated perfusion as computed by conventional laser Doppler flowmetry (LDF), demonstrating a high correlation between the two methods. The second estimated true perfusion in absolute units %RBC × mm/s separated into three distinct speed components, 0-1 mm/s, 1-10 mm/s and >10 mm/s. The ANNs removed the need for iterative optimization algorithms, resulting in more than 1000x speed-up over previous methods, and enabled real-time use in an imaging setting. The instrument was demonstrated in controlled experiments on healthy volunteers, using standardized occlusion-release provocations, and in a clinical feasibility study where the foot perfusion was monitored during endovascular interventions in patients with chronic limb-threatening ischemia. The instrument enabled continuous imaging of perfusion, with sufficiently high framerate to capture the pulsatile dynamics, or lack thereof, at each point in time. The necessity for both high spatial and temporal resolution to properly asses the microcirculation was demonstrated. The advancements to MELSCI proposed in this thesis has the potential to improve the clinical viability of the technique, increase interpretability of the results, and might lead to improved treatments based on a better understanding of the complex processes in the microcirculation. 

Ämnesord

Medicinsk teknik  (sao)

Medicinska instrument och apparater  (sao)

Blodomlopp  (sao)

Datoralgoritmer  (sao)

Engineering and Technology  (hsv)

Medical Engineering  (hsv)

Medical Laboratory and Measurements Technologies  (hsv)

Teknik och teknologier  (hsv)

Medicinteknik  (hsv)

Medicinsk laboratorie- och mätteknik  (hsv)

Medical technology  (LCSH)

Medical instruments and apparatus  (LCSH)

Computer algorithms  (LCSH)

Genre

government publication  (marcgt)

Indexterm och SAB-rubrik

Multi-exposure laser speckle contrast imaging

Microcirculatory perfusion

Machine learning

Klassifikation

610.28 (DDC)

Vpeb (kssb/8 (machine generated))