Design and Quality for Biomedical Technologies XVIII

The introduction of the Medical Device Regulations within the EU in 2017 has made the certification of medical devices more challenging for research institutes especially providing clinical benefit. Information and guidance tools have been developed to support researchers. It is essential to prepare documentation during the development process related to quality control and risk assessments. To allow time for feasibility testing and improvements the device could be labelling as a ‘research only device’. After finalization of the design, a full clinical investigation, according to MDR, can be conducted. Recently, the EU authorities have acknowledged the negative impact of the MDR and measures are taken to relieve research institutes and promote health innovations again.

Fluorescence molecular imaging (FMI) and endoscopy (FME) are technologies with great potential to guide interventions and to provide earlier, faster, and personalized diagnosis in oncology. However, FMI and FME still present challenges that can confound real-time decision making for disease management and/or treatment. Importantly, the markedly different systems hurdle the repeatability of measurements, the unbiased readout interpretation, and the wide clinical acceptability of FMI and FME. Herein we present our work for the development of multi-parametric standards to perform quality control and performance assessment of FMI systems. Moreover, we discuss examples illustrating how data analysis and the design of fluorescence standards influence performance assessment outcomes, potentially affecting comparisons between systems or studies. We, also, show the first standard tailored to the requirements of FME and demonstrate its use for quality control of a fiberscope-based FME system. The discussed performance assessment and quality control framework can accelerate the clinical translation of fluorescence molecular imaging and endoscopy and steer further developments in the field.

This study characterizes a 1.5 mm wide-angle, 400x400 pixel RGB CMOS microsensor for label-free widefield autofluorescence imaging (WFAF) to detect oral epithelial dysplasia (OED) and squamous cell carcinoma (OSCC). Utilizing 405 nm excitation, the system demonstrated effective detection of red and green autofluorescence signals in both tissue standards and resected samples. The compact design and high-resolution capabilities suggest potential for integration into multimodal imaging systems for noninvasive detection of early cancerous changes.

This study investigates the effect of several factors (e.g., light sources, warm-up time, camera focus, working distance, light/target angles) on the spectral performance of two hyperspectral imaging (HSI) cameras. HSI cameras were connected to a rigid endoscope through a zoom lens coupler to create two hyperspectral endoscopes (HySEs). The effects of the aforementioned factors on spectral performance of two HySEs were evaluated and best practice identified to ensure the spectral accuracy and reliability of HySEs. These findings provide a foundation for standardized evaluation practices, promoting the adoption of high quality HySEs in medical diagnostics for applications including early cancer detection.

Miniaturization of optical endoscopes is a crucial step to facilitate less invasive surgeries and enable image guided interventions in previously unreachable locations. Sub-wavelength diffractive meta-optics offer a promising alternative to bulky refractive lenses to realize such miniaturization and enable new modalities. In this talk, I will cover meta-optics fundamentals and describe their potential for ultra-compact optical probes, focusing on three innovative forward-viewing endoscopes, utilizing a coherent fiber bundle, scanning fiber endoscope, and single fiber based endoscope. Overall, this talk will highlight how meta-optics can replace refractive lenses, reduce optical track length, and improve imaging capabilities.

Breast cancer poses a significant public health challenge in low- and middle-income countries (LMICs) due to lack of rapid, low-cost, and non-destructive point-of-care assessments. To address this, a direct-to-digital imaging system was developed in collaboration with Peace and Love Hospital in Ghana. Our proposed system integrates tissue collection, staining, and imaging into a single process, using FIBI (Fluorescence Imitating Brightfield Imaging) and low-cost 3D printed consumables. This solution provides immediate, diagnostic-quality images within minutes, eliminating traditional slide preparation. Implementing this end-to-end rapid diagnostic system in areas without Pathology infrastructure, will increase access to early diagnosis and reduce treatment delays.

The photoplethysmography (PPG) sensors commonly used in wearable devices rely on changes in the diffuse reflection of light on the skin to detect heart rate. These changes are primarily due to the varying volume of blood in the tissue throughout the cardiac cycle. In this study, we consider three different epidermal thicknesses and study their effects on PPG signal. This is accomplished by using Monte Carlo eXtreme (MCX) to evaluate the photon path length in the tissue. We will study the phenomenon using both continuous wave (steady state) PPG which is a standard format in common wearables, and time-of-flight (time-resolved) PPG. Experimental validation on hemoglobin phantoms will be performed at common wavelengths (between 500 and 900 nm) for steady state cases whereas TOF-PPG will be studied using a system consisting of a pulse laser with wavelength at 780 nm, single-photon avalanche diodes, and a multi-channel time-correlated single photon counter.

Prior research indicates a positive bias in pulse oximeters for darker skin tone patients, often attributed to higher melanin concentration. We present a novel hybrid Monte-Carlo and analytical simulation framework that captures the effect of skin properties and system factors on reflective geometry. Our model integrates light-tissue interaction, time dynamics of blood flow and skin depth encoding of pulsatile strength. Our findings reveal that melanin variation minimally impacts SpO2 readings. Instead, variations in the depth of the pulsatility function (PF) significantly influence SpO2 bias. Additionally, wavelength shifts in red LEDs affect SpO2 accuracy more than shifts in infrared LEDs.

Tissue phantoms that mimic the optical and radiologic properties of human or animal tissue play an important role in development, characterization and evaluation of imaging systems. Phantoms which are easily produced and stable for longitudinal studies are highly desirable. In this talk, we will review the fundamental and specific features of tissue phantoms required for optical spectroscopy, multimodality imaging, and surgical guidance. We will present several new types of long-lasting phantoms made from commercially available materials. The ease of fabrication, stability, and control over optical properties will be assessed, and the contrast properties in magnetic resonance imaging (MRI) and x-ray computed tomography (CT) will also be evaluated.

One of the major challenges of biomedical engineering is to bridge the so-called valley of death, which arises from high costs, ethical barriers and technical issues. The development of reliable and low-cost platforms for general purpose phantoms may make a substantial impact. We propose a concept of hierarchical manufacturing as an approach to encode multiple mechanisms of physical contrast based on water-in-elastomer micro-emulsions made of a continuous phase of hydrophobic polydimethylsiloxane and micro-droplets of hydrophilic solutions. We show the possibility to replicate the peculiar porosity of lung tissue, by repurposing polydimethylsiloxane sponges for use as phantom materials mimicking the lungs.

Raman spectroscopy (RS) is a non-invasive and label-free technique capable of quantifying the biochemical composition of cells and tissues. With the increasing use of high wavenumber (HW) RS for translational applications, there is a need to standardize and ensure the repeatability of acquired and processed Raman data. In this study, we developed a method for fabricating tissue-mimicking HW Raman standards with well-defined biochemical composition and spectral profiles. This method and the resultant biochemical standards were verified to provide quantifiable, repeatable, and accurate spectral profiles as measured by a Raman microscope and a clinical fiber-optic RS system. Ultimately, this method can ensure accuracy and consistency of Raman spectral data from different RS systems and preprocessing methods, representing another important step in the clinical translation of Raman spectroscopy.

Clinical studies have indicated that epidermal pigmentation can significantly impact the accuracy of pulse oximetry. Towards a phantom-based test method for pulse oximetry, we present a modified and streamlined method for fabricating 0.1-mm-thick epidermal layer phantoms based on readily available chemical components. We were able to achieve optical property values within 10% of target levels across the 660 nm to 940 nm range for low, medium and high pigmentation cases. This technique provides reproducible and stable epidermal phantoms suitable for evaluating the impact of pigmentation on oximeters and other biophotonic devices.

This study introduces an in-vitro model using Ecoflex, a silicone-based material, to create skin-like replicas for investigating catheter-associated infections. This model mimics human skin texture in an excellent agreement compared to standard test samples and is validated by studying bacterial adhesion, surface roughness, and wettability on the silicone-based material. Our findings indicate the replicas accurately simulate human skin mechanical, thermal, and physical properties. The bioinspired model provides a useful realistic platform for studies on enhancing infection control strategies to improve patient outcomes and developing and testing new catheter designs and wearable sensing devices.

Explainable artificial intelligence algorithms for point-of-care cancer diagnosis are designed to make the decision-making processes of AI models transparent and interpretable, that address a critical need in clinical practice. These algorithms integrate machine learning models with interpretable features, allowing clinicians to understand how decisions are made by AI algorithms. By providing clear explanations, they bridge the gap between AI-driven predictions and clinical insights, enhancing trust and enabling informed patient management. This approach not only improves diagnostic accuracy and efficiency but also empowers healthcare professionals to make confident, evidence-based decisions, thereby advancing personalized medicine in oncology.

There is urgent need for recommendations on how to accurately characterize skin pigment for the purpose of ensuring diversity in medical device research, development and regulatory validation cohorts. We compared performance of multiple subjective and objective skin pigment assessment methods in 789 participants. We found that common methods for skin pigment characterization, such as race or subjective skin color scales, have important limitations. When applied to the same cohort, different methods yield different proportions of light, medium and dark subjects and may overestimate diversity of skin pigment. Previously published ITA thresholds for defining ‘dark’ skin may be too light.

Biophotonics continues to make major contributions to medical care. However, recent studies in pulse oximetry and other transdermal technologies have highlighted the need to re-assess the potential for skin pigmentation to impact device performance. We present an extensive review of literature to elucidate the effect of epidermal melanin on optical measurements across a range of biophotonic technologies. Fundamental optical properties and light-tissue interaction mechanisms are discussed, as well as approaches used to mitigate the impact of melanin. This work will provide a foundation for future efforts to ensure high performance of biophotonic technologies for all patients, regardless of skin color.

Cardiovascular disease is a leading cause of death in the U.S., affecting over 40% of the population alongside prevalent obesity. Health wearables, increasingly common, struggle with accuracy across diverse populations due to variations in skin tone and obesity. To address this, we developed a dynamic phantom that simulates wrist tissues' optical and mechanical properties for photoplethysmography (PPG). It contains properties across Fitzpatrick types range I-VI, low and high BMI, and connected to a pulsatile pump for pressure fluctuations. Constructed using stereolithography, fused deposition modeling, and silicone mold casting, it is tested with a NellCor Pulse Oximeter to validate PPG readings.

Photoplethysmography (PPG) is a non-invasive optical technique, but cofounders (including pigmentation) limit performance. There is a need for versatile testing platforms to understand the impact of confounders on optical sensors. This work aimed to fabricate a model with distensible vasculature channels suitable for wearable PPG devices that could incorporate a pigmented epidermal layer. A peristaltic pump was used to create pulses that were fed into an anatomical finger modelled on Solidworks and printed with a photopolymer resin using stereolithography. Various dimensions were attempted to represent vasculature, finding reproducible success in vessel diameters >2.8mm. Stereolithography 3D printing is an advantageous method to create pulsatile phantoms. Overall, with the many degrees of freedom and interchangeable components, this system can ameliorate the burden of testing on patients and animal models. This work is timely given the active investigation by regulatory bodies on pulse oximeter testing platforms and their necessity for equitable health care.

Domestic Violence (DV) is a public health crisis. There is growing evidence that the more subtle visual appearance of bruises in DV survivors with darkly pigmented skin (DPS) creates barriers for life-saving help-seeking behavior. Bruises consist of extravasated blood and water within the dermis below the melanin-containing epidermal basal layer, and the overlapping absorption spectra of blood and melanin in the visible range results in bruises having a more subtle visual appearance in DPS. In contrast, because the short-wave infrared range is sensitive to water and insensitive to melanin, we hypothesized that SWIR-imaging could enable equitable bruise assessment that overcomes the limitations of visible photography. To this end, induced and imaged bruises in a swine breed with mosaic pigmentation and found that SWIR image was able to visualize bruises with high contrast in DPS, thereby overcoming longstanding limitations of visible photography with respect to bruise documentation.

In this work, we developed phantoms that can properly emulate either benign or malignant tumors for testing the performance of continuous wave (CW) NIRS devices for Optical Mammopgraphy. The main advantage of these phantoms is that oxy- and deoxyhemoglobin are mimicked by artificial commercially available absorbents, namely a dye (ADS830WS) and an ink (Epson673), thus requiring neither special preparation protocols nor careful storage, which makes them very stable in time. Tumor-like inclusions were made of epoxy resin containing these two absorbents to emulate the absorption of the hemoglobins in a fibradenoma and an adenocarcinoma. To evaluate the approach we performed CW transmittance imaging experiments, which were validated by Monte Carlo (MC) simulations. Results indicate that the constructed tumor phantoms effectively replicate the desired targets. Moreover, the reconstruction algorithm was able to discriminate between the two types of lesions mimicked in this study.