Biopsy-based histology (the study of tissue) has been the foundation for the diagnosis and management of a variety of diseases for more than 150 years. Advances in skin imaging modalities, such as dermoscopy, reflectance confocal microscopy (RCM), and optical coherence tomography (OCT) have enhanced screening and detection capabilities, reducing the need for biopsies for some conditions. However, the diagnosis of skin cancers and some skin disorders largely remain dependent on invasive biopsies.

Although dermatologists are very skilled at evaluating skin lesions, macroscopic features on the surface of the skin do not fully reveal what is happening below the surface. Without sufficient magnification, it is also challenging to understand the changes occurring at the cellular level. By one estimate, less than 50% of biopsies of suspicious skin yield malignant diagnosesⁱ.  With 44 million dermatology visits a year in the U.S, resulting in 15 million biopsiesⁱⁱ, there is a clear need to improve the noninvasive assessment of tissue to better triage which patients truly need a biopsy.

Making the invisible digital

A long-sought goal in dermatology has been the development of an imaging modality with sufficient resolution and compositional detail to noninvasively interrogate skin histology in vivo. Impressive technologies have emerged in this field, such as RCM, multi-photon microscopy (MPM), OCT, and line-field confocal OCT (LC-OCT). These technologies have generated evidence supporting their ability to better identify disease noninvasively; however, each has limitations when used as a stand-alone tool.

Driven by practical clinical needs, we sought to develop an imaging modality that offers noninvasive, real-time cross-sectional skin imaging in a compact, intuitive form. We aimed to incorporate exceptional imaging performance to support a future of diagnostic capabilities across various skin types and conditions. In our publication, we introduce a new technology—cross-modal imaging—that addresses these clinical needs. Our cross-modal system combines multiphoton and pinhole-free reflectance confocal microscopy into one compact system with a handheld wand using a single Class 1M, 780 nm ultrafast pulsed near-infrared (NIR) light source. Cross-modal generates real-time, cross-sectional, skin images in clinically relevant color using four simultaneous optical channels: pinhole-free reflectance confocal microscopy (pf-RCM); second harmonic generation (SHG); two-photon autofluorescence, short (2PS); and two-photon autofluorescence, long (2PL).

Visualizing age effects with molecular precision

Clinically relevant color is displayed in cross-modal images because the color of light emitted and/or reflected in the skin is affected by the molecular composition, primarily reduced nicotinamide adenine dinucleotide phosphate (NADPH), keratin, collagen, elastin, and melanin. One potential benefit of compositional detail is that orientation to the fundamental layers of skin is more intuitive. The dermis, primarily made of woven collagen I molecules, typically glows blue in cross-modal images, while the overlying living epidermis is usually green, yellow, or orange. These color-based distinctions assist in identifying the dermal-epidermal junction, which is frequently altered in disease.

Another advantage of visualizing molecular composition is an opportunity to measure the impact of sun damage due to age. In our publication, the relationship of the 2PS and SHG channel intensities (color) helped identify the presence of elastin in the dermis, which should normally be composed of collagen I. This change in the dermis, called elastosis, is often a consequence of cumulative sun damage and is linked to aging. While elastosis is one example, other molecular changes revealed by color shifts or changes in cellular architecture revealed by color contrast could provide additional clinical insights into skin health.

Practical clinical use: miniature integrated system with cross-scanned multimodal imaging

The cross-modal system’s size, mobility, and image familiarity were designed to support the clinical workflow of a dermatology clinic. The system is compact, about the size of a briefcase (40 x 33 x 30 cm), and weighs under 10 kg (22 lbs). As a result, the system can be carried and used by one person.

We benefitted from the system’s compact size during its development and testing. When working with other technologies from previous projects, we experienced having to carefully roll heavy, cart-based, fragile optical systems when moving locations. Transporting systems like these to another building required expensive delivery services and had substantial risk of damage. In contrast, the cross-modal system was specifically designed to be easy to maneuver. The first time we needed to transport an early prototype to another location, we were able to fit it easily into a passenger seat of a sedan. What a relief! These days we recommend packing the system into a custom, foam-lined box, though.

To make the device compact, we integrated technologies like fiber delivery of laser excitation and MEMS-actuated scanning to enable subcellular resolution imaging without increasing the size of the hub. The handheld wand contains a small ultrafast fiber laser with low power requirements. Freehand imaging is employed for flexibility and ease of use, inspired by handheld ultrasound. The cross-modal system employed an internal depth-focusing mechanism to enable scanning through the layers of skin (in the z-direction) while keeping the probe tip in static contact with the skin, eliminating the need for rigid fixation of the imaging head to patients.

Existing versions of multiphoton microscopy capture pictures of skin en face, or parallel to the surface of the skin; however, traditional histopathology cuts the skin across the layers. Cross-modal imaging presents a cross-sectional orientation to allow visualization and identification of skin layers in a familiar orientation for interpreting dermatologists and pathologists. This improves workflow and eliminates the need for additional cumbersome training to extrapolate what the results would look like in cross-section.

"Enspectra Health’s cross-modal technology is a very exciting innovation that can transform how skin is assessed in dermatology,” said Roberto Novoa, MD, board-certified dermatologist, dermatopathologist, and co-author of the study. “I can envision a future where physicians integrate noninvasive cross-modal imaging into their workflow in order to have instantly available clinical insights enabling them to provide personalized patient care.”

Clinical validation in diverse skin

In vivo performance was evaluated with 122 human participants. With a subset of 49 participants, we analyzed signal composition in the epidermis and dermis. Participants ranged in age from 9–81 years and in skin type diversity from Fitzpatrick I-V. Moreover, 43% of participants identified as Hispanic/Latino, Asian, Black/African American, or as multiracial. We also assessed cross-modal imaging in participants with suspected skin cancer, capturing images of lesions diagnosed as seborrheic keratosis and basal cell carcinoma before surgical biopsy. Dermal elastosis in sun-damaged skin, elevated melanin in pigmented skin, basaloid nests in basal cell carcinoma, and elongated rete ridges in seborrheic keratosis were observed. Results demonstrated that cross-modal images showed features consistent with histopathology, suggesting promise in further evaluation of the technology across a broader cohort of participants.

The cross-section of imaging and AI

The digital nature of cross-modal images presents an opportunity for the application of artificial intelligence (AI) to enhance the interpretation and utility of this imaging modality. The high-resolution, cross-sectional images, which provide detailed information about cellular morphology and molecular composition, could benefit from AI-driven analysis. Analytical outputs need not be restricted to diagnoses; they could include automated image quality improvements to assist human readers, automated digital “stains” to highlight biological features or regions of interest, or curation of related, biopsy-proven cases for human review. This synergy could lead to a future of noninvasive diagnoses, personalized treatment plans, and a deeper understanding of skin pathologies.

Cross-modal technology offers important advantages for dermatology and beyond

Cross-modal technology could offer the opportunity for noninvasive evaluation of skin in clinical dermatology, with the potential to reduce barriers to inspecting inflammatory or benign-appearing lesions, shorten the time to treatment, and enhance point-of-care decision-making. It may also facilitate research by improving participant recruitment and enabling longitudinal investigation of skin lesions which could reveal changes in skin histology over time. Finally, the utility of the cross-modal technology platform could extend beyond dermatology to intraoperative or minimally invasive imaging of other organs.

Biopsy-based histology remains a standard of medical practice for insights into the diagnosis, progression, and treatment of disease. However, the invasive nature of the procedure and the logistics associated with processing samples and interpreting results fundamentally constrain access to the important cellular information revealed by histology. The advent of a new imaging modality designed to deliver noninvasive, real-time cross-sectional imaging provides rich opportunities for clinical and research advances in dermatology.

Kate Montgomery is an employee at Enspectra Health, Inc. Enspectra Health is transforming how skin health is assessed by leveraging revolutionary imaging and machine learning technology. Enspectra Health’s cross-modal system has been cleared for clinical use by the US FDA and its next generation was granted Breakthrough Device Designation for the evaluation of lesions suspicious of basal cell carcinoma and squamous cell carcinoma in select high-risk populations.

 ⁱWeinstein DA, Konda S, Coldiron BM. Use of Skin Biopsies Among Dermatologists. Dermatol Surg. 2017 Nov;43(11):1348-1357. doi: 10.1097/DSS.0000000000001188. PMID: 28562437
ⁱⁱKlipp, Jondavid. The U.S. Anatomic Pathology Market: Forecast Trends 2019-2021.