Paris-Saclay University views structure and composition of skin cancers

05 Aug 2025

Dual-mode optical imaging offers new approach to clinical assessment and diagnosis.

Traditional methods for early detection of skin cancers still rely on visual inspection and biopsy of excised tissue samples, a time consuming and potentially imprecise process.

A French team at Saint-Étienne University Hospital and Paris-Saclay University has now developed an alternative technique based on dual-mode optical imaging, intended to provide more detailed information without cutting into the skin.

As described in Journal of Biomedical Optics, the technique combines line-field confocal optical coherence tomography (LC-OCT) and confocal Raman microspectroscopy (CRM).

LC-OCT, based on OCT and reflectance confocal microscopy, can generate horizontal and vertical section images in real time with micron-scale cellular resolution, but on its own can only provide morphological information about the tissues being studied.

Combining the technique with CRM to exploit the Raman technique's ability to provide molecular fingerprints of chemical species would boost the clinical value of the approach, but attempts to do so have been limited to ex vivo imaging and required strict calibration control.

The Paris project has now combined LC-OCT and CRM in a platform designed to be compatible with in vivo and ex vivo imaging, and carried out the first large-scale study of Raman spectroscopy guided by cellular-level 3D imaging on ex vivo specimens, under conditions compatible with eventual in vivo use.

Personalized diagnosis and treatment

"Our system integrates LC-OCT for cellular-level imaging, and confocal Raman microspectroscopy to analyze the chemical composition of specific targets identified within the morphological images," said the team in its paper.

In trials over the course of one year, the system was tested in a clinical setting on 332 skin cancer samples at Saint-Etienne University Hospital, specifically nonmelanoma types like basal cell carcinoma (BCC) and squamous cell carcinoma (SCC).

The researchers used LC-OCT to locate suspicious structures, and then applied Raman microspectroscopy to gather over 1,300 chemical spectra from those areas. Interpreting this large dataset of Raman spectra required the development of a bespoke AI model, trained to recognize patterns associated with cancerous tissues.

Results showed that "the AI model performed well, achieving a classification accuracy of 95 percent for basal cell carcinoma and 92 percent when both types of cancer were included," commented the project.

This suggests that the system can reliably distinguish cancerous structures based on their chemical signatures, and further analysis of the data revealed distinct chemical differences between various cancer types, offering new insights into how these cancers develop and behave.

"LC-OCT alone has been shown to perform very well for accurate diagnosis and subtyping of BCC," said the team. "Raman characterization would provide an additional layer of chemical subtyping, which could be relevant for personalized treatment options to be explored in further studies."