Novel, Non-Invasive Skin Cancer Detection Method Shows Promise

A new dual-modal, non-invasive detection technique could make it easier—and faster—to differentiate between melanoma skin cancer and benign lesions.

The combination of optical coherence tomography (OCT) and Raman spectroscopy achieved 96.9% accuracy in differentiating melanoma and benign lesions when paired with a machine learning algorithm in a new study.¹ The method also had an area under the receiver operating characteristic curve of 0.99. The findings were reported in the Journal of Biophotonics.

Early diagnosis of melanoma can have profound impacts on patient survival, the authors noted. Patients whose cases are caught at an early stage have a 99% 5-year survival rate, but if the cancer metastasizes to distant organs, the survival rate drops to just 35%.²

Future validation studies with a larger sample size are warranted to ensure the technique will be useful in a clinical setting. | Image credit: Artur Wnorowski - stock.adobe.com

Melanoma can easily go undetected because its diagnosis tends to start with a visual inspection by a clinician or through the use of dermoscopy, followed by confirmatory histopathological evaluation.

“However, the accuracy of these visual diagnoses is highly dependent on the experience and expertise of the diagnosing physician,” the authors wrote, adding that the insufficiency of the current system reveals a clear need for credible, accurate, and noninvasive diagnostic techniques that can help clinicians act more quickly and confidently.

In the new study, the authors turned to 2 techniques to try and solve the problem. OCT is capable of generating high-resolution images that rival histological sections, they explained. Previous research has suggested that pairing this technique with machine learning could achieve a 73% accuracy rate in detecting actinic keratosis and an 81% accuracy rate in detecting basal cell carcinomas.³ Those accuracy rates offer formidable evidence of the potential of OCT and machine learning, the authors said, but there is still room for improvement.

Raman spectroscopy, meanwhile, has also been studied as a possible diagnostic tool in skin cancer. The investigators used OCT parameters, including attenuation coefficients, coefficient of determination (R²), and root mean square error (RMSE), to distinguish between possible melanoma and healthy lesions. They also integrated Raman spectroscopy into the OCT device to improve optical analysis and obtain the biochemical characteristics of different lesions, which the authors hoped would enhance the overall strength of the framework. By combining those tools with machine learning, their goal was to see if the combination would work as a noninvasive and more objective in vivo diagnostic tool.

The authors took 2 skin lesions from the thighs of a 66-year-old male volunteer. By choosing to take 2 samples from the same volunteer, the investigators were able to assume that any differences between the 2 lesions would be due to different biochemical properties and not to variations related to individual skin characteristics.

In addition to analyzing the samples using the novel noninvasive approach, the authors surgically excised both lesions for histopathological examination and definitive diagnosis. One was confirmed to be a benign nevus, while the other was malignant in situ melanoma. Using additional parameters and the pairing of OCT with Raman spectroscopy enhanced the accuracy of their findings.

“Raman spectroscopy revealed differences in carotenoid, amide-I, and CH2–CH3 structures between melanoma and nevi, supporting the OCT findings,” the authors found. “Autofluorescence background intensity variations further distinguished lesion types and enhanced lesion assessment.”

The investigators noted that future validation studies with a larger sample size are warranted to ensure the technique will be useful in a clinical setting. They said they are also exploring additional modalities, including photoacoustic tomography and high-frequency ultrasound.

“Ultimately, the goal is to achieve noninvasive diagnostics of melanoma and other types of skin cancer, minimizing the need for painful biopsies and improving patient care,” they wrote.

References

1. Wu D, Fedorov Kukk A, Panzer R, Emmert S, Roth B. In vivo differentiation of cutaneous melanoma from benign nevi with dual-modal system of optical coherence tomography and Raman spectroscopy. J Biophotonics. Published online April 21, 2025. doi:10.1002/jbio.70040
2. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74(1):12-49. doi:10.3322/caac.21820
3. Jørgensen TM, Tycho A, Mogensen M, Bjerring P, Jemec GB. Machine-learning classification of non-melanoma skin cancers from image features obtained by optical coherence tomography. Skin Res Technol. 2008;14(3):364-369. doi:10.1111/j.1600-0846.2008.00304.x