Application of multiphoton imaging and machine learning to lymphedema tissue analysis

Area of Science:

• Biomedical Engineering
• Dermatology
• Medical Imaging

Background:

• Lymphedema is a chronic condition often leading to significant tissue fibrosis.
• Accurate characterization of fibrotic changes in lymphedema is crucial for diagnosis and treatment.
• Current diagnostic methods may not fully capture the microstructural tissue alterations.

Purpose of the Study:

• To investigate in-vivo microstructural changes in lymphedema-affected skin using two-photon imaging.
• To evaluate the potential of machine learning algorithms for diagnosing lymphedema based on imaging data.
• To quantify fibrosis severity by analyzing collagen and elastin in skin tissue.

Main Methods:

• In-vivo two-photon microscopy was employed to image skin tissue from 36 lymphedema patients and 42 healthy controls.
• Image analysis included edge detection and histogram of oriented gradients to assess collagen network organization.
• A machine learning model, specifically ensemble learning, was developed for classification and diagnosis.

Main Results:

• Lymphedema tissue exhibited significant collagen network disorganization compared to healthy controls.
• An increased collagen/elastin ratio was observed in lymphedema tissue, indicating advanced fibrosis.
• The ensemble learning model achieved 96% accuracy in classifying lymphedema from healthy tissue in the testing set.

Conclusions:

• Two-photon imaging effectively visualizes fibrotic changes in lymphedema skin.
• Quantitative image analysis combined with machine learning shows high diagnostic accuracy for lymphedema.
• This approach offers a promising tool for objective lymphedema diagnosis and severity assessment.