Segmentation of mouse skin layers in optical coherence tomography image data using deep convolutional neural networks

Area of Science:

• Biomedical Imaging
• Artificial Intelligence in Medicine
• Dermatology Research

Background:

• Optical coherence tomography (OCT) provides high-resolution, non-invasive 3D imaging, widely used in ophthalmology and dermatology.
• OCT is valuable for research using animal models, but manual analysis of OCT images is challenging due to anatomical variations and lack of standards.
• Accurate segmentation of tissue layers in OCT is crucial for quantitative analysis in preclinical studies.

Purpose of the Study:

• To develop and evaluate a deep learning algorithm for automated segmentation of multiple mouse skin layers from OCT data.
• To compare the performance of the proposed algorithm against existing methods, including a baseline U-net and a previous random forest/graph-based approach.
• To assess the impact of densely connected convolutions on segmentation accuracy and outlier reduction.

Main Methods:

• A deep convolutional neural network (CNN) based on the U-net architecture was designed, incorporating densely connected convolutions.
• The proposed CNN was trained and tested on OCT image data of mouse skin.
• Performance was evaluated by comparing the proposed CNN with a baseline U-net and a prior algorithm combining random forest classification and graph-based refinement.

Main Results:

• The proposed densely connected CNN demonstrated superior performance compared to both the baseline U-net and the previous algorithm on average.
• The integration of densely connected convolutions led to a noticeable reduction in segmentation outliers.
• The algorithm achieved accurate and robust segmentation of multiple skin layers in mouse OCT images.

Conclusions:

• The developed deep learning algorithm offers an effective and automated solution for segmenting mouse skin layers in OCT images.
• The modified U-net architecture with densely connected convolutions enhances segmentation accuracy and reliability in preclinical OCT analysis.
• This automated approach can facilitate more efficient and precise research using OCT in animal models.