asoct-mcd 0.2.1

Minuscule Cell Detection in AS-OCT Medical Images

ASOCT-MCD: Minuscule Cell Detection in AS-OCT Images

A Python package for detecting minuscule cells in Anterior Segment Optical Coherence Tomography (AS-OCT) medical images.

Installation

pip install asoct-mcd

Basic Usage

from asoct_mcd.pipeline import MCDPipelineBuilder

# Create pipeline with default settings
pipeline = MCDPipelineBuilder().build()

# Detect cells in image
result = pipeline.detect_cells("image.png")

# Print results
print(f"Detected {result.cell_count} cells")
print(f"Cell locations: {result.cell_locations}")

Custom Configuration

# Using dictionary configuration
config = {
    "threshold": {"lambda_factor": 0.9, "method": "isodata"},
}

pipeline = MCDPipelineBuilder().from_dict(config).build()
result = pipeline.detect_cells("image.png")

# Using YAML configuration
pipeline = MCDPipelineBuilder().from_yaml("your_config.yaml").build()
result = pipeline.detect_cells("image.png")

Model Storage and Management

Default Model Storage Locations The ASOCT-MCD package automatically downloads and manages pre-trained models. Models are cached locally following standard ML library conventions: Default Cache Directories Linux/macOS:

~/.cache/asoct_mcd/models/
├── sam_vit_b_01ec64.pth              # SAM ViT-B segmentation model (~375MB)
└── spatial_attention_network.pth     # Cell classification model (~1MB)

Windows:

C:\Users\<username>\.cache\asoct_mcd\models\
├── sam_vit_b_01ec64.pth              # SAM ViT-B segmentation model (~375MB)
└── spatial_attention_network.pth     # Cell classification model (~1MB)

Requirements

• Python >= 3.9
• See requirements.txt for full list

Citation

arXiv: https://arxiv.org/abs/2503.12249

To cite MCD in publications, please use:

@article{chen2025minuscule,
      title={Minuscule Cell Detection in AS-OCT Images with Progressive Field-of-View Focusing}, 
      author={Boyu Chen, Ameenat L. Solebo, Daqian Shi, Jinge Wu, Paul Taylor},
      year={2025},
      journal={arXiv preprint arXiv:2503.12249}
}

Acknowledgements

Thanks to the support of AWS Doctoral Scholarship in Digital Innovation, awarded through the UCL Centre for Digital Innovation. We thank them for their generous support.