Skin Ontology Tool

Overview

The Skin Ontology Tool is a utility tool designed to assist in translating dermatology datasets to ICD-10 and ICD-11 codes. The tool processes input data, matches labels to their corresponding ICD codes, and allows users to specify different levels of granularity for the output. It is particularly useful for standardizing dermatology datasets, ensuring compatibility with ICD-based systems and combining existing datasets together.

Features

• ICD Code Translation:
  • Map dataset labels to ICD-10 and ICD-11 codes.
  • Support for hierarchical ICD-10 ontology structure.
• Granularity Levels:
  • Level 1: Broad category range (e.g., L00-L08).
  • Level 2: Specific category (e.g., L01).
  • Level 3: Detailed subcategory (e.g., L01.1).
• Unmatched Label Tracking:
  • Identify and log unmatched dataset labels.
• Customizable Processing:
  • Option to process only unmatched rows with missing or placeholder (tbd) ICD values.

Installation

1. Clone the repository:

   git clone https://github.com/your-repo/skin_ontology_tool.git
   cd skin_ontology_tool

2. Install required dependencies:

   pip install -r requirements.txt

Usage

The tool is executed via the command line and supports several options:

Command-Line Arguments

• --input (required): Path to the input CSV file containing the dataset.
• --data (required): Path to the folder containing ICD mapping datasets (JSON or CSV).
• --output (required): Path to the folder where the translated dataset will be saved.
• --granularity: Level of granularity for ICD codes (1: range, 2: category, 3: subcategory).
• --only-unmatched: Process only rows with missing or placeholder (tbd) ICD values.

Example Command

python skin_ontology_tool.py --input datasets/dermatology.csv \
                             --data icd_mappings/ \
                             --output results/ \
                             --granularity 2 \
                             --only-unmatched

Input File Format

The input CSV file must contain a column named label_to_icd with labels to be matched to ICD codes. Example:

label_to_icd	icd10	icd11
Impetigo	tbd	tbd
Cutaneous abscess	L02	NA

Output Files

1. Translated CSV: The input dataset with added ICD-10 and ICD-11 columns populated.
2. Unmatched Labels: A text file listing labels that could not be matched.

Example Output

Translated CSV:

label_to_icd	icd10	icd11
Impetigo	L01	L01.0
Cutaneous abscess	L02	NA

Unmatched Labels:

Unmatched labels saved to results/unmatched_labels.txt

Developer Guide

Hierarchical Parsing

The tool supports ICD-10 ontology in JSON format. Example structure:

{
  "code": "L00-L08",
  "description": "Infections of the skin and subcutaneous tissue",
  "subcategory": [
    {
      "code": "L01",
      "description": "Impetigo",
      "subcategory": [
        {
          "code": "L01.1",
          "description": "Impetiginization of other dermatoses"
        }
      ]
    }
  ]
}

A json containing dermatology related icd code is provided in the data folder.

Adding New Data

Place new ICD mapping files in the specified --data folder. Supported formats:

• JSON: For hierarchical ontologies.
• CSV: For flat mappings (must include label, icd10, and icd11 columns).

Future Features

• ICD-11 Hierarchical Support: Extend hierarchical parsing to ICD-11 ontology.
• Python package

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository.
2. Create a new branch for your feature/fix.
3. Submit a pull request.

License

This project is licensed under the MIT License. See the LICENSE file for details.