IFCB Flow Metric Microservice

A stateless microservice that computes flow metric anomaly scores for IFCB (Imaging FlowCytobot) bins using a pre-trained Isolation Forest model.

Overview

This service extracts 26 statistical features from IFCB point cloud data (ADC files) and computes anomaly scores using an Isolation Forest classifier. Higher scores indicate more unusual flow distributions that may indicate instrument issues or data quality problems.

API

GET /flow_metric/{bin_id}

Computes the flow metric anomaly score for the specified IFCB bin.

Parameters:

• bin_id (path): IFCB bin identifier (e.g., D20120101_T120000)

Response:

• Content-Type: text/plain
• Body: Single float value representing the anomaly score

Example:

curl http://localhost:8001/flow_metric/D20120101_T120000

Response:

-0.0523

Setup

Prerequisites

1. IFCB data directory with bin files (must contain ADC files)
2. Pre-trained classifier model file (classifier.pkl)

Environment Variables

Required environment variables (set these on your host):

• IFCB_DATA_DIR: Path to IFCB data directory on host (e.g., /path/to/ifcb/data)
• MODEL_DIR: Path to directory containing classifier.pkl on host (e.g., /path/to/models)

You can copy .env.template to .env and modify the paths:

cp .env.template .env
# Edit .env with your actual paths

Running with Docker Compose

docker compose up --build

The service will be available at http://localhost:8001

# Replace with an actual bin ID from your data
curl http://localhost:8001/flow_metric/D20120101_T120000