Attune

App Description

A mid-sized biomedical research lab focused on immuno-oncology needed to improve the speed and consistency of its immune cell profiling workflows. The lab routinely processed large sample volumes (e.g., PBMCs from clinical cohorts) and faced bottlenecks with traditional flow cytometers that relied on hydrodynamic focusing.

  • Overview
  • Project:Attune NxT Flow Cytometer
  • Platform:Desktop
  • Role:UX Designer (New Feature Update)
  • Audience:Lab technicians, academic researchers, scientists
  • Safe :This project is presented in a portfolio-safe, external-facing format.

Case Study - New Feature

A research lab using the Attune™ NxT Cytometric Software (v6.1) aimed to automate cell detection and counting from brightfield images. Manual annotation was time-consuming and inconsistent, especially when dealing with large datasets (~500 images per sample).

The Challenge
  • Manual cell identification introduced user bias
  • Difficulty in consistently marking cell centers vs boundaries
  • Large datasets slowed down annotation workflows
  • Need for repeatable AI model training
  • Implemented Attune™ NxT image annotation workflow with centroid-based (“Hooke A”) approach
  • Enabled rapid model setup with intuitive model configuration and tagging
  • Selected optimized training datasets with recommended image subsets
  • Used manual and automatic centroid annotation tools (Insert/Suggest Center)
  • Applied preprocessing controls (brightness, contrast, size filtering) for consistency
  • Trained and evaluated models using built-in accuracy metrics and learning curves

Research & Insights

The Attune™ NxT workflow demonstrates a shift toward centroid-based annotation (“Hooke A”), enabling faster and simpler cell detection by focusing on cell centers rather than full segmentation. This reduces annotation time while maintaining reliable performance. The combination of user-guided and automated tools improves accuracy and consistency, while preprocessing controls ensure robustness across variable image conditions. By using optimized subsets of data and built-in performance metrics, the system supports rapid, scalable model development and makes AI-driven cytometric analysis more accessible.

What We Did
  • User interviews with lab technicians and academic researchers
  • Workflow walkthroughs of common ProQuantum tasks
  • Heuristic evaluation against Komodo UX principles
  • Collaboration with product managers and application scientists
  • Collaboration with product managers and application scientists
  • Users rely on step-by-step mental validation during assays
  • Clear system feedback builds more trust than advanced visuals
  • Progressive disclosure works better than information-dense dashboards
What We Did
  • User interviews with lab technicians and academic researchers
  • Workflow walkthroughs of common ProQuantum tasks
  • Heuristic evaluation against Komodo UX principles
  • Collaboration with product managers and application scientists
  • Collaboration with product managers and application scientists
  • Users rely on step-by-step mental validation during assays
  • Clear system feedback builds more trust than advanced visuals
  • Progressive disclosure works better than information-dense dashboards
  • Consistency across steps reduces anxiety during high-stakes runs
What We Did
  • User interviews with lab technicians and academic researchers
  • Workflow walkthroughs of common ProQuantum tasks
  • Heuristic evaluation against Komodo UX principles
  • Collaboration with product managers and application scientists
  • Collaboration with product managers and application scientists
  • Users rely on step-by-step mental validation during assays
  • Clear system feedback builds more trust than advanced visuals

The Breakthrough

The key breakthrough was the shift from complex, time-intensive segmentation to a fast and efficient centroid-based (“Hooke A”) detection approach. This significantly reduced annotation time while maintaining strong accuracy. By combining user input with AI-assisted suggestions, the workflow enabled reliable model training using smaller, optimized datasets. Ultimately, it empowered even non-expert users to build and deploy AI-driven analysis workflows with ease and consistency.

Our Process

Here is my Services. Where you will find my creativity and my working talents.

User Research

Conducted interviews & usability tests

Wireframing

Created new task flows & wireframes

UI Design

Built and modern, intuitive UI

UX Artifacts

Here is my Services. Where you will find my creativity and my working talents.

Outcome & Reflection

The implementation of the centroid-based (“Hooke A”) workflow led to a significant improvement in efficiency, reducing annotation time and accelerating model development without compromising accuracy. The system delivered consistent and reproducible results across varied image conditions, demonstrating strong robustness. By simplifying the process and integrating guided AI tools, it enabled broader user adoption, including those without prior machine learning expertise. Overall, the workflow not only streamlined image analysis but also highlighted the value of intuitive, scalable AI solutions in advancing cytometric research.