Reimagining VoxelBox: From Information Overload to Role-Based Clarity

This redesign aimed to reduce friction points observed in VoxelBox v1 and seamlessly integrate VoxelBox into the daily work routine of healthcare professionals

VoxelBox dashboard on a laptop, showing the All Patients table with scan status cards above it

Team

Team

Me (Intern), Senior Designer and 2 PMs

Timeline

Timeline

Nov '24 - Apr '25

My Role

My Role

Concept, UX and Visual Design

Industry

Industry

Neuroscience & Health-tech

Platform

Platform

Web

Project Overview

The Challenge

Healthcare professionals such as neurosurgeons, radiologists, and technicians work on tight schedules, leaving little to no time to learn new software.

In the first release of VoxelBox, they struggled with usability and adoption. This highlighted the need for a redesign that made VoxelBox intuitive, time-efficient, and fit seamlessly in their workflows.

Problem Statement

How do we redesign the VoxelBox dashboard that allows users to get the relevant data as quickly as possible?

Results

The filter and stat card changes shipped during my internship. Usage data showed an increase in filter adoption after the stat cards were introduced; radiologists and admins were now reaching filtered views without being prompted.

The IMS roles and permissions work was handed off before I left. I didn't get to see it in production, but the permission matrix and role-based views were validated with the product and engineering teams beforehand.

Increased

Filter adoption across radiologist and admin users

Reduced

Time spent scanning the table to find actionable scans

Shipped

Role-based permission system across 7 distinct user types

VoxelBox dashboard before the redesign. The All Patients table has Add Filter and Sort By controls above it, and each row shows the patient, study ID, provisional diagnosis, scan date, uploaded files, a process status of Success, Processing or Failed, and a Start Reporting action.

BSAI Dashboard - Before

VoxelBox dashboard after the redesign. Two groups of stat cards sit above the All Patients table: Processing shows Success 06, Failed 07 and Ongoing 12, and Reports shows Prepared 07 and Incomplete 07.

BSAI Dashboard - After

Annotated view of the new stat cards. Note on the Processing cards: radiologists and admins were manually scrolling to find failed or ready scans, so these cards give an at-a-glance count, and clicking any card filters the table below. Note on the Reports cards: report status was the second recurring question, which scans are ready for me to report on, and Prepared and Incomplete mirror that mental model.

New Additions

“ The filters existed. The problem wasn't capability, it was discoverability.”

Radiologists already had an "Add Filter" dropdown - the interaction existed, but it required intent to discover.

Rather than redesigning the filter UI itself (which would fragment the existing internal pattern), we introduced stat cards as a second, passive, but direct entry point.

The count is visible without any action. The filter is one click.

Same logic, lower friction.

As VoxelBox expanded to more hospitals, the same dashboard now had to serve seven different users, each with different information needs and levels of access.

Full view of IMS Dashboard - Subjects Tab

BrainSightAI needed more than a clinical tool.

They needed visibility into billing, scan package tracking, and role-specific access across a growing network of partner hospitals.

The same dashboard became the foundation for an Internal Management System, serving seven distinct roles from radiologists to BrainSightAI's own sales team.

Roles & Permissions

With seven distinct roles across clinical and business functions, the permissions editing flow had to be both granular and manageable.

A radiologist and a psychiatrist both need report access but a BSAI sales admin needs invoice and payment visibility that clinical staff should never see.

The challenge was designing a single editing interface that could handle this complexity without overwhelming a hospital admin managing it.

Roles and Permissions tab for Manipal Hospital. A table lists each role's ID, name and members, its permissions such as Manage Plans, Manage roles and permissions, View subjects and Download Results, the date it was last updated, and an Edit Permissions button.

Managing Roles & Permissions

Onboarding a New Member

Adding a new member to a hospital account required assigning a role at the point of entry, not as an afterthought.

This ensured permissions were set correctly from day one rather than requiring a separate editing step later.

Members tab for Manipal Hospital. A pending invite for John Doe as Hospital Admin has Resend Invite and Revoke Invite buttons. The Members table below lists Nishant Kumar as Hospital Admin in Neurology, and a toast confirms that Nishant Kumar has accepted the invite to join BAI Eloquent as Hospital Admin.

Managing Roles & Permissions

Plans & Pricing

BrainSightAI's two billing models - Pay Per Use and Subscription, needed to be presented clearly enough for a hospital admin to make the right choice without sales intervention.

The plans screen surfaces the key differentiators without overwhelming the user with pricing details.

Plans tab for Manipal Hospital showing an active BAI Eloquent + Cognitive + Rehab plan: the brain networks and tracts it covers, a list of members, pay per use license details such as package ID, plan validity, price per unit and discount, a billing section with a Download Bill button, and a Request Cancellation button.

Plans and Pricing Dashboard

What happened with the design…last I checked

Zoomed-out Figma canvas for the project, with a BSAI Dashboard section and a BSAI IMS section holding rows of screens for the Pay Per Use and Subscription package flows, the create role and edit permissions flow, the invite member flow, and Subjects.

The roles and permissions work was designed and handed off before my internship ended. I didn't get to see it shipped or used in production, so I can't speak to real-world adoption for this part the way I can for the filter/stat card work (which had usage data behind it).

What I'd Do Differently

The filter and stat card work (Part 1) was grounded in real usage data and direct feedback from radiologists and admins - that's what made it a confident decision.

The roles and permissions work (Part 2), by contrast, was based on stakeholder requirements and my understanding of each role's needs, but I never got to validate it with the actual users in each role (a neurosurgeon's needs vs. a psychiatrist's, for instance, may differ more than the spec assumed).

If I revisited this, I'd want to run lightweight usability sessions with at least one person per role before finalizing the permission matrix, even a quick walkthrough would surface mismatches between assumed and actual needs.

Brainstorming

One of the many discussions I had with the senior designer and PM about IMS during the term of my internship.

It was a great learning experience.

Screenshot of a remote brainstorming call: a FigJam board with sticky notes about dashboard levels and roles, next to a Google Meet window where the senior designer is presenting.