ARIA v1.9.1 Build Week Edition

Note: This analysis is based solely on the self-reported, unverified project description provided by the author. No external verification, revenue data, customer base, or historical usage is available.

The description states that ARIA v1.9.1 Build Week Edition is a non-diagnostic coronary CT acquisition support AI for radiologic technologists. It performs two main functions:

• Main Mode: Uses deterministic logic to generate formal acquisition strategies, including:
  • Confirmed rotation and beat strategy
  • Acquired temporal resolution
  • Reference heart-rate range
  • Phase strategy
  • Candidate comparison
  • Selection rationale
  • Decision trace

• Buddy Mode: Provides conversational explanations using GPT-5.6, which receives the formal result as read-only context and does not recalculate or replace it.

The system also includes:

• Bilingual (English/Japanese) interface support
• Real-time streaming responses
• Offline synthetic review workflow
• Optional live GPT-5.6 workflow
• Portable Windows build

Inference: The product is described as a desktop application with a hybrid deterministic + LLM architecture, where the LLM serves an educational or explanatory function rather than decision-making.

The description states that ARIA was created to make coronary CT acquisition planning more structured, reproducible, and explainable without replacing medical professionals. It is positioned as a tool that supports radiologic technologists by standardizing expert knowledge into a deterministic framework while offering conversational explanations through GPT-5.6.

During the Build Week extension:

• The system was enhanced to clearly separate formal strategy generation from explanation.
• The author emphasizes that GPT-5.6 does not recalculate or override decisions, but only explains them.
• The tool is described as non-diagnostic and remains within a strict authority boundary between deterministic logic and LLM interaction.

Inference: The positioning evolved from an existing deterministic system to one with added conversational layer for usability and education, without shifting focus from clinical decision-making responsibility.

The description explicitly identifies the target customer as radiologic technologists who perform coronary CT acquisition planning under time pressure. These professionals must consider factors such as heart rate, variability, motion, phase strategy, temporal resolution, and scanner constraints.

Inference: The ICP is narrow — focused on medical imaging professionals working in cardiac CT environments, particularly those using desktop tools in structured clinical workflows.

Not evidenced.

The description does not mention any pricing model, monetization strategy, or business model details beyond the self-reported nature of the project.

Note: No evidence of revenue streams, customer acquisition costs, or commercial viability is provided.

• Built with Python and several technologies including FastAPI, Streamlit, PyInstaller, GPT-5.6, OpenAI API, local relay, and Windows platform.
• Uses a strict authority boundary:
  • Deterministic logic generates formal result
  • GPT-5.6 explains result as read-only context
  • No recalculation or replacement of decisions
• Includes secure handling of API keys via temporary input and child process communication
• Supports offline synthetic review and portable Windows packaging
• Implements automated regression, routing, packaging, and security tests

Inference: The technical architecture shows a clear separation between deterministic logic and LLM interaction, with emphasis on safety, portability, and secure handling of data.

Not evidenced.

There is no mention of actual users, customers, or adoption beyond the hackathon demonstration. No revenue, headcount, or product usage data are included.

Note: The project appears to be a prototype or proof-of-concept developed for a single event and lacks evidence of real-world deployment or traction.

Not evidenced.

The description does not reference competitors, existing tools in the market, or competitive positioning within the medical imaging or AI-assisted radiology space.

Note: No information is provided about how ARIA compares to other systems or whether similar tools already exist.

• Unverified claims: The system is described as non-diagnostic and not replacing professionals, but this has no independent verification.
• No production history: There is no evidence of prior use beyond the hackathon submission.
• Limited scope: The demonstration uses only synthetic cases with no patient data, which limits understanding of real-world applicability.
• Dependency on proprietary LLMs: Reliance on GPT-5.6 and OpenAI APIs introduces potential dependency risks.
• Single-person team: A team size of one raises questions about scalability, maintenance, and long-term development capacity.

Inference: While the architecture is well-defined, the lack of prior traction or real-world validation makes it difficult to assess commercial viability or risk of failure.

1. Has ARIA been used in clinical settings beyond this demo?
2. What is the current maturity level of the deterministic logic engine outside of the Build Week extension?
3. Are there any plans for integrating feedback from radiologic technologists into future versions?
4. How does the system handle edge cases or unexpected inputs not covered in the synthetic demo?
5. Is there a plan to move beyond the local relay and hosted options?
6. What are the long-term goals for ARIA’s development, including scalability and integration with existing PACS or imaging systems?