Comprehensive Analysis of Commercial AI Solutions for ICH Volume Measurement

Overview

This systematic review evaluates FDA- and CE-approved deep learning-based AI software for detecting and quantifying intracerebral hemorrhage (ICH) on non-contrast CT scans. It synthesizes clinical validation data, performance metrics, and regulatory profiles of 11 commercially available tools, highlighting the current landscape and gaps in comparative evaluation.

Background

Intracerebral hemorrhage (ICH) is a critical neurological emergency requiring rapid diagnosis and volume assessment to guide management. Advances in deep learning (DL) have enhanced non-contrast CT (NCCT) capabilities for ICH detection and volumetric quantification, improving triage and radiological workflow efficiency. Multiple FDA- and CE-certified AI software solutions have emerged, but systematic comparative data on their analytical performance and regulatory status remain limited. This review addresses this gap by systematically analyzing literature and regulatory databases to inform clinical decision-making.

Data Highlights

A total of 22 studies were included after a two-step systematic literature review following PRISMA 2020 guidelines. The FDA database search identified 90,834 records, narrowed to 44 using product codes, and further filtered to 11 eligible AI software products for ICH detection and volume quantification. The European EUDAMED database search yielded no comprehensive public data on AI-enabled radiological tools. Six of nineteen manufacturers provided publication lists, aiding in study identification.

Key Findings

• Eleven FDA- and CE-approved AI software tools utilizing deep learning for ICH detection and volumetric segmentation were identified and evaluated.
• Performance metrics across studies demonstrated improved detection accuracy and reliable volume quantification compared to traditional methods.
• The FDA database (1996–2023) was a valuable resource for identifying regulatory-approved AI tools, while the European EUDAMED database currently lacks detailed public information on such devices.
• Most manufacturers used the term 'ICH' interchangeably for intraparenchymal and other intracranial hemorrhage types, complicating precise classification.
• The systematic review highlighted a lack of standardized comparative studies directly evaluating multiple commercial AI solutions head-to-head.

Clinical Implications

Clinicians should be aware of the availability of multiple FDA- and CE-certified AI tools that enhance rapid detection and volumetric assessment of ICH on NCCT, potentially improving patient triage and workflow efficiency. However, the absence of standardized comparative data necessitates cautious interpretation of performance claims and consideration of individual software validation in clinical settings. Regulatory transparency, especially in Europe, remains limited, underscoring the need for improved public access to device information.

Conclusion

This comprehensive review consolidates current evidence on commercially available AI solutions for ICH detection and volume measurement, revealing promising performance but highlighting gaps in comparative evaluation and regulatory transparency. Continued efforts are needed to standardize assessments and improve data accessibility to optimize clinical integration.

References

1. PRISMA 2020 Guidelines -- Systematic Review Reporting Standards
2. FDA Product Code QAS -- Radiological Computer-Assisted Triage and Notification Software
3. European Commission Directorate-General for Health and Food Safety -- EUDAMED Database Status