DeepBT® Brain Tumor AI System
The flagship brain-tumor AI module on the Liger platform
DeepBT® integrates tumor detection, segmentation, classification, localization, measurement, tracking and report drafting into an end-to-end clinical workflow. Detector Plus holds U.S. FDA 510(k) and TFDA clearance; the 3rd-generation Detector A+ supports five tumor types and is under TFDA review. All AI results are preliminary references — physicians retain final review and confirmation.
- FDA 510(k) · K252190
- TFDA · 008460
- 5 tumor types (A+, in review)
From Detector to Detector A+
Three generations: more tumor types, classification and LLM reporting
| Capability | Gen 1 · Detector | Gen 2 · Detector Plus | Gen 3 · Detector A+ |
|---|---|---|---|
| 3 tumor types · pre-op contouring | supported | supported | supported |
| 3 tumor types · pre- & post-op contouring | not supported | supported | supported |
| 5 tumor types · pre- & post-op contouring | not supported | not supported | supported |
| 5-class tumor classification | not supported | not supported | supported |
| LLM-assisted report drafting | not supported | supported | supported |
| Regulatory status | TFDA cleared · 007906 | TFDA 008460 + FDA 510(k) K252190 | Under TFDA review (not yet cleared) |
DeepBT® Detector A+ (Gen 3) is under TFDA review and not yet cleared; the performance shown is documented / internal-validation data.
Five common brain tumor types
Detection, contouring and classification (Detector A+)
Compared with most brain-tumor AI that supports a single tumor type or task, DeepBT® Detector A+ offers broader disease coverage and differentiated clinical value.
T1W+C + T2W
DeepBT® uses T1W+C and T2W bi-parametric input, integrating more imaging information than single-parameter products — a world-first, core patent-protected technique.
T1W+C
Post-contrast T1
T2W
T2-weighted
More than segmentation — an end-to-end clinical pipeline
Detect → segment → classify → locate → review → measure → report
- 01
Tumor detection
- 02
Initial segmentation
- 03
False-positive filtering
- 04
5-class classification
- 05
Brain-region & intra/extra-axial localization
- 06
Physician review
- 07
Measurement & tracking
- 08
LLM report drafting
Performance (Detector A+ · documented data)
Internal / documented validation — under regulatory review
Performance summary from multicenter retrospective validation; pending regulatory / clinical / QA review before formal publication
Patient-wise Sensitivity
病患層級敏感度
Pending reviewSpecificity
特異度
Pending reviewLesion-wise Sensitivity
病灶層級敏感度
Pending reviewFP / scan
每次掃描偽陽性
Pending reviewDice Coefficient
Dice 係數
Pending review5-class overall accuracy
五類腦瘤總體辨識率
Pending review- ~36s
- Inference / case
- 1,000+
- MR images auto-processed
- 5.37→2.23
- Report turnaround (days, median)
- −58%
- Turnaround reduction
AI-generated contours are preliminary references only and must be reviewed, confirmed, or modified by qualified medical professionals before clinical use.
All AI-generated contours must be reviewed, confirmed, or modified by qualified medical professionals before use in downstream treatment-planning workflows.
A two-way physician–AI workflow
AI proposes; the physician reviews, corrects and confirms
AI inferences are reviewed, corrected and confirmed by physicians; confirmed results form traceable data supporting quality monitoring and product improvement — a closed loop that connects AI with clinical judgment.
AI generates detection, segmentation, classification and measurements
Physician reviews and edits in the existing viewer / workstation
After confirmation, a standardized, traceable report draft is produced
Physicians retain final review, correction and confirmation — consistent with AI-assisted clinical decision principles.
Want DeepBT® in your hospital workflow?
Reach out about clinical deployment, a product demo, or how DeepBT® runs on the Liger platform.