Fosvel PSMA
Fosvel PSMA · Prostate cancer PET/CT

Small lights, counted with care.

A research model that aims to outline PSMA-avid findings on whole-body PET/CT, studied on 68Ga-PSMA-11 and 18F-PSMA-1007 scans.

Research use only · Not a medical device · Not CE-marked
Public research data
Why it is needed

A scan full of small things that matter.

PSMA PET/CT has changed how prostate cancer is staged and followed. One scan can show many small lymph nodes and bone lesions, each to be found, placed and counted region by region.

The tracer also gathers in normal glands, kidneys, bowel and some benign bone spots. Reading means telling these apart, lesion by lesion.

What it does today · research
01 · Detect

Head to thigh, in one pass.

One pass over the whole-body scan. The research model aims to outline PSMA-avid findings on 68Ga-PSMA-11 and 18F-PSMA-1007 PET/CT.

Red = what the research model flags. Public research data, 18F-PSMA-1007.
Flagged by the research modelNormal uptakePublic research data, 68Ga-PSMA-11
02 · Tell apart

Disease, or normal uptake?

The model aims to keep findings apart from normal uptake in glands, kidneys and bowel. Small benign bone spots on 18F scans are still its hardest case, and the focus of our current research.

03 · Map

Region by region.

Each finding placed in the regions physicians use to report PSMA scans, in a miTNM-style research list.

T · Prostatelocal
N · Pelvic lymph nodesregional
M1a · Distant lymph nodesdistant
M1b · Bonedistant
M1c · Other organsdistant
Illustration of the report structure, not a patient result.
Left: first scan · Right: second scanSame finding, matched by the research modelPublic research data, 68Ga-PSMA-11
04 · Follow up

The same lesion, found again.

Two scans of the same person, months apart. The research model finds each lesion again on the later scan, so its size can be compared over time.

Tested on public data · research

What we measured, and on what.

Training
420 scans
264 patients · public dataset
Internal test
89 scans
57 patients · internal test · public dataset
Lesions ≥ 0.5 mL found
89%
n = 89 scans · internal test · public dataset
Dice median · internal
0.73
n = 89 scans · internal test · public dataset
False findings / scan · internal
7.3
n = 89 scans · internal test · public dataset
External test
50 scans
external test · 1 public dataset
Dice median · external
0.62
n = 50 scans · external test · public dataset
Key parameter
99%
miPSMA score agreement · n = 2,609 lesions · internal test

Trained on 420 public PSMA PET/CT scans; tested on 89 internal and 50 external scans from 2 public datasets.

Internal: same dataset as training, other patients (slightly optimistic). External: other public datasets. Dice: overlap with expert outlines, 0 to 1. Agreement with expert outlines, not clinical outcomes. miPSMA: PROMISE V2 expression score. Research use only. Not a medical device. Results on public, retrospective datasets; not validated for clinical use.

What we hope next

The whole picture, in the physician's own words.

Our goal: fewer false alarms on small benign bone spots, and structured research reports in the style of the PROMISE framework, followed from scan to scan.

One day, also along PSMA-targeted therapy. These are goals of our research, not features of a product.

Why it matters to us
PSMA is where PET is moving fastest: new tracers, new therapies, more scans every year. A careful reader could save time where it counts.

Show us a hard case.

contact@fosvel.com
Please do not send patient data.