FOUNDER
Dr Yiannis Skarparis
Consultant Interventional Radiologist
1.Founder of The Radiology Academy
2.Curates the evidence pipeline
3.FRCR trained
Your evidence-
backed
second opinion.
The radiology AI that refuses to make things up. Every claim traced, tiered and checked against curated medical evidence — built by a consultant radiologist for the whole imaging team.
9:41•••
MyRadAssistant
Verified with caveats
After TACE for hepatocellular carcinoma, what does the guidance recommend for assessing treatment response on follow-up imaging — and how do CT and MRI compare for detecting residual or recurrent tumour?
How I approached this · 72s
Evidence: Moderate
Key points
- Assess with EASL/mRECIST viability criteria on dual-phase CT or MRI; the LI-RADS Treatment Response Algorithm gives a standardised Viable / Nonviable framework.
- MRI is more sensitive than CT for residual tumour — especially after Lipiodol TACE, where oil obscures enhancement on CT.
Show full answer — evidence & limitations
Continue with: What interval is recommended for follow-up imaging after treatment?
87% confidence24 verified29 citations · 8 sources
9:41•••
REPORT CRITIQUE
CTPA · query PE
Needs improvement5/10
Ambient dictation
Talk naturally — the report builds itself
?
2 CRITICAL5 MAJOR0 MINOR
OriginalIssues (7)Draft
CRITICALFindings section
No assessment of RV/LV ratio documented — needed for PE risk stratification.
CRITICALTechnique section
No comment on contrast opacification quality or timing adequacy.
MAJORFindings section
Subsegmental pulmonary arteries not explicitly evaluated.
33 RCR criteria checked
Apply to draft
Licensed content sources
Why
MyRadAssistant?
Built around four guarantees that general AI can't make.
01
Traced
Every claim links to source text. No fabricated citations.
02
Checked
Hallucination checks cross-reference each response.
03
Honest
Confidence shown, not hidden. Tier 1–3 on every answer.
04
UK-first
NICE, RCR, NHS pathways prioritised by default.
Three tiers, every answer.
Tier 1
Guidelines · Meta-analyses · RCTs
Highest-quality evidence and direct citation coverage.
Tier 2
Cohort & case-control
Strong observational support, transparent limits.
Tier 3
Case reports · Expert opinion
Clearly labelled, lower confidence.
Every modality.
Every role.
Every report.
Fast, evidence-based answers, cited back to source — for radiologists, reporting radiographers, sonographers and radiology nurses alike.
01
For the radiologist
Building a differential.
“Ring-enhancing lesion — discriminators between abscess and high-grade glioma?”
02
For the reporting radiographer
Signing off a chest film.
“Incidental nodule on a chest film — when do BTS guidelines advise CT follow-up?”
03
For the sonographer
At the machine.
“Sonographic features of acute cholecystitis — key signs and pitfalls?”
04
For the radiology nurse
On the unit.
“Contrast extravasation — what does RCR guidance say about aftercare?”
Different roles, one evidence base.
Built by a consultant radiologist. Made for the whole imaging team.
FRACTIONAL CTO
Mr Rory Oliver
Fractional CTO for HealthTechs
1.AI · Cloud · Cyber · React
2.Behind GPNotebook (75% of UK GPs)
3.Healthcare-grade infra & data security
UI/UX CONSULTANT
Mrs Christine Lancaster
Product Designer
1.Product Designer across South Africa, GCC, Switzerland & the US
2.Mobile · Web · Design Systems · Startup Product Design
3.Scaled social platform growth by 132% DAU in 2 months
CSO
Dr Ben Stenberg
Consultant Sonographer
1.New technology implementation
2.Founder of the BMUS AI group
3.Over 25 years of Radiology experience
Regulated,
registered, verified.
Only credentials we currently hold are listed here, each independently verifiable. Anything in progress is deliberately left off.
MHRA Class I
Retrieval & answer function registered with the MHRA as a Class I medical device (UK MDR 2002).
MHRA register ↗Content licences
Full content licences for the RCR iRefer referral guidelines and Radiopaedia.
Licence on file
FAQ:
Everything about
MyRadAssistant
MyRadAssistant uses retrieval-augmented generation (RAG). Before answering anything, it searches a curated radiology and clinical evidence base for relevant source material, then asks a language model to synthesise an answer using only that retrieved text, with citations linking each claim back to the source. A generic LLM like ChatGPT generates from learned patterns alone, which is why it confidently invents references and conflates similar-sounding studies. RAG anchors every answer to text we actually have. If we don't have it, the system says so rather than guessing.
Stop second-guessing.
Get a verified second opinion.
Cancel anytime