TrustCatPay-it-Forward

Free, private AI “second look” for MRI/CT

Not a diagnosis — just a compass. Upload a de-identified scan, get AI guidance in minutes, and a cryptographic receipt proving what ran and when. Funded by the community. Built on enterprise-grade GPUs.

Open the App Pay it Forward
ENS: trustcat.ethChain: BasePrice: $0 (donation-funded)

How it works

1) Upload (de-ID only)

You provide a de-identified DICOM or ZIP. We reject files with personal identifiers.

2) AI guidance

Models highlight patterns (e.g., adrenal/brain). Output is informational, not a medical diagnosis.

3) Proof & purge

You get a signed receipt (model hash, time, GPU). Files auto-purge within 24 hours.

⚠️ Not medical advice. TrustCat outputs are informational and can be wrong. Do not ignore professional advice. Seek urgent care for serious symptoms.

Privacy first

  • We only accept de-identified studies.
  • Files are processed stateless; default retention ≤ 24h.
  • No advertising, selling, or sharing of data.
  • Receipts contain hashes, never personal info.

Why TrustCat exists

TrustCat turns scars into signal — giving people a free, private AI second look when the system moves too slow.

Open source vaultCryptographic receiptsCommunity-funded

Why TrustCat exists

I’m a dad of four who spent years misdiagnosed with adrenal insufficiency (Addison’s) and Type 1 diabetes — crushed by fatigue and brain fog. TrustCat is a pay-it-forward project so others get signal faster. It’s not medical advice; it’s a private, auditable second look you can pair with a clinician.

No ads or trackers Uploads auto-purge ≤ 24h Receipts you can verify

Under the hood: MONAI

  • MONAI Core pipelines for 3D MRI/CT (resampling, spacing, intensity normalization).
  • Model Zoo / Bundles for reproducible checkpoints & configs; each run logs immutable model/container hashes.
  • MONAI Deploy-style GPU inference containers. No training on your upload by default.
  • Optional MONAI Label for interactive annotation later — strictly opt-in.
  • Verifiable outputs: model+weights+container digests included in your proof for third-party re-run.
✅ We use MONAI ✅ GPU receipts ✅ Opt-in attestation ❌ No training on your files

Learn more at monai.io.

Adrenal & Brain-fog — what we’re screening for (not a diagnosis)

Common symptoms (adrenal insufficiency)

  • Persistent fatigue, weakness, exercise intolerance
  • Dizziness or low blood pressure (orthostatic)
  • Unintentional weight loss, poor appetite, nausea
  • Salt craving, abdominal discomfort
  • Skin changes (e.g., darkening) in some cases

“Brain-fog” can feel like…

  • Trouble concentrating, slower processing, word-finding issues
  • Short-term memory slips; mental fatigue after simple tasks
  • Sleep disruption or non-restorative sleep

Possible drivers to discuss with a clinician

  • Endocrine: adrenal insufficiency (primary/secondary), thyroid imbalance
  • Metabolic/autoimmune: diabetes, B12/iron deficiency, celiac
  • Medication effects (sedatives, anticholinergics, steroids)
  • Neurologic: migraine, prior concussion, small-vessel changes
  • Sleep: apnea, circadian disruption, chronic insomnia
  • Autonomic: POTS/dysautonomia
  • Infection/inflammation: post-viral (e.g., long COVID), CNS inflammation
  • Mood/attention: depression/anxiety/ADHD
  • Other: dehydration, over-training, low caloric intake

⚠️ Not medical advice. These lists are informational only. If symptoms are severe, worsening, or new, seek professional care.

What imaging we accept

✅ Supported (DICOM or ZIP of DICOM)

  • Brain MRI: T1, T2, FLAIR, DWI/ADC, SWI (with/without contrast)
  • Abdomen MRI (adrenal protocol): in-/out-of-phase, T2-weighted, dynamic post-contrast if available
  • CT abdomen (adrenal protocol): non-contrast, portal venous, delayed phases

❌ Not supported (for now)

  • Single JPG/PNG screenshots
  • PACS exports with personal identifiers
  • Ultrasound, PET, plain X-ray

De-identify first. Remove names, dates of birth, MRNs. Keep series/sequence labels. You may compress the study into a ZIP.

What you receive — sample report

TrustCat Report — Example (not a diagnosis) 
Report ID: TC-EXAMPLE-1234
Study: Brain MRI (T1/T2/FLAIR/DWI)
Run date: 2025-08-xx 14:05 UTC
Model: MONAI Bundle <brain_t1_t2_flair_dwi_v1>
Model hash: sha256: <xxxxxxxx>
Container: ghcr.io/trustcat/monai-infer@sha256:<yyyyyyyy>
GPU: RTX 5090 | Driver: <ver>

FINDINGS (AI-assisted, errors possible)
• White-matter hyperintensity pattern: mild, nonspecific distribution (periventricular predominance). 
• No acute diffusion restriction identified.
• No intracranial hemorrhage pattern detected on SWI.

Confidence (0–1): WMH 0.62 | DWI 0.85 | SWI 0.88

RECOMMENDATIONS (for discussion with a clinician)
• Correlate with blood pressure, sleep quality, migraine history, and vascular risk (e.g., A1c, lipids).
• Consider formal radiology read and clinical follow-up if symptoms persist or worsen.

ATTESTATIONS
• Inputs: DICOM hashes (Merkle root): bafy…(CID)
• Outputs: report.json, overlays.nii.gz, thumbnails.zip (CID: bafy…)
• Reproducibility: model+container digests above; rerun is possible by a third party.
✅ MONAI pipelines ✅ Reproducible digests ❌ No training on your upload

Enterprise-grade stack

Compute & orchestration

  • NVIDIA RTX 5090 / 5000 fleet with datacenter power/load monitoring.
  • Kubernetes for isolation, autoscale, and zero-downtime rollouts.
  • Containerized inference (immutable digests) with pinned model bundles.
  • Job queue + receipts: every run emits model & container hashes.
NVIDIA GPUs Kubernetes Container-native

Audit & attestations

  • EAS (Base) attestation option on each job (schema + digest + timestamps).
  • IPFS CIDs for reproducible artifacts (reports/overlays), default purge ≤ 24h.
  • Optional Chainlink proof/verification hooks for external consumers.
  • Metrics: GPU/queue health exposed for transparency (no PHI).
EAS attest IPFS receipts Chainlink (opt-in)

Security & privacy posture

We do

  • De-identify only: uploads must be scrubbed before submit.
  • TLS everywhere + scoped, expiring presigned URLs.
  • Workload isolation per job/tenant; read-only model volumes.
  • Short retention: default purge ≤ 24h; artifacts optional & user-controlled.
  • Immutable provenance: model/container digests & GPU info in the receipt.
Least-privilege Short-lived data Provenance

We don’t

  • We don’t train on your uploads by default.
  • We don’t sell data, run ads, or embed trackers.
  • We don’t keep long-term copies without explicit consent.
  • We don’t upsell gray-hat “recovery” or spam follow-ups.
No trackers No upsells No data resale

Why TrustCat vs hyperscalers

Transparency

  • GPU receipts & reproducible digests
  • ENS-anchored identity (trustcat.eth)
  • Community-funded, open roadmap

Control

  • Edge-first, no silent model swaps
  • Opt-in attestations (EAS/Chainlink)
  • Short retention; user chooses artifacts

Cost & lock-in

  • Flat, fair pricing; no surprise egress
  • No forced accounts across 6 services
  • No data-network effects used against you
Edge, not black box Verifiable runs Fair pricing

Enterprise-grade stack

Compute & orchestration

  • NVIDIA RTX 5090 / 5000 fleet with datacenter power/load monitoring.
  • Kubernetes for isolation, autoscale, and zero-downtime rollouts.
  • Containerized inference (immutable digests) with pinned model bundles.
  • Job queue + receipts: every run emits model & container hashes.
NVIDIA GPUs Kubernetes Container-native

Audit & attestations

  • EAS (Base) attestation option on each job (schema + digest + timestamps).
  • IPFS CIDs for reproducible artifacts (reports/overlays), default purge ≤ 24h.
  • Optional Chainlink proof/verification hooks for external consumers.
  • Metrics: GPU/queue health exposed for transparency (no PHI).
EAS attest IPFS receipts Chainlink (opt-in)

Security & privacy posture

We do

  • De-identify only: uploads must be scrubbed before submit.
  • TLS everywhere + scoped, expiring presigned URLs.
  • Workload isolation per job/tenant; read-only model volumes.
  • Short retention: default purge ≤ 24h; artifacts optional & user-controlled.
  • Immutable provenance: model/container digests & GPU info in the receipt.
Least-privilege Short-lived data Provenance

We don’t

  • We don’t train on your uploads by default.
  • We don’t sell data, run ads, or embed trackers.
  • We don’t keep long-term copies without explicit consent.
  • We don’t upsell gray-hat “recovery” or spam follow-ups.
No trackers No upsells No data resale

Why TrustCat vs hyperscalers

Transparency

  • GPU receipts & reproducible digests
  • ENS-anchored identity (trustcat.eth)
  • Community-funded, open roadmap

Control

  • Edge-first, no silent model swaps
  • Opt-in attestations (EAS/Chainlink)
  • Short retention; user chooses artifacts

Cost & lock-in

  • Flat, fair pricing; no surprise egress
  • No forced accounts across 6 services
  • No data-network effects used against you
Edge, not black box Verifiable runs Fair pricing