The problem
Most of us read inside software built by people who have never had 40 unread studies at 6 pm. It shows. I lose time re-hanging studies and moving between PACS and dictation, so I built Slate.MD to get that time back.
What Slate.MD is
Slate.MD combines a PACS with dictation and reporting for teleradiology groups, imaging centers and community hospitals. I still read daily and test the package on my three-monitor rig, including chart context, the worklist, dispatch and admin.
What works today
The read
Hanging protocols fill my monitors. Scrolling runs from workstation memory. In STAT mode, Sign opens the next study. One key opens 2x2 MPR with draggable crosshairs.
The dictation
Slate punctuates my dictation and routes findings into organ sections. The speech engine uses the exam's vocabulary and the words I've taught it.
Generate drafts an impression from my findings. If the draft mentions a measurement or an organ I never dictated, it stops and shows me. I read every impression before I sign.
Words appear while I talk, then a second pass cleans them up. Uncertain words turn amber. Fix a word twice and Slate learns it, or once with Always fix.
The report
- I place or dismiss unassigned findings before signing.
- I use my own template, word for word. Findings carried from a comparable prior keep their measurements, sides and negations. I confirm carried measurements before signing.
- What I sign is the text on my screen, byte for byte. Signing takes one click.
- Dictating a critical-result call creates its documentation line.
The chart
The chart pane links each line to its source. It runs on synthetic test data today; a pilot uses your site's HL7 feed.
Available to configure
One on-site appliance hosts the archive and models for speech recognition and drafting.
Remote readers keep temporary image caches. Workstations retain unsigned drafts for crash recovery until signed or cleared there.
I go through remote reading, exports and EHR connections with your site before they turn on. Your site controls share links, disc exports and report delivery.
- HL7 contact lookup includes roles, callback numbers and source freshness. I confirm the contact and place the call.
- Report delivery can hold a report, recall it before it sends, and retry. It stays off until your site sets it up.
- The optional credentialing workspace holds records, documents, reviewed AI suggestions, forms, tasks and unsigned packets. Provider connections require organization setup. Suggestions do not verify credentials.
What's next
- Chart checks are built and tested, awaiting activation.
- SMART on FHIR launch and live HL7 need site registration, configuration and EHR-team validation.
- Triage flags on the worklist will come from FDA-cleared vendors.
Where it stands
The reading loop runs end to end on real, de-identified studies. Slate.MD isn't at a customer site yet.
My viewer isn't FDA-cleared for primary diagnosis. Pilot primary reads will use an embedded FDA-cleared viewer.
Pilots are paid at design-partner pricing, with my cell number and a weekly call.
What I'd love from you
Give me 30 minutes. I'll read a real case in front of you, open to sign.
Tell me what interrupts your reading. Send this to a group owner who should see it.
Eli Lim, MD · eli@slate.md
Group inquiries · hello@slate.md