Ophthalmology on transtorm.ai

Qwen3.8-Flash-Next applies the fundus laterality rule backwards

Sat, 26 Sep 2026 00:00:00 +0000

Qwen3.8-Flash-Next locates the optic disc correctly in 120 decisions out of 120, then gets the eye wrong on 51 images out of 60. The protocol, the numbers and the fix we adopted.

Injection Registry

Wed, 19 Aug 2026 00:00:00 +0000

Context

In an anti-VEGF practice, the injections are routine; keeping a clean longitudinal record is not. Visual acuity sits in one place, OCT measurements in another, the drug and lot number on a label, the next appointment in the front desk’s memory. When the specialist wants to know whether an eye is actually responding to treatment, the data exists, but rarely in front of them in a readable form.

Operational Challenge

Scattered data: Acuity, OCT, drug, lot and appointments live in different systems or on different media.
Fragile data entry: Free text and hand-copied values produce inconsistent records that are hard to work with.
Invisible treatment response: Comparing an eye’s course or outcomes by drug requires manual reconstruction.
Tracking overdue eyes: Spotting eyes that are genuinely behind on their protocol depends on the team’s vigilance.

Solution Architecture

The intravitreal injection registry brings each eye together in a single longitudinal record. Data entry is structured and guided by a clinical context card; fields use controlled vocabularies for drugs and protocols, and carried-over values must be confirmed, never silently inherited. Reading takes a single glance: ETDRS and CST timelines per eye, cohort-level outcome comparison by drug, and a pseudonymized CSV export ready for a clinical study. Everything runs on your own infrastructure.

Kiosk Reception

Wed, 19 Aug 2026 00:00:00 +0000

Context

In a practice, a patient’s arrival sets off a small chain of tasks. The receptionist identifies them, checks their appointment and opens their record, then someone decides which exams to run and adds them. The check-in kiosks on the market handle the start of this chain and stop at the door of the clinical record: adding the exam is still done by hand.

Operational Challenge

A bottleneck at reception: Each check-in takes up about 85 seconds of the receptionist’s time. Five patients arriving together means a queue several minutes long.
Information re-entered by hand: The type of consultation is almost always the one planned at the previous visit. So it is already in the record, but someone still has to look it up and enter it by hand.
Duplicate records: A patient who is not found correctly gets created again, and their history ends up split.
The language barrier: A patient who does not speak the practice’s language needs someone to help them.

Solution Architecture

The patient check-in kiosk recognizes the patient by their appointment time and initials, completed by their year or date of birth: no card, no code, no SMS. It finds the appointment in the practice’s calendars, then the patient’s record, which it creates if it does not exist yet. It works out the type of consultation from the last follow-up and adds the exam itself, with the right practitioner. By the time the patient sits down, their exam is already on the worklist.

LASIK Planning

Wed, 19 Aug 2026 00:00:00 +0000

Context

In refractive surgery, astigmatism is never just a number: behind every cylinder there is an axis, an origin, and a consistency to verify against keratometry. Surgeons already go through this reasoning at every planning session, but their tool is often a hand-filled spreadsheet. For a straightforward myopia, that is enough. As soon as the axes diverge or the cornea fails to account for the full cylinder, the file reaches its limits, and it reaches them silently.

Astigmatism in LASIK Planning: Why the Spreadsheet Has to Go

Sun, 05 Jul 2026 00:00:00 +0200

How a pre-LASIK astigmatism planner replaces the spreadsheet: vector analysis and ORA computed for every eye, all the way to the laser plan.

How a Reception Kiosk Builds the Clinical Worklist Itself

Sat, 06 Jun 2026 00:00:00 +0000

How a patient check-in kiosk for an ophthalmology practice goes beyond mere registration: it identifies the patient, infers the consultation type from the last follow-up, and automatically creates the worklist item in the patient record, with a mobile companion, secured by physical presence, on which patients can see how late their doctor is likely to be running, self-hosted and GDPR-friendly. A comparison with Phreesia, Clearwave, Qmatic and the patient kiosk market.

A Purpose-Built Registry for Intravitreal Injections

Tue, 02 Jun 2026 00:00:00 +0200

How a medical-retina clinic encodes anti-VEGF injections safely and reads ETDRS/CST outcomes at a glance, and how the same engine adapts to clinical research.