Altis Labs raises $25M to make an AI scan read a trial endpoint
A cancer-trial startup wants its model's reading of a CT scan to be the number a drug is judged by — and Apple's Siri home hub is already in employees' hands.
Altis Labs raised $25 million to make an AI reading of a CT scan the thing a cancer trial is measured against. The round, announced September 17 and co-led by OrbiMed and Qiming Venture Partners USA, funds a company whose product is not a drug and not a chatbot: IPRO, a model that reads routine CT scans taken during oncology trials and returns a prediction tied to how long the patient will live. At the IASLC World Conference on Lung Cancer this month, Altis presented data showing IPRO's response rate detected treatment effects that ORR — the objective response rate, the tumor-shrinkage yardstick regulators have used for decades — missed entirely in MARIPOSA, a Phase 3 trial in EGFR-mutated advanced non-small-cell lung cancer.
That is the real argument. ORR is a surrogate: it measures whether a tumor got smaller, not whether the patient lived longer, and the two diverge often enough that drugmakers routinely run multi-year survival follow-ups to prove what a scan already knew. If an AI endpoint tracks overall survival more tightly than the current standard, trials get smaller, read out faster, and fail earlier — which is worth far more to a pharma sponsor than the $25 million is to Altis.
Treat the claim with the usual caution: this is company-presented conference data, not a peer-reviewed result, and the model's value depends on regulators accepting an AI-derived number as a trial endpoint. Altis has spent three years building the case — a Johnson & Johnson Innovative Medicine collaboration, a partnership with University Health Network, and repeated ASCO, ESMO and AACR presentations arguing IPRO beats RECIST and TNM staging at predicting survival. The Series A is the market saying the argument is at least plausible.
Apple's Siri-powered home hub is in employee testing and could launch as soon as October, according to Bloomberg's Mark Gurman. The device, known internally as J490, is a roughly square screen that sits on a stand or mounts to a wall, runs a new operating system built around Siri AI, and pairs with an iPhone for setup. A front camera identifies who is looking at it — without a Face ID-grade depth sensor — and reshuffles the screen around that person's apps, notes, calendar, music and podcasts. Gurman describes it as closest to an Amazon Echo Show with a smaller display.
It is the first hardware category Apple has launched around the new Siri rather than bolting it onto an existing one, and it lands in a market Amazon and Google have owned for a decade without making anyone much money. Apple's bet is that the assistant, not the speaker, is what finally makes a home screen useful. The company has not commented on the report, and the timeline has slipped before.
Which would you trust more in a cancer trial — a radiologist's read of a scan or a model's prediction? Tell us in the comments.
