---
title: "Northeastern's RECOVER Uses GPT-4o to Monitor Cancer Patients at Home"
subtitle: "A voice-based chatbot plus clinician dashboard uses chain-of-thought prompting to cover 94% of postoperative check-in questions with zero hallucinations across 200 samples."
url: "https://alphasignal.ai/news/northeastern-s-recover-uses-gpt-4o-to-monitor-cancer-patients-at-home"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-04-15T00:00:00.000Z"
updated: "2026-08-23T21:30:41.805Z"
type: "Paper"
topics:
  - "Llms"
  - "Agents"
subtopics:
  - "Vision Language"
  - "Hallucinations"
  - "Tool Use"
upvotes: 3920
source_url: "https://arxiv.org/abs/2502.05740"
comments: 0
truncated: true
---

# Northeastern's RECOVER Uses GPT-4o to Monitor Cancer Patients at Home

> A voice-based chatbot plus clinician dashboard uses chain-of-thought prompting to cover 94% of postoperative check-in questions with zero hallucinations across 200 samples.

Postoperative gastrointestinal cancer patients live in a dangerous window. GI cancer causes roughly 3.4 million deaths per year worldwide, and 5-year survival rates fall below 30% in many cases after surgery. Complications like sepsis or anastomotic leakage can appear suddenly at home, but the tools clinicians use to monitor recovery (phone calls, patient portals, long questionnaires) either exhaust patients or drown providers in unstructured data. A team from Northeastern, Johns Hopkins, MedStar, and UC Irvine built [RECOVER](https://arxiv.org/abs/2502.05740), an LLM-powered remote monitoring system, to test whether GPT-4o can close that gap without introducing new safety risks.

## Two interfaces, one language model

RECOVER splits the workload between patients and providers. Patients talk to an Alexa Echo Dot each day, activating the skill with "Alexa, Open Recover Bot," while clinical staff review results through a web dashboard that shows symptom severity with color-coded dots and LLM-generated summaries. The backend runs on Flask with PostgreSQL, and the OpenAI service is hosted on Azure under a HIPAA-compliant configuration that keeps conversation logs out of third-party hands.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [RECOVER: Designing a Large Language Model-based Remote Patient Monitoring System for Postoperative Gastrointestinal Cancer Care](https://arxiv.org/abs/2502.05740)
- [RECOVER (HTML version)](https://arxiv.org/html/2502.05740)
- [PACMHCI Publication (DOI 10.1145/3788068)](https://doi.org/10.1145/3788068)