Helping You Remember What Matters Most

Objectives

• Develop an AI-based platform to assist dementia patients with memory recall.
• Provide continuous reminders and hints tailored to individual needs.
• Create a user-friendly interface accessible to both patients and caregivers.
• Enhance cognitive support to manage daily life and reduce memory lapses.
• Offer caregiver tools for tracking and supporting patient progress.
• Evaluate the system’s effectiveness in improving memory recall and quality of life.

---

Methodology

Step 1: Memory Upload

Doctors upload photos and a short story describing a memory.

Step 2: Embedding Generation

• Text is converted into vector embeddings using an AI model.
• Embeddings are stored in a vector database for semantic search.

Step 3: AI Agent Interaction

• Patients are shown photos and asked to recall the story.
• AI compares patient responses to stored embeddings.
• If correct → AI confirms memory.
  If incorrect → AI provides hints or prompts to try again.

Step 4: Continuous Learning

AI continuously refines its understanding through a feedback loop, improving recall accuracy over time.

---

Visual Workflow

The workflow diagram below summarizes the entire system pipeline.

  ┌──────────────────────────────┐
  │ Doctor uploads memory photos │
  │ and story description        │
  └──────────────┬───────────────┘
                 │
                 ▼
   ┌────────────────────────────┐
   │ Story converted into       │
   │ vector embeddings          │
   └──────────────┬─────────────┘
                 │
                 ▼
   ┌────────────────────────────┐
   │ Embeddings stored in       │
   │ Vector Database (Pinecone) │
   └──────────────┬─────────────┘
                 │
                 ▼
   ┌────────────────────────────┐
   │ Patient views photo and    │
   │ attempts to recall memory  │
   └──────────────┬─────────────┘
                 │
                 ▼
   ┌────────────────────────────┐
   │ AI agent validates response│
   │ via vector similarity check│
   └──────────────┬─────────────┘
        ┌─────────┴─────────┐
        │                   │
 (✅ Yes - Correct)   (❌ No - Incorrect)
        │                   │
        ▼                   ▼

┌────────────────────┐ ┌───────────────────────────────┐
│ Confirms memory is │ │ Provides hints and asks user  │
│ recalled correctly │ │ to guess again                │
└────────────────────┘ └───────────────────────────────┘

---

Tech Stack

Component	Technology Used	Purpose
Backend	Node.js, Express.js	API and business logic
Frontend	React.js	User interface for doctors and patients
AI Models	Python, Gemma, Google Gemini	Embedding generation and semantic understanding
Database	MongoDB	Patient and memory storage
Vector DB	Pinecone	Semantic search of embeddings
Hosting	Cloud Storage + Secure APIs	Image and data management

---

Outcomes (Till Date)

1. Image Upload & Vectorization
   Successfully implemented photo upload and conversion of stories into vector embeddings for storage and retrieval.

2. Reminiscence Therapy Agent
   Developed an AI-powered question-answer system that engages patients for memory recall.

3. Response Analysis & Feedback
   Implemented timed response logic and integrated hints when incorrect answers are detected.

4. Voice Detection Integration
   Added real-time speech input and analysis for natural recall sessions.

---

Future Scope of Research

1. On-Device Processing
   Edge-based AI for improved privacy and faster responses on personal devices.

2. Privacy Enhancements
   Role-based access control, encryption, and secure cloud storage for sensitive memory data.

3. AI Hallucination Mitigation
   Prevent false memory enforcement by verifying hints and embeddings before confirmation.

---

References

1. General Psychology — Book by Baron
2. Moon S., Lee J.M., Kang M., Kim K. M. (2020). “The effect of digital reminiscence therapy on people with dementia: A pilot study.” The Open Nursing Journal. DOI:10.2174/1874434602014010231
3. Y Pu et al. (2025). “Reminiscence therapy delivery formats for older adults with dementia or mild cognitive impairment: A systematic review and network meta-analysis.” Psychology Journal. ScienceDirect
4. G.T. Grossberg et al. (2021). “A systematic, automated digital reminiscence therapy platform.” Alzheimer’s Journal. Wiley Online Library

---

"Using AI to bring back precious memories — one story at a time."