Student Innovation-Cutting-edge technology in these sectors continues to be in demand. Recent shifts in healthcare trends, growing populations also present an array of opportunities for innovation.

Description:	Student Innovation-Cutting-edge technology in these sectors continues to be in demand. Recent shifts in healthcare trends, growing populations also present an array of opportunities for innovation.
Organization:	AICTE
Department:	AICTE, MIC-Student Innovation
Category:	Hardware
Theme:	MedTech / BioTech / HealthTechtd>
Youtube Link:	Not provided
Dataset Link:	Not provided
Contact info:	Not provided

SIH 2026 Problem Research Dossier: PulseGrid-MedEdge Point-of-Care Hardware & Health Ecosystem

1. Executive Snapshot

Executive Summary:
Problem statement SIH26215 is an open-theme Hardware innovation challenge under the MedTech / BioTech / HealthTech domain, sponsored by the AICTE, MIC-Student Innovation cell [PS]. It targets pressing national challenges driven by demographic expansion, rural-urban healthcare access disparities, and shifting epidemiological trends (such as the rising burden of non-communicable diseases like hypertension, cardiac arrhythmias, and chronic respiratory disorders) [KNOWN]. To deliver an outstanding, winning submission, a 4–6 student team cannot rely on a generic software mock-up or a simple breadboard sensor loop. The proposed solution—PulseGrid-MedEdge—is a low-cost, ruggedized, portable multi-parameter diagnostic hardware hub integrated with an on-device TinyML anomaly classification engine, an offline-first store-and-forward mobile app for healthcare workers, and an Ayushman Bharat Digital Mission (ABDM) compliant cloud tele-consultation platform [INFERENCE]. The hardware unit captures clinical-grade single-lead ECG, SpO2, heart rate, non-invasive blood pressure (NIBP), and body temperature, executing instant on-device filtering and risk triage even in remote off-grid locations without cellular connectivity [INFERENCE].

Metadata Table

Field	Value	Evidence Tag
Problem Statement Number	SIH26215	[PS]
Title	Student Innovation-Cutting-edge technology in these sectors continues to be in demand. Recent shifts in healthcare trends, growing populations also present an array of opportunities for innovation.	[PS]
Sponsoring Organization	AICTE	[PS]
Department	AICTE, MIC-Student Innovation	[PS]
Category	Hardware	[PS]
Theme	MedTech / BioTech / HealthTech	[PS]
Deadline	30 September 2026	[PS]
Ideas Counter	1/500	[PS]
Scraped Timestamp	2026-09-05	[PS]

Fundamental System Objective:
To engineer a low-cost (< ₹12,000 / $140 USD), portable, clinical-grade multi-parameter vital signs hardware kit powered by on-device TinyML triage, offline-first mobile synchronization, and native ABDM/FHIR interoperability for doorstep primary healthcare delivery in low-resource environments [INFERENCE].

2. Problem Deconstruction

Atomic Requirements Extraction

[PS Statement]: "Student Innovation-Cutting-edge technology in these sectors continues to be in demand. Recent shifts in healthcare trends, growing populations also present an array of opportunities for innovation."

Explicit Requirements [PS]

1. REQ-EXP-01 (Hardware Deliverable): The project must culminate in a physical hardware device/prototype, not a standalone software application [PS].
2. REQ-EXP-02 (Domain Focus): The innovation must directly address MedTech / BioTech / HealthTech domain challenges [PS].
3. REQ-EXP-03 (Institutional Context): The submission is evaluated under the AICTE MIC Student Innovation framework, demanding original engineering, feasibility, and societal impact [PS].
4. REQ-EXP-04 (Healthcare Trend Alignment): The solution must leverage modern technological advancements (e.g., microelectronics, edge computing, biosensors) to solve problems arising from growing population demands and evolving healthcare trends [PS].

Inferred Requirements & Engineering Implications [INFERENCE]

1. REQ-INF-01 (Multi-Vital Acquisition Hub): To be clinically useful for population health, the hardware must integrate biosensors capable of capturing core primary diagnostic vitals: Electrocardiogram (ECG), Pulse Oximetry (SpO2), Body Temperature, and Blood Pressure (NIBP) [INFERENCE].
2. REQ-INF-02 (Edge AI / TinyML Processing): To handle high-rate analog signals (e.g., 250 Hz ECG sampling) and provide real-time diagnostic guidance without active cloud connectivity, the hardware must run lightweight, quantized ML models directly on the microcontroller unit (MCU) [INFERENCE].
3. REQ-INF-03 (Offline Store-and-Forward Telemetry): Because primary healthcare workers (ASHAs/ANMs) operate in rural sub-centers with poor network coverage, the hardware-mobile pipeline must store diagnostic sessions locally and auto-synchronize when connectivity is restored [INFERENCE].
4. REQ-INF-04 (Indian Regulatory & Interoperability Compliance): The architecture must adhere to Central Drugs Standard Control Organization (CDSCO) Class B non-invasive device rules [RESEARCHED] and conform to Ayushman Bharat Digital Mission (ABDM) FHIR R4 data standards [KNOWN].

3. Stakeholders & User Needs

Stakeholder	Goal	Current Pain	System Need	Evidence Tag
ASHA / ANM Field Workers	Conduct rapid health screenings at village doorsteps and log accurate vitals.	Carrying multiple heavy, uncalibrated standalone tools; manual paper register entries; no automated diagnostic assistance.	A single, lightweight, battery-operated hardware kit with 1-click test workflows, audio/visual cues, and offline mobile sync.	[INFERENCE]
Rural Citizens / Patients	Access affordable, timely diagnostic screening for chronic conditions without traveling to urban centers.	Long travel distances (30–50 km) to Primary Health Centres (PHCs), high out-of-pocket expenses, late detection of non-communicable diseases.	Doorstep diagnostic vital checks, instant digital health record creation (linked to ABHA), and immediate triage alerts.	[KNOWN]
PHC Doctors / Medical Officers	Remotely review patient vitals and issue tele-prescriptions with high clinical confidence.	Self-reported or inaccurate patient vital logs; lack of continuous waveform data (e.g., raw ECG traces); unstructured communication over chat apps.	CDSCO-compliant, calibrated sensor data streams, automated TinyML anomaly highlights, and structured FHIR diagnostic reports.	[RESEARCHED]
District Public Health Administrators	Monitor epidemiological trends, detect NCD spikes, and allocate medical resources efficiently.	Fragmented paper records, delayed reporting from rural sub-centers, lack of real-time spatial health data.	Real-time aggregated geospatial dashboard showing screening coverage, disease prevalence heatmaps, and referral tracking.	[INFERENCE]

4. Requirements Traceability

Proposed System Feature	Justifying Requirement	Implementation Component	Verification Method	Evidence Tag
Multi-Parameter Biosensing Hardware Module	REQ-EXP-01, REQ-EXP-02, REQ-INF-01	Custom PCB featuring ESP32-S3 MCU, AD8232 ECG AFE, MAX30102 SpO2/HR module, MLX90614 IR Temp sensor, NIBP pump/sensor.	Benchtop multimeter validation, comparison with gold-standard patient simulators.	[INFERENCE]
On-Device TinyML Arrhythmia Classifier	REQ-EXP-04, REQ-INF-02	Quantized 1D-CNN model running on ESP32-S3 via TensorFlow Lite for Microcontrollers (TFLM).	Confusion matrix testing on MIT-BIH Arrhythmia benchmark dataset.	[RESEARCHED]
Offline-First Mobile App with SQLite Sync	REQ-INF-03	React Native Android app with SQLCipher local storage and background MQTT/HTTPS queue sync.	Offline field simulation (airplane mode data capture followed by reconnection sync test).	[INFERENCE]
ABDM / ABHA ID & FHIR Microservice	REQ-INF-04	Node.js gateway generating FHIR R4 Observation and DiagnosticReport JSON objects; ABDM Sandbox API integration.	ABDM Sandbox API verification suite and FHIR validator tool.	[RESEARCHED]
Web-Based RMP Teleconsultation Portal	REQ-EXP-02, REQ-INF-04	React.js web dashboard with real-time WebRTC video/audio, canvas ECG waveform plotter, and digital prescription builder.	End-to-end simulated clinical teleconsultation session with RMP feedback.	[INFERENCE]

5. Constraints & Environment Realities

1. Connectivity Realities: Rural health sub-centers and tribal hamlets in India frequently experience 0–1 bar 2G or complete cellular outages [KNOWN]. Design Impact: The hardware unit must buffer up to 500 patient diagnostic sessions in onboard SPI Flash memory or local SD card storage, transmitting data to the mobile app via Bluetooth Low Energy (BLE 5.0) without requiring an internet connection [INFERENCE].
2. Power Supply & Grid Instability: Unannounced load shedding lasting 8–12 hours is common in tier-3/rural regions [KNOWN]. Design Impact: The hardware hub must be powered by an internal 3.7V 3000mAh LiFePO4 battery with integrated TP4056/CN3791 charging circuit supporting USB-C and 5V solar panel charging [INFERENCE].
3. Language & Digital Literacy: ASHA workers have varied educational backgrounds and may not be fluent in English [KNOWN]. Design Impact: The mobile app interface must feature multi-lingual support (Hindi, Marathi, Tamil, Telugu, Kannada, Bengali) with voice guidance and visual icon-driven workflows [INFERENCE].
4. Physical & Environmental Field Conditions: High ambient dust, humidity (80%+ during monsoons), and temperatures exceeding 42°C in Indian summers [KNOWN]. Design Impact: The hardware enclosure must be 3D-printed/injection-molded using durable ABS plastic with silicone gaskets (aiming for IP54 ingress protection) and medical-grade silicone cables [INFERENCE].
5. Cost & Scalability Constraints: Existing commercial multi-vital monitors cost upwards of ₹60,000–1,500,000 ($700–$18,000 USD), making wide public health deployment prohibitive [RESEARCHED]. Design Impact: Target Bill of Materials (BoM) cost must be capped under ₹12,000 ($140 USD) by utilizing modular, mass-produced biosensing ICs [ASSUMPTION].

6. Research Questions

1. Sensor Precision & Noise Suppression: What is the optimal Analog Front-End (AFE) configuration for single-lead ECG signal filtering (e.g., AD8232 vs MAX30001) to eliminate 50Hz AC mains interference and muscle tremor artifacts in non-clinical settings? [RESEARCHED]
2. Regulatory Classification & Safety: Does an edge-AI multi-vital screening device fall under CDSCO Class A or Class B risk classification, and what ISO standards (e.g., ISO 13485, IEC 60601-1-2) must be planned for in production? [RESEARCHED]
3. On-Device Machine Learning Execution: How can a 1D Convolutional Neural Network (CNN) for arrhythmia detection be compressed (INT8 quantization) to run within <100 KB RAM constraints on an ESP32-S3 MCU while maintaining >92% sensitivity? [RESEARCHED]
4. Digital Health Ecosystem Integration: What specific APIs and OAuth 2.0 flows are required to fetch ABHA numbers, request patient consent, and upload FHIR-compliant diagnostic bundles to the Ayushman Bharat Digital Mission (ABDM) Health Repository? [RESEARCHED]
5. Market Competitor Gaps: What are the operational limitations of existing Indian MedTech hardware products (e.g., Agatsa SanketLife, Sunfox Spidtech, Tricog InstaECG) regarding multi-parameter integration and offline rural functionality? [RESEARCHED]

7. Data Requirements & Sources

Data Type	Purpose	Ideal Real-World Source	Hackathon Substitute	Access Method	Availability	Privacy Sensitivity	Confidence
ECG Waveform Signal	Real-time cardiac rhythm visualization & arrhythmia model training.	MIT-BIH Arrhythmia Database / PhysioNet [RESEARCHED].	PhysioNet WFDB Python library & live hardware AD8232 output.	Open-access download / Serial BLE stream.	Immediate.	High (Anonymized in public datasets).	[RESEARCHED]
Photoplethysmogram (PPG) / SpO2	Blood oxygen saturation & pulse rate calculation.	MIMIC-III Waveform Database [RESEARCHED].	MAX30102 live sensor readings & synthetic PPG signal generator script.	Open-access download / I2C bus stream.	Immediate.	Medium.	[RESEARCHED]
NIBP / Oscillometric Pressure Data	Systolic/Diastolic blood pressure estimation.	Clinical NIBP patient datasets / NIBP simulator.	Pressure sensor (MPX5050GP) ADC conversion curves & simulated pump cycles.	Microcontroller ADC sampling.	High (Simulated).	Low.	[INFERENCE]
Patient Demographics & ABHA Profile	Patient identification & record linking.	ABDM Sandbox ABHA API [RESEARCHED].	ABDM Mock API Server generating synthetic ABHA profiles & QR codes.	REST API / Mock JSON payload.	Immediate.	Critical (Requires DPDP Act compliance).	[RESEARCHED]
Geospatial Health Screenings	Epidemic monitoring & spatial distribution heatmaps.	District Health Information System (DHIS2) / MoHFW public reports [KNOWN].	GeoJSON generator producing synthetic GPS coordinates within target districts.	Local GeoJSON generation script.	Immediate.	Low (Aggregated data).	[INFERENCE]

Data Classification Breakdown

• A. Explicitly Mentioned Data: None explicitly detailed in the brief PS text [PS].
• B. Required by Inference Data: ECG raw lead voltage, PPG infrared/red AC/DC values, skin temperature (°C), NIBP pressure curve (mmHg), ABHA ID token [INFERENCE].
• C. Data to be Collected: Real-time biosensor streams from team members during live demo testing [ASSUMPTION].
• D. Synthetically Generated Data: Pre-recorded MIT-BIH arrhythmia samples for edge AI classification stress-testing and synthetic district-level patient screening logs [ASSUMPTION].
• E. Data Needing Validation: Clinical accuracy validation datasets against CDSCO-certified hospital monitors [NEEDS VALIDATION].

8. Existing Solutions & Gap Analysis

Existing Solution	Key Features & Architecture	Strengths	Weaknesses	Relevance to SIH26215	Remaining Technical Gap	Source & Confidence
Agatsa SanketLife	Pocket 12-lead ECG device paired with a smartphone app via Bluetooth.	CDSCO cleared, highly portable, well-designed consumer UX.	Single-parameter (ECG only); requires continuous cloud connection for AI report generation; high device cost (> ₹15,000).	Direct competitor in portable ECG space.	Lacks multi-parameter vital signs (SpO2, Temp, NIBP), lacks on-device offline TinyML triage, closed proprietary ecosystem.	[RESEARCHED]
Sunfox Spidtech (Spido)	Ultra-portable 12-lead ECG sensor connecting to mobile phones.	Extremely compact, highly affordable (~ ₹5,000–₹8,000), robust build.	Limited to ECG; reliant on smartphone app for processing; no multi-vital hub; no native ABDM FHIR gateway.	Highly relevant hardware benchmark.	Does not measure NIBP/SpO2/Temp; lacks dedicated edge display/standalone hardware unit; no offline store-and-forward queue for health workers.	[RESEARCHED]
Tricog Health (InstaECG)	Cloud-connected clinical ECG machine for clinics and ambulances.	Medical-grade 12-lead ECG, rapid expert doctor validation (< 10 mins).	High hardware cost (> ₹40,000); mandatory monthly cloud subscription; requires stable 4G network; bulky form factor.	Clinical gold standard comparison.	Not designed for doorstep screening by ASHA workers; zero offline capability; cost prohibitive for mass rural scaling.	[RESEARCHED]

9. Candidate Solution Approaches

+--------------------------------------------------------------------------------------------------+
|                                 CANDIDATE ARCHITECTURAL APPROACHES                                |
+-----------------------------------+--------------------------------+-----------------------------+
| Approach A: Pure Cloud Relay      | Approach B: Smartphone-Centric | Approach C: Dedicated Edge  |
| - Low MCU processing              | - Offloads processing to phone | - On-device MCU TinyML      |
| - Continuous 4G required          | - Bluetooth bandwidth limits   | - Standalone offline storage|
| - Fails in off-grid rural areas   | - High phone battery drain     | - Local screen + ABDM sync  |
+-----------------------------------+--------------------------------+-----------------------------+

Approach 1: Pure IoT Sensor Relay + Cloud AI Processing

• Mechanism: ESP32 hardware reads sensors and streams raw bytes continuously over 4G/Wi-Fi to AWS IoT Core; cloud server processes signals and runs heavy deep learning models [INFERENCE].
• Data Requirements: Continuous high-bandwidth sensor byte stream (250 Hz * 16-bit = 4 kbps per user continuously) [INFERENCE].
• Complexity: Low hardware complexity, high cloud backend complexity.
• Cost: Low initial hardware cost (~ ₹4,500), but high ongoing cloud server & cellular data costs.
• Explainability: High (cloud can run complex XAI algorithms like SHAP/Grad-CAM).
• Scalability & Reliability: Highly fragile in rural settings due to network latency and outages [INFERENCE].
• Hackathon vs. Production Feasibility: Easy to build in a lab with Wi-Fi; totally unviable for real-world rural deployment [INFERENCE].

Approach 2: Smartphone-Centric Processing (BLE Sensors + Mobile App ML)

• Mechanism: Dumb biosensor modules stream raw data via BLE to the ASHA worker's Android phone; the mobile app runs TFLite models and manages storage [INFERENCE].
• Data Requirements: BLE throughput (~20 kbps continuous), local smartphone storage.
• Complexity: Medium hardware complexity, high mobile app software complexity.
• Cost: Moderate hardware cost (~ ₹6,500).
• Explainability: Medium.
• Scalability & Reliability: Dependent on variable smartphone hardware specifications, RAM limits, and BLE connection stability across low-end Android devices [INFERENCE].
• Hackathon vs. Production Feasibility: Moderate build effort; risk of BLE packet drops during ECG waveform transmission.

Approach 3 (Proposed): Dedicated Edge-AI Multi-Vital Hardware Hub + Offline Store-and-Forward Ecosystem

• Mechanism: Dual-core ESP32-S3 MCU executes on-device signal filtering, peak detection, and quantized TinyML arrhythmia triage directly in C/C++; results and raw signals are buffered locally in SPI Flash and synced via BLE to a React Native app with SQLite encrypted storage; data syncs to an ABDM-compliant cloud microservice when online [INFERENCE].
• Data Requirements: Local flash buffering (4 MB), compressed protobuf BLE payload, FHIR JSON backend schema.
• Complexity: High embedded firmware and circuit design complexity; moderate cloud backend complexity.
• Cost: Optimized BoM (~ ₹9,500 / $115 USD) [ASSUMPTION].
• Explainability: High (rule-based deterministic vital bounds combined with edge-ML classification flags).
• Scalability & Reliability: Maximum reliability in off-grid environments; zero performance dependency on phone processing power or active network [INFERENCE].
• Hackathon vs. Production Feasibility: Excellent hackathon presentation (physical hardware working offline live on stage) with a clear, realistic path to CDSCO/ABDM production compliance [INFERENCE].

10. Recommended Approach & Why

Selection: Approach 3 (Dedicated Edge-AI Multi-Vital Hardware Hub)

Justification for Selection:
Approach 3 is selected because it directly solves the environmental realities of rural healthcare delivery in India [INFERENCE]. By performing signal filtering and TinyML classification on the ESP32-S3 MCU, the device operates completely independently of network availability and smartphone performance constraints [INFERENCE]. The hardware hub remains a standalone, reliable clinical instrument, while the mobile app serves purely as an interface for patient record entry and background ABDM cloud synchronization [INFERENCE].

Rejection of Alternatives:

• Approach 1 (Pure Cloud Relay) was rejected because 2G/3G network dead zones in rural health sub-centers cause packet loss, rendering real-time cloud diagnostic streaming impossible [INFERENCE].
• Approach 2 (Smartphone-Centric) was rejected because low-end Android phones supplied to public health workers often suffer from restricted background processing, inconsistent BLE drivers, and high battery drain, leading to dropped diagnostic data during field use [INFERENCE].

11. Technical Architecture

+--------------------------------------------------------------------------------------------------+
|                                    PULSEGRID-MEDEDGE ARCHITECTURE                                |
+--------------------------------------------------------------------------------------------------+

  [ HARDWARE LAYER ]
  +----------------------------------------------------------------------------------------------+
  |  AD8232 (ECG AFE)   |  MAX30102 (SpO2/HR)  |  MLX90614 (IR Temp)  |  MPX5050GP (NIBP Sensor) |
  +----------------------------------------------------------------------------------------------+
                                            | (Analog/ADC, I2C Bus)
                                            v
  +----------------------------------------------------------------------------------------------+
  |  ESP32-S3 Microcontroller (Dual-Core 240MHz, 512KB SRAM, 8MB PSRAM, BLE 5.0)                |
  |  - FreeRTOS Kernel Tasks: Sensor Sampling (250Hz), DSP Filtering, TFT Display Driver          |
  |  - TinyML Engine: Quantized 1D-CNN (TFLite Micro) Arrhythmia & Anomaly Detection             |
  |  - Local SPI Flash Buffer (4MB) for Offline Store-and-Forward                                |
  +----------------------------------------------------------------------------------------------+
                                            | (BLE 5.0 Encrypted Protocol / Protobuf)
                                            v
  [ LOCAL EDGE APP LAYER ]
  +----------------------------------------------------------------------------------------------+
  |  React Native Mobile App (ASHA / ANM Field Tool)                                              |
  |  - SQLCipher Encrypted Local Storage (Patient Profiles, Diagnostic Session Logs)             |
  |  - ABHA ID QR Code Scanner & Demographics Auto-Fill Module                                   |
  |  - Background Sync Manager (Exponential Backoff, Queue Processor over MQTT/HTTPS)             |
  +----------------------------------------------------------------------------------------------+
                                            | (HTTPS / TLS 1.3 - Restored Network)
                                            v
  [ CLOUD SERVICE LAYER ]
  +----------------------------------------------------------------------------------------------+
  |  Node.js API Gateway / Auth (JWT + OAuth 2.0)                                                |
  |  +----------------------------------------------------------------------------------------+  |
  |  | Python FastAPI ML Service | PostgreSQL + PostGIS DB | ABDM FHIR R4 Adaptor (M1/M2/M3) |  |
  |  +----------------------------------------------------------------------------------------+  |
  +----------------------------------------------------------------------------------------------+
                                            | (WebSockets / WebRTC)
                                            v
  [ PRESENTATION & GOVERNANCE LAYER ]
  +----------------------------------------------------------------------------------------------+
  |  React.js RMP Tele-Consultation Web Portal    |  District Public Health Analytics Dashboard  |
  |  - Real-time ECG Waveform Canvas Render        |  - Geospatial Heatmaps & Disease Trends      |
  |  - E-Prescription Builder (List O/A/B)         |  - ASHA Field Screening Progress Tracker     |
  +----------------------------------------------------------------------------------------------+

Detailed Data Flow Narrative

1. Signal Acquisition: The ASHA worker attaches the hardware probes to the patient. The AD8232 AFE conditions raw single-lead cardiac signals, MAX30102 measures optical red/IR absorption, MLX90614 reads infrared body temperature, and the MPX5050GP transducer measures oscillometric cuff pressure [INFERENCE].
2. On-Device Edge Processing: FreeRTOS tasks on the ESP32-S3 sample the ECG at 250 Hz. Digital bandpass filters (0.5 Hz – 40 Hz) strip 50 Hz AC mains hum and baseline wander. The TFLite Micro model evaluates a 2-second ECG window, classifying the trace as Normal, Atrial Fibrillation, Premature Ventricular Contraction (PVC), or Bradycardia/Tachycardia [INFERENCE].
3. Local BLE Synchronization: The hardware displays vitals and risk flags on its built-in 2.4" TFT screen and transmits compressed Protocol Buffer payloads over BLE 5.0 to the React Native field app [INFERENCE].
4. Offline Mobile Storage: The mobile app links the vitals payload with the patient's scanned ABHA ID, encrypts the record using AES-256 via SQLCipher, and saves it to a local sync queue [INFERENCE].
5. Cloud Store-and-Forward Sync: When 3G/4G/Wi-Fi becomes available, the mobile app background worker flushes the queued records to the Node.js API Gateway over HTTPS/TLS 1.3 [INFERENCE].
6. ABDM Mapping & Clinical Portal: The cloud backend converts records into standard FHIR R4 Observation bundles, registers the health document with ABDM Sandbox APIs, and alerts the assigned RMP on the React.js web portal for tele-consultation review [RESEARCHED].

12. Technology Stack

Layer	Selected Technology	Why Selected	Alternative Considered	Why Alternative Rejected	Evidence Tag
Microcontroller / MCU	ESP32-S3 (Dual-Core, BLE 5.0, Vector Extensions)	Integrated BLE/Wi-Fi, hardware AI acceleration vector instructions, low cost (~ ₹450).	Raspberry Pi 4	High power consumption (5V/3A), slow boot time, requires active cooling, high cost (> ₹5,000).	[RESEARCHED]
ECG Analog Front-End	AD8232 Single-Lead Heart Rate Monitor IC	Integrated operational amplifiers, fast right-leg drive (RLD) noise rejection, low cost (~ ₹250).	MAX30001	Excellent biopotential AFE, but significantly higher component cost and complex BGA footprint for prototype hand-soldering.	[RESEARCHED]
SpO2 & Pulse Sensor	MAX30102 High-Sensitivity Pulse Oximeter Module	Combined Red/IR LEDs and photodetector, standard I2C communication, low power.	Analog PPG Discrete Photodiode Circuit	Highly susceptible to ambient light leakage and mechanical misalignment.	[RESEARCHED]
Mobile App Framework	React Native (TypeScript)	Cross-platform, rich ecosystem for BLE libraries, fast UI development for multi-lingual field screens.	Flutter	Good alternative, but team has existing JS/TS expertise and faster SQLCipher plugin ecosystem.	[ASSUMPTION]
Local App Storage	SQLCipher (Encrypted SQLite)	AES-256 transparent database encryption, compliant with DPDP Act data-at-rest rules.	AsyncStorage	Plaintext JSON storage on device storage; vulnerable to root extraction and regulatory non-compliance.	[RESEARCHED]
Cloud Backend API	Node.js (Express) + Python (FastAPI)	Node.js handles async I/O & ABDM API integration efficiently; Python FastAPI runs heavy cloud ML validation.	Ruby on Rails	Slower performance for high-concurrency WebSocket waveform streaming.	[INFERENCE]
Database	PostgreSQL + PostGIS Extension	Relational integrity for patient records, native JSONB support for FHIR payloads, spatial GIS indexing.	MongoDB	Lack of native GIS indexing performance and ACID transaction guarantees for clinical diagnostic workflows.	[RESEARCHED]
Edge ML Engine	TensorFlow Lite for Microcontrollers (TFLM)	C++ micro runtime optimized for ARM Cortex and Xtensa LX7 MCU architectures.	Edge Impulse C++ SDK	Excellent, but TFLM gives direct low-level control over model memory buffers and quantization.	[RESEARCHED]

13. AI/ML Justification

1. On-Device 1D-CNN ECG Arrhythmia Classifier

• Purpose: Automatically detect life-threatening cardiac arrhythmias (Atrial Fibrillation, PVC, extreme Bradycardia/Tachycardia) during initial field screening [INFERENCE].
• Model Type: Quantized (INT8) 1D Convolutional Neural Network (3 Conv layers, 1 MaxPool, 1 Dense layer) [RESEARCHED].
• Input: 500-sample array (2 seconds of single-lead ECG sampled at 250 Hz, normalized) [INFERENCE].
• Output: Multi-class probability vector: [Normal: 0.85, AFib: 0.02, PVC: 0.11, Bradycardia: 0.02] + Diagnostic Risk Flag [INFERENCE].
• Data Required: MIT-BIH Arrhythmia Database (PhysioNet) containing 48 half-hour two-channel ambulatory ECG recordings [RESEARCHED].
• Why AI is Preferable: Classical QRS peak-detection rule engines struggle to distinguish complex morphological anomalies like PVCs or subtle fibrillatory waves under noisy field conditions [INFERENCE].
• Evaluation Metric: Sensitivity, Specificity, and F1-Score on unseen MIT-BIH test beats (Target: > 92% Sensitivity) [RESEARCHED].
• Failure Mode & Fallback: If signal quality is corrupted by motion artifacts, the Signal Quality Index (SQI) module flags "High Noise / Re-take Test" and suppresses the ML inference output to prevent false alarms [INFERENCE].
• Human Oversight: The AI output is strictly Advisory under Indian Telemedicine Practice Guidelines 2020. Final diagnostic confirmation and prescription rest solely with the Registered Medical Practitioner (RMP) [RESEARCHED].

2. Rule-Based Deterministic System (Where AI is NOT Used)

• Deterministic Vital Bounds: Vital parameters such as Body Temperature (> 100.4°F = Fever), Pulse Oximetry (< 90% SpO2 = Hypoxia), and NIBP (> 140/90 mmHg = Stage 2 Hypertension) MUST NOT use machine learning models [INFERENCE]. These thresholds are calculated using deterministic clinical lookup rules [KNOWN]. Using AI for basic thresholding introduces non-deterministic risk and false negatives, which is unacceptable in clinical diagnostics [INFERENCE].

14. Data Architecture & Offline/Low-Network Design

Database Schema (Cloud PostgreSQL + FHIR R4 JSONB)

-- Patients Table
CREATE TABLE patients (
    patient_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    abha_address VARCHAR(50) UNIQUE,
    full_name VARCHAR(100) NOT NULL,
    gender VARCHAR(10),
    date_of_birth DATE,
    contact_number VARCHAR(15),
    created_at TIMESTAMP WITH TIMEZONE DEFAULT CURRENT_TIMESTAMP
);

-- Diagnostic Sessions Table (Stores FHIR JSON + Raw Signal References)
CREATE TABLE diagnostic_sessions (
    session_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    patient_id UUID REFERENCES patients(patient_id),
    asha_id UUID NOT NULL,
    device_hardware_id VARCHAR(50) NOT NULL,
    systolic_bp INT,
    diastolic_bp INT,
    spo2_percentage INT,
    heart_rate_bpm INT,
    temperature_celsius NUMERIC(4,2),
    ecg_arrhythmia_label VARCHAR(30),
    ecg_raw_waveform JSONB, -- Compressed 250Hz sample array
    fhir_observation_bundle JSONB NOT NULL, -- Standard FHIR R4
    sync_status VARCHAR(20) DEFAULT 'PENDING',
    location GEOMETRY(Point, 4326), -- PostGIS Spatial Location
    created_at TIMESTAMP WITH TIMEZONE DEFAULT CURRENT_TIMESTAMP
);

Offline Synchronization & Conflict Handling

1. Local Queue Architecture: Diagnostic sessions captured offline are assigned a cryptographically secure client-generated UUID (v4) and stored in local SQLCipher with a status of SYNC_PENDING [INFERENCE].
2. Synchronization Algorithm: A background task checks for network availability every 60 seconds. When online, records are pushed via HTTP POST in batches of 10 with idempotent transaction handling (using session_id as the deduplication key) [INFERENCE].
3. Conflict Resolution: Since diagnostic screening logs are append-only time-series events, updates to patient demographics use a "Last-Write-Wins" strategy based on cryptographically verified server timestamps [INFERENCE].
4. Bandwidth Reduction: Raw ECG waveforms are compressed on-device using Delta Encoding followed by Base64 encoding, reducing payload size by ~65% before BLE and HTTP transmission [INFERENCE].

15. Security, Privacy & Governance

1. Digital Personal Data Protection (DPDP) Act 2023 Compliance:
   • Consent Management: The ASHA mobile app captures explicit digital patient consent (audio confirmation or digital touch signature) prior to diagnostic screening [RESEARCHED].
   • Data Minimization: Only vitals and mandatory demographic fields required for clinical evaluation are collected [INFERENCE].
   • Data at Rest: Hardware buffers clear raw RAM upon test completion. Mobile app databases use AES-256 SQLCipher encryption. Cloud databases utilize encrypted storage volumes (LUKS / AWS EBS Encryption) [RESEARCHED].
2. Data in Transit: All communication between the hardware hub, mobile app, and cloud servers is enforced over Bluetooth Security Mode 4 (LE Secure Connections) and HTTPS / TLS 1.3 with Certificate Pinning [INFERENCE].
3. Regulatory Compliance & Boundaries:
   • CDSCO Medical Device Rules 2017: The device is designed as a Class B non-invasive diagnostic monitoring system [RESEARCHED]. The SIH output represents a working functional prototype and proof-of-concept [INFERENCE]. Formal commercial distribution would require Form MD-14/MD-22 manufacturing licensing and ISO 13485 QMS certification [RESEARCHED].
   • Telemedicine Practice Guidelines 2020: The platform facilitates tele-consultation by supplying validated vitals to Registered Medical Practitioners (RMPs) [RESEARCHED]. AI flags are explicitly labeled as "Triage Guidance Only" to preserve full RMP diagnostic liability and authority [RESEARCHED].

16. MVP Scope (Hackathon Build)

+--------------------------------------------------------------------------------------------------+
|                                      HACKATHON MVP SCOPE MAP                                     |
+----------------------------------------------------+---------------------------------------------+
| IN SCOPE (MUST & SHOULD BUILD)                     | OUT OF SCOPE (DESCRIBED / PRODUCTION ONLY)  |
| - Custom Hardware Hub PCB Prototype                | - ISO 13485 Cleanroom Manufacturing         |
| - AD8232 ECG + MAX30102 SpO2 + Temp + NIBP Probes | - Form MD-14 CDSCO Regulatory Clearance     |
| - Quantized 1D-CNN Arrhythmia Model on ESP32-S3    | - 12-Lead Multi-Electrode Simultaneous AFE  |
| - React Native ASHA App with Offline SQLCipher Sync| - Direct GSM/eSIM Cellular Module Hardware  |
| - ABDM Sandbox ABHA Linking Microservice           | - Multi-Hospital EHR Enterprise ERP Adapters|
| - React.js Doctor Tele-Consultation Portal         | - Continuous 24/7 Wearable ICU Monitoring   |
+----------------------------------------------------+---------------------------------------------+

In Scope

MUST BUILD (Core Target for Winning Entry)

• Working multi-vital hardware prototype on PCB/perfboard housing ESP32-S3 MCU, 2.4" TFT LCD screen, AD8232 ECG sensor, MAX30102 SpO2 sensor, and MLX90614 Temperature sensor [INFERENCE].
• Embedded FreeRTOS C++ firmware performing 250 Hz ECG signal acquisition, digital filtering, and live waveform plotting [INFERENCE].
• On-device TFLite Micro 1D-CNN arrhythmia classifier distinguishing Normal vs. Abnormal ECG beats [RESEARCHED].
• React Native Android application communicating via BLE to display vitals, log patient profiles offline, and store records in encrypted SQLite [INFERENCE].
• React.js web dashboard for doctors displaying real-time WebRTC teleconsultation video, ECG canvas plots, and e-prescription building [INFERENCE].

SHOULD BUILD

• Integrated NIBP cuff module driven by micro-air pump, solenoid valve, and pressure sensor [INFERENCE].
• ABDM Sandbox API integration for fetching mock ABHA profiles via QR code scan [RESEARCHED].
• District public health heatmap dashboard showing GIS mapping of screening counts and elevated blood pressure/arrhythmia alerts [INFERENCE].

NICE TO HAVE

• Multi-lingual voice assistance prompts on the mobile app (Hindi & English) guiding the field worker through probe attachment [INFERENCE].
• Solar charging circuit integration on the hardware enclosure [INFERENCE].

Described Only / Out of Scope (Do Not Build During Hackathon)

• Production injection-molded IP67 casing [INFERENCE].
• Commercial CDSCO Form MD-15/MD-22 certification and clinical trial documentation [RESEARCHED].
• Full 12-lead simultaneous ECG cable assembly (single-lead/3-electrode modified lead II is sufficient for MVP) [INFERENCE].

17. Live Demo Simulation Plan

To deliver an undeniable physical and digital demonstration in front of SIH judges without relying on venue Wi-Fi or unpredictable medical conditions, the team will execute a structured 5-minute live simulation script:

+--------------------------------------------------------------------------------------------------+
|                                  5-MINUTE LIVE DEMO WORKFLOW                                     |
+--------------------------------------------------------------------------------------------------+
  [MIN 0:00 - 1:00] PHYSICAL HARDWARE SENSING
  - Place finger on MAX30102 (SpO2/HR) & hold ECG leads (AD8232).
  - Show live 250Hz ECG trace and real-time vitals on the built-in 2.4" TFT LCD screen.
  
  [MIN 1:00 - 2:00] ON-DEVICE TINYML TRIAGE (OFFLINE)
  - Toggle a hardware switch to inject a pre-recorded MIT-BIH AFib waveform sample into ESP32 ADC.
  - Show instantaneous on-device TinyML classification flag: "Arrhythmia Detected (AFib: 94%)".
  
  [MIN 2:00 - 3:00] BLE SYNC & OFFLINE-FIRST FIELD APP
  - Enable "Airplane Mode" on the ASHA worker Android phone.
  - Scan synthetic ABHA QR Code; save diagnostic session locally in encrypted SQLCipher.
  
  [MIN 3:00 - 4:00] STORE-AND-FORWARD CLOUD SYNC
  - Disable "Airplane Mode". Show background MQTT queue automatically syncing buffered records to cloud.
  
  [MIN 4:00 - 5:00] DOCTOR PORTAL & ABDM FHIR BUNDLE
  - Open React.js Doctor Portal on laptop; show instant notification, live WebRTC video call,
    canvas ECG trace render, and generated ABDM FHIR R4 JSON object.

18. Implementation Roadmap

Alignment Deadline: 30 September 2026 [PS].

+--------------------------------------------------------------------------------------------------+
|                                     IMPLEMENTATION TIMELINE                                      |
+--------------------------------------------------------------------------------------------------+
  PHASE 1: Hardware Circuit Design & Biosensor Benchmarking (Weeks 1 - 2)
  - Design schematic in KiCad; bench-test AD8232, MAX30102, MLX90614 on breadboard.
  - Benchmark ADC sampling rates and signal-to-noise ratios.
  
  PHASE 2: Embedded Firmware & TinyML Model Quantization (Weeks 3 - 4)
  - Develop FreeRTOS C++ firmware on ESP32-S3 for multi-sensor task scheduling.
  - Train 1D-CNN arrhythmia model on MIT-BIH dataset in PyTorch; quantize to INT8 via TFLite.
  - Deploy TFLite Micro model onto ESP32-S3 SRAM and verify < 20ms inference time.
  
  PHASE 3: Mobile App & Encrypted Local Storage (Weeks 5 - 6)
  - Build React Native app UI for multi-lingual ASHA workflows.
  - Implement BLE 5.0 custom GATT service handler and SQLCipher local database queue.
  
  PHASE 4: Cloud Backend, ABDM Integration & Web Dashboard (Weeks 7 - 8)
  - Develop Node.js/Express REST & WebSocket backend with PostgreSQL/PostGIS.
  - Connect ABDM Sandbox API for ABHA authentication and FHIR bundle construction.
  - Build React.js Doctor Tele-consultation portal with real-time canvas rendering.
  
  PHASE 5: Enclosure Integration, System Testing & Mock Trials (Weeks 9 - 10)
  - 3D-print ergonomic ABS enclosure; assemble battery, PCB, screen, and sensor ports.
  - Conduct full offline store-and-forward field stress tests and false positive/negative validation.

19. Testing & Validation Plan

1. Hardware & Sensor Calibration Testing: Validate AD8232 ECG waveforms against a certified patient simulator (or reference benchtop ECG monitor). Ensure MAX30102 SpO2 readings are within ±2% accuracy compared to a commercial pulse oximeter [INFERENCE].
2. TinyML Model Performance Testing: Evaluate the quantized 1D-CNN model on 5,000 hold-out ECG signal beats from the MIT-BIH dataset. Target metrics: Sensitivity ≥ 92%, Specificity ≥ 94%, Inference Latency ≤ 30 ms [RESEARCHED].
3. False Positive / False Negative Stress Testing: Test the Signal Quality Index (SQI) module by artificially shaking sensor cables and loosening electrode contacts. Verify that noisy signals correctly trigger "Re-take Reading" rather than generating false arrhythmia flags [INFERENCE].
4. Offline Resilience & Sync Validation: Conduct a 100-record continuous offline capture session in Airplane Mode. Restore network connection and verify 100% data recovery with zero duplicate records in PostgreSQL [INFERENCE].
5. Security & Cryptographic Audit: Verify AES-256 encryption on the SQLite database using string-extraction tools on rooted Android test devices. Confirm all cloud API endpoints enforce HTTPS / TLS 1.3 and reject unauthenticated requests [INFERENCE].

20. Risk Register

Risk Event	Likelihood	Impact	Severity	Mitigation Strategy	Evidence Tag
Severe Motion Artifacts corrupting ECG/SpO2 readings in field settings	High	High	High	Implement hardware RC filtering + software Signal Quality Index (SQI) check that rejects noisy frames prior to ML inference.	[INFERENCE]
Bluetooth Low Energy (BLE) packet loss during live wave streaming	Medium	Medium	Medium	Implement BLE packet buffering on ESP32-S3 local PSRAM and apply lightweight Delta compression.	[INFERENCE]
ESP32-S3 SRAM exhaustion running TFLite Micro model	Medium	High	High	Quantize ML model weights to INT8 and restrict tensor arena size to < 80 KB; utilize external PSRAM for display buffer.	[RESEARCHED]
ABDM Sandbox API downtime during live judging demo	High	Medium	Medium	Build an offline/local Mock ABDM Gateway server that returns identical FHIR JSON structures during network failures.	[INFERENCE]
LiFePO4 battery depletion during prolonged field testing	Low	High	Medium	Integrate CN3791 USB-C fast-charging IC supporting pass-through power and 5V solar panel input.	[INFERENCE]

21. Cost & Scalability

Hackathon Prototype Build Cost

Component / Subsystem	Quantity	Estimated Unit Cost (INR)	Total Cost (INR)	Evidence Tag
ESP32-S3 Microcontroller Board (N16R8)	1	₹550	₹550	[RESEARCHED]
AD8232 ECG Sensor Module + 3-Lead Electrode Cable	1	₹350	₹350	[RESEARCHED]
MAX30102 SpO2 / Pulse Oximeter Module	1	₹220	₹220	[RESEARCHED]
MLX90614 Non-Contact IR Temperature Sensor	1	₹650	₹650	[RESEARCHED]
MPX5050GP Pressure Sensor + Mini Air Pump & Valve (NIBP)	1	₹1,400	₹1,400	[RESEARCHED]
2.4" SPI TFT LCD Display Screen	1	₹450	₹450	[RESEARCHED]
3.7V 3000mAh LiFePO4 Battery + TP4056 USB-C Charger	1	₹450	₹450	[RESEARCHED]
Custom PCB / Circuit Board Fabrication + 3D Printed ABS Enclosure	1	₹1,200	₹1,200	[INFERENCE]
Total Hardware MVP Cost	--	--	₹5,320 (~ $64 USD)	[INFERENCE]

Projected Commercial Production Cost (Scale: 10,000 Units)

• Mass-manufactured SMD PCB with integrated components: ~ ₹3,200 / unit ($38 USD) [ASSUMPTION].
• Injection-molded IP54 medical-grade casing: ~ ₹600 / unit [ASSUMPTION].
• CDSCO compliance, testing, and distribution overhead: ~ ₹1,200 / unit [ASSUMPTION].
• Target Commercial Sale Price to Public Health Departments: ₹8,500 – ₹10,000 per unit (representing an 80%+ savings compared to standard clinical monitors) [INFERENCE].

22. Differentiation

+--------------------------------------------------------------------------------------------------+
|                                    COMPETITIVE DIFFERENTIATION                                   |
+------------------+----------------------------------+--------------------------------------------+
| DIMENSION        | GENERIC SIH SUBMISSION           | PULSEGRID-MEDEDGE SOLUTION                 |
+------------------+----------------------------------+--------------------------------------------+
| Hardware Form    | Loose breadboard wires connected | Custom PCB, 3D enclosure, internal battery,|
|                  | to a laptop USB port.            | 2.4" LCD display, standalone field unit.   |
+------------------+----------------------------------+--------------------------------------------+
| AI Architecture  | Cloud API dependency (Fails      | On-device quantized TinyML (INT8 1D-CNN)   |
|                  | offline).                        | running directly on ESP32-S3 MCU.          |
+------------------+----------------------------------+--------------------------------------------+
| Data Handling    | Local mock JSON or Firebase      | Encrypted SQLCipher offline queue +        |
|                  | plain-text logs.                 | ABDM FHIR R4 native microservice.          |
+------------------+----------------------------------+--------------------------------------------+
| Vital Parameters | Single sensor (e.g., just ECG or | Full multi-parameter hub (ECG, SpO2, Temp, |
|                  | just temperature).               | NIBP) in one portable kit.                 |
+------------------+----------------------------------+--------------------------------------------+
| Clinical Safety  | Raw AI output shown directly.    | Deterministic safety bounds + SQI artifact |
|                  |                                  | rejection + RMP advisory boundary.         |
+------------------+----------------------------------+--------------------------------------------+

23. Future Scope

1. Multi-Lead ECG Reconstruction: Upgrade embedded firmware to synthesize a 6-lead ECG from 3 physical electrodes using spatial lead transformation algorithms [RESEARCHED].
2. LoRaWAN / Satellite Telemetry: Add an optional low-power LoRaWAN transceiver (865–867 MHz band in India) to stream critical emergency vital alerts directly from deep forest/mountainous regions without cellular towers [KNOWN].
3. Pneumonia & Respiratory Sound Analysis: Integrate a digital MEMS stethoscope probe into the hardware hub to record lung sounds and execute on-device TinyML classification for early pneumonia and COPD detection [RESEARCHED].
4. CDSCO Class B Formal Certification: Conduct formal clinical validation trials in collaboration with a recognized government medical college to obtain CDSCO manufacturing licensing for large-scale deployment under the National Health Mission (NHM) [RESEARCHED].

24. Assumptions & Open Items Log

Consolidated Assumptions Log [ASSUMPTION]

• ASSUMPTION-01: The student team possesses access to basic electronics laboratory equipment (soldering station, digital multimeter, oscilloscope) and 3D printing facilities.
• ASSUMPTION-02: Single-lead ECG (Modified Lead II configuration) is clinically acceptable for primary community screening and arrhythmia triage before referral to a district hospital for full 12-lead diagnostic ECGs.
• ASSUMPTION-03: ASHA workers will be provided basic Android smartphones running Android 9.0+ with BLE 4.2+ support.
• ASSUMPTION-04: Mass production BoM cost estimates assume bulk procurement of components in batch sizes of 10,000 units.

Open Items & Validation Required [NEEDS VALIDATION]

• NEEDS VALIDATION-01: Clinical accuracy validation of the custom AD8232/MAX30102 PCB assembly against CDSCO-certified hospital monitors under varying skin temperature and motion conditions.
• NEEDS VALIDATION-02: Exact production approval processing timelines for CDSCO Class B medical devices via the SUGAM portal under the Medical Device Rules 2017 updates.
• NEEDS VALIDATION-03: Live throughput tests for ABDM Sandbox M1/M2/M3 production gateways during high-volume simultaneous record pushing.

Final Recommendation

The team should definitely choose to solve SIH26215 by building PulseGrid-MedEdge. Open-theme hardware challenges under AICTE Student Innovation are highly rewarded by judges when a team demonstrates a complete, physically real, low-cost hardware unit combined with sophisticated edge AI and production-ready cloud standards (ABDM/FHIR) [INFERENCE]. Avoid pure software submissions for this problem statement, as the category is explicitly tagged as Hardware [PS].

Recommended Stack

• Hardware/Firmware: ESP32-S3 (Dual-Core MCU), AD8232 (ECG), MAX30102 (SpO2/HR), MLX90614 (IR Temp), MPX5050GP (NIBP), FreeRTOS, TFLite Micro C++.
• Mobile App: React Native, TypeScript, SQLCipher, react-native-ble-plx.
• Cloud Backend: Node.js, Express, Python FastAPI, PostgreSQL, PostGIS, Redis, Docker.
• Web Frontend: React.js, Tailwind CSS, Canvas API, WebRTC.

Biggest Dependency

Availability of a working physical ESP32-S3 development board and functional biosensing modules (AD8232, MAX30102) by Week 2 to enable hardware-software co-design [INFERENCE].

Biggest Risk

Signal noise and motion artifacts corrupting ECG and SpO2 sensor streams during live demonstrations in front of judges [INFERENCE]. Mitigation: Implement hardware RC active filters and an automated Signal Quality Index (SQI) algorithm to discard noisy frames [INFERENCE].

Biggest Technical Challenge

Compressing and quantizing the 1D-CNN ECG arrhythmia model to fit within < 100 KB RAM and executing inference in under 30 ms on the ESP32-S3 MCU without causing task starvation on the FreeRTOS display thread [RESEARCHED].

First 5 Actions

  [ACTION 1] Procure ESP32-S3 MCU, AD8232 ECG kit, MAX30102 SpO2 module, MLX90614 temp sensor,
             and 2.4" SPI TFT LCD screen.
  [ACTION 2] Download the MIT-BIH Arrhythmia Database from PhysioNet and set up a PyTorch 1D-CNN
             training pipeline.
  [ACTION 3] Flash FreeRTOS firmware to ESP32-S3 to verify simultaneous ADC sampling (250Hz) and
             I2C sensor reading.
  [ACTION 4] Register a developer account on the NHA ABDM Sandbox portal to obtain client credentials
             for ABHA linking.
  [ACTION 5] Initialize the React Native mobile app repository and test local AES-256 database read/write
             using SQLCipher.

Validation Before Coding

• Verify sensor signal quality on an oscilloscope to ensure 50 Hz AC mains interference is attenuated by at least 30 dB [INFERENCE].
• Verify that the exported TFLite Micro model compiles cleanly in ESP-IDF or Arduino IDE without memory buffer overflow errors [RESEARCHED].

Buildability Score /100

88 / 100 (Extremely high buildability for a 4–6 engineering student team combining Electronics, Embedded C++, Mobile, and Web skills over a multi-week timeline) [INFERENCE].

Confidence

HIGH (Supported by official SIH26215 metadata [PS], established CDSCO/ABDM regulatory frameworks [RESEARCHED], public PhysioNet clinical datasets [RESEARCHED], and verified hardware component specifications [RESEARCHED]).

Research Sources

1. Central Drugs Standard Control Organization (CDSCO): Guidance Document on Medical Device Software & Medical Devices Rules, 2017.
   URL: https://cdsco.gov.in [RESEARCHED]
   Supported Claim: CDSCO risk-based classification (Class A to D), Software as a Medical Device (SaMD) rules, and SUGAM licensing pathways in India.

2. Ministry of Health and Family Welfare (MoHFW): Telemedicine Practice Guidelines, 2020.
   URL: https://www.mohfw.gov.in [RESEARCHED]
   Supported Claim: Legal framework for RMP remote consultation, List O/A/B prescription rules, patient consent, and AI advisory boundaries.

3. National Health Authority (NHA): Ayushman Bharat Digital Mission (ABDM) Sandbox & FHIR R4 Data Specifications.
   URL: https://sandbox.abdm.gov.in [RESEARCHED]
   Supported Claim: ABHA ID creation, health document linking, and FHIR R4 Observation payload structures.

4. PhysioNet / MIT-BIH Arrhythmia Database: Standardized ECG Waveform Clinical Benchmark.
   URL: https://physionet.org/content/mitdb/ [RESEARCHED]
   Supported Claim: Benchmark dataset for training and evaluating 1D-CNN cardiac arrhythmia classification models.

5. Analog Devices: AD8232 Single-Lead Heart Rate Monitor Front-End Datasheet.
   URL: https://www.analog.com/media/en/technical-documentation/data-sheets/AD8232.pdf [RESEARCHED]
   Supported Claim: Hardware AFE specifications, integrated right-leg drive (RLD) circuit, and 2-pole adjustable high-pass filtering.