MycoSpec — Fungal Monitoring Device

The Problem

Invisible threat,
real consequences

HVAC systems create warm, humid, low-light environments — ideal conditions for fungal proliferation. Yet most buildings have no monitoring system specifically designed to detect mold-conducive conditions before growth becomes visible.

For immunocompromised patients — post-transplant, chemotherapy, or on immunosuppressants — inhaled Aspergillus spores can cause invasive aspergillosis, a life-threatening pulmonary infection with mortality rates exceeding 50% in at-risk populations.

Climate change is accelerating this risk. Severe storms like those that flooded Duke Medical Center in winter 2024 dramatically increase post-flood mold proliferation in hospital infrastructure.

⚠ Clinical Consequence

Undetected HVAC mold can cause invasive aspergillosis in immunocompromised patients. This was the motivation behind MycoSpec — a real, preventable harm.

47%

of US homes contain detectable mold or dampness

4.6M

US asthma cases attributed to indoor mold exposure

>50%

mortality rate of invasive aspergillosis in immunocompromised patients

hospital-grade HVAC mold monitors currently on the market at affordable scale

Hardware

The Device

A compact, clip-on sensor module engineered to mount directly on standard HVAC vent fins — no tools, no drilling, no disruption to hospital operations.

CAD diagram of MycoSpec device with annotated components

MycoSpec mounted on HVAC vent in real environment

🏗

3D Printed Enclosure

PLA enclosure protecting internal electronics. Designed for low-profile mounting against standard HVAC vents. Future iterations target IP54-rated injection-molded housing.

🔩

Stainless Steel Mounting Brackets

Tool-free clip-on attachment system. Metal hooks insert into HVAC fin slots for secure, non-invasive mounting. Accommodates standard residential and commercial vent profiles.

💨

MQ138 VOC Sensor

Primary mold-detection sensor. Sensitive to volatile organic compounds (VOCs) associated with fungal metabolic activity, including ethanol, acetone, and terpene byproducts.

🧪

MQ3 Sensor

Secondary VOC sensor providing redundant readings. Cross-referenced with MQ138 data to improve threshold accuracy and reduce false positives.

🌡

Temperature + Humidity Sensor (DHT)

Monitors ambient HVAC conditions. Aspergillus fumigatus proliferates optimally at 70–90% humidity and 30–45°C — sensor alerts when conditions enter this danger zone.

📡

ESP32 Microcontroller + WiFi

Transmits live sensor data over WiFi to the MycoSpec web dashboard. Readings logged every 5 seconds for continuous monitoring. Future work targets improved power and connectivity stability.

Operation

How It Works

Five steps from installation to mold risk reading. No technical expertise required — designed for clinical and residential lay users.

📌

Mount Device

Insert stainless steel hooks into HVAC vent fins until the device clips securely into place.

📱

Scan QR / Visit Dashboard

Scan the QR code on the device or visit the web dashboard directly to access live readings.

🟢

Enable Live Readings

Toggle live mode on the dashboard to begin continuous sensor data collection from the ESP32.

⏱

Wait 10 Minutes

Allow the device to collect a full data window. Sensors stabilize after initial power-on drift period.

📊

Read Risk Level

Dashboard displays LOW / MEDIUM / HIGH mold risk based on MQ138/MQ3 PPM thresholds and environmental conditions.

Validation

Testing & Results

Controlled experiments conducted at Duke's Washington Lab facilities using positive and negative mold culture controls to validate sensor thresholds.

Positive control — sealed container with Aspergillus mold cultures

Left: Positive control — active Aspergillus cultures. Right: Laboratory testing environment at Duke Washington Lab.

The study design used two sealed container environments: a negative control (no mold cultures) and a positive control (Aspergillus fumigatus cultures on PDA/MEA/YPA agar plates). Both MQ138 and MQ3 sensors were logged simultaneously over ~350 seconds.

Sensor PPM readings — MQ138 and MQ3 positive vs. negative controls over time

MQ138 & MQ3 PPM readings (7.5s – 350s). Positive and negative controls show clearly distinct bands.

MQ138 Negative Control ~10.5–12.0 PPM stable band → LOW threshold

MQ138 Positive Control ~11.0–12.4 PPM, slightly elevated vs. negative

MQ3 Negative Control ~15.1–16.4 PPM stable baseline

MQ3 Positive Control ~15.8–17.2 PPM, consistently above negative

The separation between positive and negative control bands was used to define min/max detection thresholds. The algorithm takes the lowest value of the negative control and highest value of the positive control to set LOW / MEDIUM / HIGH classification boundaries.

Engineering Realities

Major Technical Challenges

Building a reliable mold detection system required overcoming three fundamental hurdles — each exposing real-world constraints that shaped the final design of MycoSpec v1.

Challenge 3.1

False Positives & Sensor Specificity

MQ-series VOC sensors are broad-spectrum by design — they respond to a wide range of volatile organic compounds, not exclusively mold metabolites. In real HVAC environments, this creates noise from cleaning agents, occupant activity, off-gassing materials, and cooking byproducts, all of which can mimic mold-like VOC signatures and trigger false alarms.

This was encountered directly during early lab testing, where ambient lab chemicals produced elevated PPM readings indistinguishable from early mold activity at a single-sensor level.

Dual-sensor fusion (MQ138 + MQ3) with correlated threshold logic reduces false positive rate. Both sensors must independently exceed baseline before elevating risk level. Positive/negative control calibration anchors thresholds to actual mold culture data rather than arbitrary PPM values.

Challenge 3.2

Mold Species Differentiation

Different fungal species emit distinct VOC profiles. Aspergillus fumigatus — the focus of MycoSpec v1 validation — produces a characteristic mix of sesquiterpenes and alcohols. Other clinically significant species such as Stachybotrys and Penicillium emit different compound profiles at different concentrations.

The current system detects "mold-like conditions", not a specific species. It is presently unknown whether the PPM thresholds derived from Aspergillus cultures generalize to other species — creating a risk of both missed detections and false positives depending on the species present.

Acknowledged as a known limitation of v1. Future work (see Roadmap) includes multi-species validation with Stachybotrys, Penicillium, and Cladosporium to determine whether species-specific threshold bands are needed, or whether a unified mold presence score is achievable.

Interpretation

Mold Risk Levels

The dashboard outputs one of three risk classifications based on aggregated sensor data from both VOC sensors and environmental conditions.

Low

VOC levels fall within the negative control baseline band. Temperature and humidity are outside the optimal fungal growth range. No immediate action required.

→ Continue routine monitoring

Medium

VOC readings are elevated above baseline and approaching positive control levels, or environmental conditions (humidity 70–90%, temp 30–45°C) fall within the optimal Aspergillus growth window.

→ Inspect HVAC system; increase monitoring frequency

High

VOC readings match or exceed positive control levels. High probability of active fungal metabolic activity present. Immediate professional assessment recommended.

→ Contact facility management; refer to EPA mold guidelines

For remediation guidance at MEDIUM–HIGH risk: EPA Mold Cleanup Guidelines →

Roadmap

Future Opportunities

MycoSpec v1 is a functional proof of concept. The following engineering improvements are prioritized for future iterations.

🛡 IP54 Environmental Rating

Upgrade housing to meet IP54 standards — full dust ingress protection and splash water resistance. Critical for real HVAC environments with condensation and particulate exposure.

💉 Plastic Injection Molding

Transition from 3D-printed PLA to injection-molded ABS or PC enclosures for mass producibility, tighter tolerances, and improved surface finish.

⚡ Stable Power Architecture

Explore PoE (Power over Ethernet), battery-backed power supplies, or USB-C PD to eliminate inconsistent sensor readings caused by voltage fluctuations in current USB power setup.

📡 Improved WiFi Stability

Streamline ESP32 data pipeline with local buffering and reconnection logic. Investigate BLE-to-gateway fallback for robust hospital WiFi environments.

🍄 Multi-Species Validation

Expand testing beyond Aspergillus fumigatus to include Stachybotrys, Penicillium, and Cladosporium. Refine PPM thresholds with larger sample sizes to reduce false positives.

📦 Modular Form Factor

Design a modular sensor cartridge system to enable field-replaceable sensors and future-proof the hardware platform against sensor technology improvements.

The Team

Built at Duke

MycoSpec was developed for EGR490/590 Climate Tech Studio at Duke University, in collaboration with the Washington Lab.

⚙️

Saranyaa Kashyap

M.Eng Mechanical Engineering '26 · Hardware & Enclosure Design

💻

Felix W.

B.Eng Electrical & Computer Engineering '29 · Firmware & Dashboard

🔬

Rose

BS Marine Biology '27 · Testing & Validation

Testing conducted at Duke University Washington Lab facilities. Advised on clinical context by Duke Hospital stakeholders.

Team Process

How We Built It

MycoSpec was a fully cross-disciplinary build — mechanical, electrical, biological, and software work developed in parallel across a three-person team. Each artifact below reflects a distinct workstream that had to come together in the final integrated device.

Circuit wiring diagram showing ESP32 connected to MQ-138, MQ-3, DHT22, and SD card

Electrical

Wiring & Firmware Architecture

The ESP32 microcontroller serves as the central hub: it reads analog PPM values from the MQ-138 (VOCs) and MQ-3 (alcohol/ethanol) via GPIO pins A0/A1, collects temperature and humidity from the DHT22 on GPIO 5, logs all data to an SD card over SPI, and transmits readings via WiFi to Google Sheets every 2 seconds. Each reading is averaged over n=10 samples to reduce noise. Both MQ sensors require a 10 kΩ load resistor and a clean-air R₀ calibration step before deployment.

VOC compound sensitivity table comparing MQ-2 through MQ-138 sensors

Sensor Selection

VOC Compound Sensitivity Reference

Sensor selection was informed by a cross-referenced sensitivity matrix of MQ-series sensors against known mold VOC metabolites. The MQ-138 showed strong (S) responses across the broadest set of fungal-relevant compounds — alcohols, ketones, aldehydes, and aromatics. The MQ-3, while more narrowly tuned to ethanol, provides a useful orthogonal signal. Using both sensors in combination allows the algorithm to cross-validate readings and reduce false positives from single-compound interference.

CAD render of MycoSpec 3D-printed enclosure with mounting bracket

Mechanical

Enclosure & Mounting System

The enclosure was designed in CAD and 3D-printed in PLA. The modular clip-on form factor allows the device to mount directly onto standard HVAC duct flanges without tools. The top plate provides access ports for the two MQ sensors and DHT22, while the body houses the ESP32, SD card module, and power wiring. The "M" branding panel is removable for sensor servicing. Future iterations target injection-molded ABS with IP54 environmental sealing.

Skills Gained

What We Built Along the Way

⚙️

Hardware & Fabrication

Sensor selection and characterization, 3D-printed enclosure design, HVAC integration, PCB prototyping and soldering.

💻

Firmware & Software

ESP32 firmware development, real-time sensor data pipelines, threshold algorithm design, and web dashboard development.

🔬

Scientific Validation

Designing controlled lab experiments, working with live fungal cultures, interpreting sensor data against biological ground truth.

🗣️

Communication & Stakeholder Engagement

Translating technical findings for clinical audiences, collaborating with Duke Hospital stakeholders, and presenting to non-expert reviewers.

Reflection

What We'd Do Differently

Start testing earlier

Hardware-software integration and lab validation took longer than anticipated. Earlier testing cycles would have surfaced sensor calibration issues before they became time pressure.

Nail down the power architecture sooner

Voltage inconsistencies from USB power introduced noise late in development. Locking in a stable power supply earlier would have saved significant debugging time.

Test with more mold species from the start

Limiting v1 validation to Aspergillus was pragmatic but left open questions about generalizability. Multi-species testing in parallel would have strengthened our threshold confidence.

Document sensor behavior from day one

Baseline drift and warm-up behavior weren't systematically logged until mid-project. Earlier longitudinal data collection would have improved calibration and reproducibility.

Early-Warning Fungal Monitoring

Invisible threat,real consequences