AI for Ventilator Management
Budget / Salary₹1,500–12,500
TypeFreelance project
LocationRemote
Posted1 hour ago
Hi, I’m working on my PhD project on **AI-based detection of patient–ventilator asynchrony (PVA) from ventilator waveforms**, and I’m looking for a Python/AI developer.
The project will involve:
1. **Physics-based synthetic waveform generation** using lung-mechanics equations to generate realistic Paw, flow and volume signals, initially focusing on normal breathing and double triggering.
2. **Synthetic dataset creation** with waveform signals, images, labels and physiological parameters.
3. Addition of realistic **noise, artefacts and physiological variability**, followed by diffusion-based augmentation.
4. Development of AI models using **time-series and waveform images**, including Transformers, CNNs, Bi-LSTM and Swin Transformer.
5. **Multimodal fusion** of signal and image features for PVA classification.
6. **Synthetic-to-real domain adaptation** using DANN, CORAL and MMD.
7. **Explainable AI (XAI)** using Grad-CAM, Integrated Gradients, Score-CAM and Attention Rollout, with comparison against expert annotations.
8. **Temporal modelling and uncertainty estimation** to improve clinical reliability.
9. Eventually, development of a **simple web-based interface** for waveform upload, PVA prediction, probability, highlighted regions and uncertainty.
I will provide the **clinical definitions, physiological specifications, research literature and expert validation**.
I need the developer to handle the **Python mathematical modelling, synthetic data generation, deep-learning implementation, training, evaluation, XAI and deployment**.
**Immediate requirement:** develop a scientifically valid Python model that can generate and visualise **normal and double-triggering ventilator waveforms**, with adjustable physiological and ventilator parameters. We can then progressively build the complete AI pipeline.
The project will involve:
1. **Physics-based synthetic waveform generation** using lung-mechanics equations to generate realistic Paw, flow and volume signals, initially focusing on normal breathing and double triggering.
2. **Synthetic dataset creation** with waveform signals, images, labels and physiological parameters.
3. Addition of realistic **noise, artefacts and physiological variability**, followed by diffusion-based augmentation.
4. Development of AI models using **time-series and waveform images**, including Transformers, CNNs, Bi-LSTM and Swin Transformer.
5. **Multimodal fusion** of signal and image features for PVA classification.
6. **Synthetic-to-real domain adaptation** using DANN, CORAL and MMD.
7. **Explainable AI (XAI)** using Grad-CAM, Integrated Gradients, Score-CAM and Attention Rollout, with comparison against expert annotations.
8. **Temporal modelling and uncertainty estimation** to improve clinical reliability.
9. Eventually, development of a **simple web-based interface** for waveform upload, PVA prediction, probability, highlighted regions and uncertainty.
I will provide the **clinical definitions, physiological specifications, research literature and expert validation**.
I need the developer to handle the **Python mathematical modelling, synthetic data generation, deep-learning implementation, training, evaluation, XAI and deployment**.
**Immediate requirement:** develop a scientifically valid Python model that can generate and visualise **normal and double-triggering ventilator waveforms**, with adjustable physiological and ventilator parameters. We can then progressively build the complete AI pipeline.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.