MobilePTX: AI-based Analytical Method for Limited POC Ultrasound Imaging data

PAGE TITLE

Overview

 

PAGE SUMMARY

This technology is a novel artificial intelligence (AI) system designed for point-of-care ultrasound (POCUS) imaging, utilizing a sparse coding framework to enhance the extraction and interpretation of ultrasound features. This explainable AI methodology can help create diagnostic tools that are accurate, transparent, and practical for real-world healthcare settings. By focusing on clinically meaningful anatomical structures and motion patterns, the system can achieve high diagnostic performance with far fewer labeled training examples. The innovative technical architecture utilizes an expert-informed workflow focusing on pleural-line localization and sparse spatiotemporal feature encoding, which aligns with data-scarce settings and may improve interpretability over generic AI models. This framework has been successfully demonstrated for pneumothorax detection. Further development and commercialization could help address a critical, time-sensitive condition—pneumothorax detection at the point of care—relevant across emergency, trauma, ICU, military, and remote settings. In addition, this Drexel technology could be adapted to other ultrasound-based applications, offering a scalable platform for developing next-generation diagnostic support tools that run directly on portable and mobile devices.

 

 

ADVANTAGES

TITLE:Key Advantages

 

Improved diagnostic accuracy through sparse coding, which enables effective extraction of meaningful ultrasound features.

Improved interpretability of AI outputs, facilitating clinician trust and clinical decision support.

Real-time execution on mobile hardware platforms, promoting accessibility in diverse clinical contexts.

Robust model architecture combining state-of-the-art object detection (YOLOv4) with advanced video feature analysis.

Demonstrated superior performance in comparison to existing AI models tailored for lung ultrasound diagnostics.

 

 

Problem Solved

TITLE:Problems Solved

 

Addresses challenge of AI model opacity in medical imaging by providing interpretable outputs that clinicians can understand quickly

Facilitating rapid, on-site diagnostic decisions in critical settings through integration with portable ultrasound devices

Improves extraction of clinically meaningful features from noisy and variable ultrasound images, enabling more accurate diagnostic interpretation

Addresses the scarcity of annotated medical imaging data by using sparse coding techniques that learn effective representations from limited training examples

 

 

APPLICATIONS

TITLE: Market Applications

 

Point-of-care diagnosis of pulmonary conditions including pneumothorax using lung ultrasound imaging.

Screening and detection of COVID-19 respiratory manifestations through ultrasound analysis.

Extension to other ultrasound-based diagnostic applications where interpretability and accuracy are critical.

Deployment on mobile and portable devices such as smartphones or tablets for bedside or remote clinical use.

Use in medical training environments to assist clinicians in understanding AI-driven diagnostic decisions.

 

 

IP STATUS

Intellectual Property and Development Status

US Patent pending 19/124,912

 

 

 

PUBLICATIONS

Technical References

 

Pubinfo should be the citation for your publication. Publink is the full url linking to the publication online or a pdf.

MobilePTX: Sparse Coding for Pneumothorax Detection Given Limited Training Examples

 

 

Drexel News article

https://drexel.edu/news/archive/2025/March/Drexel-Researchers-Develop-New-DNA-Test

 

 

 

 

Commercialization Opportunities

Drexel is seeking an industry partner to license, develop and commercialize this technology.

 

----------------------------------------------

 

 Contact Information      

 

 

For more information, contact Dr. Robin Stears, Director of IP & Agreements, at rls457@drexel.edu or applied_innovation@drexel.edu.

 

 
Patent Information: