Skip to content

Houston Methodist

Deep learning technology

Client Background
Challenges & Objectives
Solution
Client Background

Houston Methodist is a leading US healthcare system and academic medical center.  They are dedicated to providing exceptional patient care and prioritize advanced technology, maintaining high standards of quality and safety.  They operate a main site in the Texas Medical Center and six community hospitals serving the Greater Houston area. 

Challenges & Objectives

Houston Methodist surgeons needed to quickly and accurately identify the specific type of cardiovascular implant (e.g., a heart valve) in a patient before surgery, especially in emergencies.  Relying on visual inspection of medical images (like X-rays or CT scans) was unreliable and time-consuming.  This presented a significant challenge, as misidentification or delays in identification could have serious consequences for patient safety and surgical planning.  They needed a faster, more accurate, and more reliable way to identify these implants.

Solution

To improve the speed and accuracy of cardiovascular implant identification, Softway developed HMIAP, a deep learning system that analyzes medical images, automatically detects and classifies implants, and delivers results via a user-friendly web application, streamlining surgical planning for Houston Methodist.

Our approach

01

Data Preparation and Model Training: We developed a custom tool to convert DICOM images to JPEG and facilitate efficient labeling of implants.  Using this labeled dataset, we trained a deep learning model (Faster R-CNN) to accurately detect and classify different cardiovascular implants.

02

Cloud-Based Application Development: We built a web application using Flask and deployed it on AWS, leveraging S3 for image storage and MySQL RDS for the database.  This cloud-based approach ensured scalability, reliability, and accessibility for Houston Methodist clinicians.

03

Seamless Workflow Integration:  The HMIAP application was designed for easy integration into Houston Methodist’s existing clinical workflows.  A user-friendly interface and robust search functionality simplified the process of accessing and utilizing implant information.

04

Agile Development and Collaboration: We employed an agile methodology with regular reviews and close collaboration with Houston Methodist. This iterative approach allowed for flexibility and ensured the solution met their specific needs.

houston-methodist-scan

AI-powered healthcare solutions, improving patient care.

Application

Cardiovascular Implant Detection

HMIAP (Houston Methodist - Cardiovascular Implant Detection) is a web application that uses machine learning to automatically detect and classify up to ten types of cardiovascular implants (like heart valves) in medical images. 

Trained on a large dataset of labeled medical images, the AI model powers HMIAP's ability to quickly and accurately identify implants, streamlining surgical planning and improving patient care. The application features a medical device image library, UI image upload capabilities, database updates, image categorization logic, and secure user access.

Screenshot 2025-04-17 at 11.12.13 AM

Project Gallery


The results

Results-graph

01

Faster Implant Identification: HMIAP speeds up implant identification, saving time for surgeons.

02

Improved Accuracy and Efficiency: HMIAP's AI model increases accuracy and efficiency by reducing the time and effort required for implant identification.

03

Scalable and Accessible Solution: The cloud-based architecture of HMIAP ensures scalability to handle large volumes of medical image data.

04

Foundation for Future Development: The HMIAP proof-of-concept provides a solid foundation for further development and expansion of AI-powered implant detection capabilities.

Key insights & takeaways

Deep Learning

This project demonstrated the effectiveness of applying targeted deep learning models to specific, well-defined problems within healthcare. 

By focusing on cardiovascular implant detection, we were able to achieve a high degree of accuracy and efficiency, highlighting the value of specialized AI solutions in the medical field. 

Data Quality

The success of HMIAP underscored the critical role of high-quality, labeled training data in deep learning.  Our close collaboration with Houston Methodist clinicians, combined with a streamlined data preparation process, was essential for developing an effective model.  

Ready to grow? Let's talk.

We are excited to help with your technology, communication, and culture change needs. Let’s talk shop as soon as you’re ready.

AdobeStock_707763291
BeyondHumain_hero-3

AI is changing everything.
Are you ready?

AI is here. And the businesses that don’t adapt? They’ll be left behind. Beyond HumAIn is the podcast designed to help leaders like you navigate the AI revolution without losing sight of what matters most—your people.

Our way of working is Wall Street Journal and USA TODAY bestseller.

Written by four of the leaders at Softway®, Love as a Business Strategy® offers a people-first framework for achieving any business outcome by creating an environment of high performance.

Get ready to ditch the status quo and embrace humanity.

Learn more
LAABS_Floating-Stack-2