Phlebot - Autonomous Venipuncture Robot

Awarded Best System Performance Robot, Best Construction Quality, and Best Coolness Factor Robot

Venipuncture is the process of gaining access to veins for the purpose of blood sampling or therapy.

It is often difficult to locate and accurately insert the needle, even for train professionals leading to lengthy service times and increased discomfort for the patient.

Existing solutions for locating veins cost thousands of dollars and don't assist with the injection step. They also have difficulty with small veins, those with decreased elasticity, bruising, obesity and tattooed patients.

We are developing an autonomous solution to locate veins and insert needles for venipucture called

PhleBot.

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

Introduction 

Opening an intravenous line for some patients can be very challenging due to difficulty in visualizing the vein. Market solutions to this problem include infrared finders and NIR based sensors which reach prices upwards of $750.

 This project aims to create a low budget medical device that would be capable of opening an intravenous line to simplify the venipuncture process. The value of this project aims to provide a cheap alternative to the many current visualization tools out in the market currently, which in addition has the ability to successfully perform venipuncture and open an I.V. safely without a trained healthcare professional. 

The final design aims to be able to autonomously determine the vein and perform the injection without any external help. 

Further information about this can be found at this link below.

https://sites.google.com/andrew.cmu.edu/cmumechatronics2023teamd/home

Functional Architecture  

The process will begin with the user inserting their arm into the robot in order to create an alignment that we can then follow. An external operator would then tourniquet the arm to stimulate blood flow and the robot will then localize and center itself after locating the notch of the arm.The sensing and computer vision processing occurs here as sensors in the form of cameras would be used to locate the vein for puncturing. The actuation phase begins here as the robot navigates to the appropriate tool head and moves to the location of the vein. According to the desired output, the robot will anchor at this position and swap tool heads to complete each desired action such as disinfecting, injecting, removing, and bandaging the entry.

System Description  

• Mechanical

  • A carriage (Green and Grey) moves along the length of the arm as the toolhead (Blue) moves along the clamping arm (Yellow) to detect veins and insert the needle.  

• Electronics

  • The system runs on 12V to power the entire system including a Jetson NANO, linear rail, solenoids and pneumatics that operate at 100PSI. The camera records arm photos that then are labelled by the Jetson NANO. The resulting injection site is calculated and then traversed to using the solenoids and linear rail. Pneumatic are used to inject at a constant pressure every time to ensure no errors. 

• Perception

  • Using a trained neural network, given RGB-D images, the intent is to be able to identify possible vein locations to inject into with high accuracy and precision. Provided a vein label, the robot will use inverse kinematics to position and inject into the arm properly. 

My contributions were all focused on the solenoid mechanism, electronics and assisting in perception.

MECHANICAL SUBSYSTEM ELECTRONICS SUBSYSTEM PERCEPTION SUBSYSTEM

System Performance  

Conclusion

The Phlebot resulted in the successful development of an autonomous robot capable of finding and injecting into veins on its own. The robot's hardware and software demonstrated high performance and reliability as well. With my implementation of the electrical and help for perception, the robot was able to properly function as intended. Over a range of 50 tests, the robot was able to always identify the vein 94% of the time and inject into it.