Determining patient weight in the emergency department using optical sensors

In an emergency, every second counts: in the emergency department, doctors give patients medication to counteract the consequences of heart attacks and cerebral infarctions on the basis of their weight. In cooperation with the neurological head clinic of the University of Erlangen and Siemens Healthcare, the Libra3D project has set itself the task of supporting doctors in determining weight by means of a camera system.

Motivation

A stroke patient with a brain haemorrhage is admitted to the University Hospital in Erlangen. The patient is not responsive. The emergency doctor gives the attending doctors a brief description of the symptoms, as well as the medication already administered. For further treatment, doctors administer a dose adapted on the basis of the patient’s weight. When every second counts, they resort to a common method: they estimate the weight. If a patient is given an overdose or underdose as a result of this, doctors can only correct this error in the further course of treatment once the exact weight has been determined by weighing the patient after the fact.

Previous studies conducted by Erlangen University Hospital have shown that when medical staff estimate a patient’s weight, about one third of the time the patient receives a dose that deviates from the ideal dose by more than ten percent. This is where a sensor system comes into play to estimate the patient’s weight. In contrast to an estimate made by a doctor, the system determines the weight on an objective basis. Doctors can easily estimate the weight of patients, if they have the same approximate stature as the patient in question.

Sensor systems

The Libra3D sensor system is based primarily on a Microsoft Kinect camera. In the clinical treatment room, this camera is hidden behind the ceiling. Various 3D sensors are suitable for compiling a three-dimensional image of the patient, but the Kinect camera stands out due to its affordable price. Many ideas investigated in research papers cannot be applied in practice. They either involve expensive sensor networks or require the patient to stand up or be moved. The appeal of this particular project is that we can obtain stable results using low-cost consumer electronics. Libra3D estimates the patient’s mass by measuring the surface area and modelling the human body. The density of the patient can then be inferred using anthropometric characteristics.

Algorithmics

One of the most difficult tasks is locating the patient and filtering out the hospital beds and objects that do not need to be included in the weight estimation. The weight of the patient in the bed is issued within just a few moments. 

Results

The initial results are now in: the system, developed by Nuremberg Tech, is accurate in its weight estimations to within 10% in almost 80% of cases; doctors, on the other hand, are successful in only 70 percent of cases. This preliminary result shows the potential of the system. This year, the system will be extended to include a thermal imaging camera. This is already integrated in the ceiling of the treatment room, but still needs to be calibrated to the Kinect camera. The researchers hope that the additional information provided by the thermal imaging camera will enable them to determine weight even more accurately.

Publications

• Christian Pfitzner, Stefan May, and Andreas Nüchter: Evaluation of Features from RGB-D Data for Human Body Weight Estimation, In Proceedings of the 20th World Congress of the International Federation of Automatic Control, 9-14 July 2017, Toulouse, France, 2017.
  
• Christian Pfitzner, Stefan May, and Andreas Nüchter: Neural Network-based Visual Body Weight Estimation for Drug Dosage Finding. In Proceedings of the SPIE Medical Imaging Conference on Image Processing, San Diego, CA, USA, March 2016 (accepted)
• Christian Pfitzner, Stefan May, Christian Merkl Device and method for optically detecting a person’s weight, German patent. Date of filing: 29 February 2016.
• Christian Pfitzner, Stefan May, Christian Merkl, Lorenz Breuer, Martin Köhrmann, Joel Braun, Franz Dirauf, and Andreas Nüchter. Libra3D: Body Weight Estimation for Emergency Patients in Clinical Environment with a 3D Structured Light Sensor, in Proceedings of the IEEE International Conference Robotics and Automation (ICRA '15), Seattle, WA, USA, May 2015.
• Christian Pfitzner. Robotic Vision in Medical Applications: Visual Weight Estimation for Emergency Patients. Workshop Proposal to the Ph.D. Forum at theInternational Conference Robotics and Automation (ICRA '15), Seattle, WA, USA, May 2015. 

Project participants from Nuremberg Tech

• Project management: Prof. Stefan May
• Academic/research staff Christian Pfitzner, Rainer Koch
• Head of Research: Christian Merkl, Eduard Roth
• Final theses: Christian Merkl
• Student project groups:

1. Andreas Geiger, Eduard Roth, Kristina Sawinsky, Miriam Schmidmeier, Venelin Venelinov, Laura Werder
2. Johann Delchmann, Chris Espig, Marion Kaiser, Deniz Neufeld 

Project partners

Contact at Siemens: 

• Dr Joel Braun
• Franz Dirauf

Contact at Erlangen University Hospital: 

• Dr Martin Köhrmann, Senior Consultant and visiting lecturer
• Dr Lorenz Breuer, Stroke Unit ward doctor

Contact at Julius Maximilian University of Würzburg:

Prof. Andreas Nüchter, Chair of Computer Science VII

Funding

The Libra3D project is funded by the German Federal Ministry of Education and Research (BMBF). Reference number: 03FH040PX3

Funding period: 01 October 2013 to 30 September 2016

Contacts

Christian Pfitzner