Trenažer resuscitace

Abstract

The aim of this thesis is to verify whether accelerometer measurements acquired during cardiopulmonary resuscitation can be used to derive parameters that provide meaningful feedback on compression quality, especially compression rate and assignment to qualitative classes. The thesis deliberately works with a minimal hardware configuration corresponding to commonly available devices equipped with a single accelerometer and examines whether such a solution can be implemented within a web application. The opening chapter briefly summarizes commercial CPR simulators and the types of feedback they provide, points out limitations that motivate the search for a more open and accessible alternative, and at the same time defines the need to evaluate accelerometer data within a clearly specified mechanical configuration.

The following chapters focus on the design and implementation of a measurement chain with an accelerometer worn on the arm. First, the feasibility of acquiring raw data from the eZ430-Chronos watch and displaying them in a web application was verified. Subsequently, a custom prototype based on a microcontroller with an acceleration sensor and wireless transmission via BLE was developed. The processing chain includes calibration, band-pass filtering, compression cycle detection, and calculation of metrics used for classification and rate assessment.

The proposed procedure was from the outset based on the assumption of a clearly defined mechanical configuration and was subsequently validated using the selected CPR simulator and in the web application environment. Individual compression cycles were automatically delineated from the accelerometer recordings and pulse-shape parameters were calculated for each cycle. Analysis of their distributions confirmed that a combination of amplitude and area under the curve makes it possible to distinguish shallow, target-range, and deep compressions and to define threshold values for classification. These classes are not treated as absolute physical ground truth in millimeters, but as operationally defined training categories derived from the feedback of the simulator in the given configuration. Experimental validation also confirmed that this classification principle is not universally transferable without regard to the mechanical configuration of the system, but requires clearly defined conditions and a corresponding calibration dataset. An independent validation set acquired on the same rigid surface further showed that the derived thresholds remain reproducible within the same methodology and mechanical setup, rather than constituting absolute physical truth. The same principle was implemented in the web application environment and adapted for real-time evaluation.

The result is a hardware and software prototype that shows that even a single simple accelerometer can provide practically useful feedback for compression training. The main contribution of the thesis is the design and experimental verification of a classification algorithm together with a definition of the conditions under which it remains valid. The prototype links commonly available wearable electronics with a web application and provides a technical foundation for further development, towards broader compatibility with commercial devices, the design of a user-friendly interface, and the gradual integration of educational, motivational, and regulatory requirements into a comprehensive training tool.