Automatic differentiation between Veress needle events in laparoscopic access using proximally attached audio signal characterization

Abstract The access to the abdomen and the creation of a pneumoperitoneum is an initial and particularly critical step of minimally invasive laparoscopic procedures. Insertion instruments such as the Veress needle need to be introduced blindly into the abdominal cavity, which is associated with inadvertent visceral and vascular injuries. To ensure safe positioning of the instrument, information about the entry path advancement of the tip through the abdominal wall is needed. The main objective of this work is to demonstrate the capability to acquire information about intracorporeal tissuetool interactions of the Veress needle tip, utilizing acoustic emissions recorded at the extracorporeal end of the needle. In an experimental setup, a Veress needle was inserted in a multitissue- layer phantom with a defined insertion speed. Acoustic emissions were recorded with a MEMS microphone attached to the extracorporeal end of the needle. In addition, the counteraction forces during insertion of the needle were measured and a video of the experiment was recorded as reference. With this setup, an audio database of characteristical insertion events was generated. For the classification of characteristic audio events and detection of tissue-layer crossing, features were calculated in the time and frequency domain. Subsequently, a feature dimensionality reduction was performed. The distribution clustering of the audio database in the three-dimensional feature subspace allows a distinction between certain characteristic audio events. The preliminary results show the capability of this acoustic emission based method to detect events related to the insertion of a Veress needle, such as tissue-layer crossing.