Achieving negative margins is important in surgical cancer treatment; however, there is a trade-off between achieving a negative margin and preserving healthy tissue. Mass spectrometry -based tools are becoming available that enable classification of in vivo tissue in real time to improve margin definition. In tandem, robotic surgical platforms are being developed that enable precise and repeatable motion control.
The objective of this work was to demonstrate a compact soft robotic platform with autonomous localisation and guidance capabilities based on mass spectrometry classification data.
A hybrid soft robot was used to guide an optical fibre and collection tube, for laser ablation of tissue and subsequent collection of the aerosol produced. A 10600 nm CO2 laser (Omniguide, USA) was used for ablation, while a mass spectrometer (Xevo G2-S QToF, Waters Corporation, USA) coupled with a Rapid Evaporative Ionization Mass Spectrometry (REIMS) source was used for molecular analysis of the aerosol. The mass spectral results are processed into tissue models using multivariate statistical analysis (Principal Component Analysis – Linear Discriminant Analysis) which models are used for quasi-real (>100 ms) time tissue classification. The hybrid soft robot is a combination of two robotic mechanisms, capable of gross and fine positioning respectively, and three cameras for both monocular tool tracking and stereoscopic tissue surface reconstruction.
The gross positioning system consisted of a soft robotic parallel mechanism with a large workspace, which guides the base of a thermally actuated fibre robot, the fine positioning system. Tracking of the gross positioning robot was achieved using computer vision; specifically, a bespoke marker on the robot shaft was tracked using graph-based deep learning methods that enabled sub-millimetre accuracy. As such, the pose of the robot tip was tracked while mass spectrometry data was collected to produce 3D point clouds.
To autonomously steer the robot, the direction of motion of the soft robotic platform was altered depending on the classification data from the mass spectrometry instrument. Algorithms were designed such that the robot first identified and then followed the boundaries between healthy and unhealthy tissue.
The autonomous guidance and localisation capabilities were demonstrated in a series of experiments on mouse skin tissue containing cancerous regions.
The hybrid robot platform successfully produced mass spectrometry point clouds of several ex vivo mouse skin tissue samples, each containing multiple cancerous lesions. Based on the classification mappings, the robot was able to identify boundaries of the cancerous regions, generate trajectories in 3D space, then execute these trajectories while ablating the tissue.
The initial tissue models consisted of thin sections (mouse brain, pork liver) mounted on glass slides and were analysed by the robotic platform coupled with a REIMS-equipped mass spectrometer. Based on the molecular profiles (which mainly consist of metabolites, lipids and other small molecules), the structures of the hippocampal region were successfully identified.
One of the main advantages of the robot is its small size and flexibility. The robotic platform can be utilised to analyse normally hard-to-reach areas such as the head and neck area, or internal body cavities such as colorectal areas, where it will have the capability to autonomously perform molecular profile-based surgical interventions. Live animal experiments were performed in pigs to demonstrate the possibility of applying the system in an in vivo environment.
Due to the molecular pathology-based decision-making of the system, the robot has great potential for successful resection of diseased tissues where the preservation of function is important (such as skin cancer on the face, or head and neck area). The setup was successfully used to analyse the head and neck area of a human cadaver to demonstrate the capability to sample deep tissue areas without causing significant disruption to the surrounding tissues.
Currently ex vivo tissue collection and analysis is being prepared for skin cancer and cervical cancer samples. Ethical approval is being sought for the in vivo demonstration of the robotic system, for skin cancer and cervical cancer cases. The robot was able to autonomously ablate cancerous tissue while preserving healthy tissue.
This work demonstrates the feasibility of autonomous surgical interventions using mass spectrometry data to guide a robotic tool. The soft robotic device used has multiple potential applications, including endoluminal minimally invasive interventions, for example, which will be the focus of this project in the future. Furthermore, the robot hardware can be manufactured rapidly and economically, increasing the accessibility of this technology.