RESEARCH STATEMENT and OVERVIEW
Automation of robotic surgery can improve the performance of surgeons and the quality of the life of patients. However, there are challenges in the automation of robotic surgeries that must be resolved. Surgical tools are widely used in healthcare and medicine for various purposes.
Problem Statement: Automating learnings in robot-assisted surgery for surgeon’s training and assistance
Motivation: The research problem's outcome can be automated to give scores on training or assistance during the surgery. The surgical robots may be thought of as information-driven surgical tools that enable human surgeons to treat individual patients with greater safety, improved efficacy, and reduced morbidity than would otherwise be possible.
The research topic involves understanding three main areas: medical surgery, robotic path planning, and machine learning (deep learning).
In Robot-Assisted Surgery or MIS, deployment is based on an indirect manipulation of specialized surgical tools and operations through an indirect visualization of the surgical field captured by an endoscope (camera). The effectiveness of existing medical procedures can potentially be improved using robot-assisted tool steering. A typical surgical tool requires searching through all possibilities in a surgical environment. Investigation of path planning of a dextrous tool renders both computationally expensive and theoretically challenging problems to address.
We need algorithms to avoid collision of the tool with critical organs and reduce the search space while applying optimal path planning algorithms. Procedural complexity is added during the planning of a surgical trajectory by considering additional non-linear trajectories and different risk structures.
Also, the quality of surgery and the surgeon’s capability can be improved using intraoperative assistance systems. These systems aid in recognizing surgical actions during any intervention facilitate the interaction with technical additives and provide support depending on the surgery context and surgical actions.
The decoupling between the planning stage and the execution stage relies on the underlying assumption that the surgical tool will execute the motion strategy computed by the planning stage successfully. Machine learning methods can be used to address both motion planning and motion execution problems.
The experience of doctors is fairly needed to apply ML methods. For this, the patient-specific information needs to be combined with doctors' statistical information about human anatomy, physiology, and disease to produce a comprehensive computer representation of the patient, which can then be used to produce an optimized interventional plan.
01
SURGICAL ROBOTICS
Robotic surgery, or robot-assisted surgery, allows doctors to perform many types of complex procedures with more precision, flexibility and control than is possible with conventional techniques.
02
MOTION PLANNING
The motion planning problem consists of finding a valid path for an object from a start configuration to a goal configuration.
03
LEARNING ALGORITHMS
Learning algorithms include machine learning concepts that are useful to automate any surgical procedure while considering uncertainties.
Deep learning methods are used for real-time planning which includes formulating an intraoperative assistance system.
Pre-operative planning:
Path planning of a tool in Robot-assisted surgery can provide assistance to the surgeons by giving solutions for faster and safe tool movements during the surgery.
Pre-operative images are used to segment obstacles in the surgical environment. Interval Method to plan path for a surgical tool.
Projected Environment
Shubhangi Nema and Leena Vachhani, "Safe and Fast Path Planner for Minimally Invasive Surgery", IEEE/RSJ International Conference on Intelligent Robots and Systems 2021 (IROS'2021) Sept, 2021, Czech Republic.
Real-Time planning:
Intraoperative assistance system
-
Applying machine learning methods for surgical action analysis.
-
For assisting surgeons in the field of robot-assisted surgery and training the newly budding surgeons.
-
Raw laparoscopic dataset with single or multiple surgical robotic and rigid instruments is used for the analysis.
Dataset: Robotic Rigid
Endovis sub-challenge: Instrument segmentation and tracking. https: //endovissub-instrument.grand-challenge.org/
-
Interaction with expert surgeons associated with BETIC Lab IITB.
-
Interaction with newly budding surgeons in remote areas.
My work decomposes the higher dimensional problem in the lower dimension by considering the constant value of the variable used for reducing the dimension. The current work renders a safe and fast planner in lower dimensions. In order to use the lower dimensional results in higher dimensional path planning, working towards the direction of developing an intraoperative assistance system would investigate parallel processing of lower dimensional results.