Current PhD projects


PhD Candidate

Hirenkumar Nakawala

Advisors

Elena De Momi, Prof. Giancarlo Ferrigno

Title

Knowledge-driven machine-learning based context-aware framework in surgery

Description

 Surgical Applications: The Ph.D. thesis is focused on automatic analysis of workflows of two surgical procedures:

  1. Thoracentesis: It is performed to withdraw the fluid from the chest cavity in my disease conditions such as kidney failures. It is simple procedure, although it is associated with very high complications.
  2. Robot-Assisted Partial Nephrectomy (RAPN): It is performed to remove kidney tumor from kidney. The workflow of this procedure is very length and it takes very long time for trainees to understand this procedure. It is also associated with many complications.

Purpose: The purpose of Ph.D. thesis to analyze the surgical workflows of above mentioned procedure automatically and give the contextual information i.e. information on the surgical activity in progress, at a specific instance of time. The aim is to create a context-aware system that could help in surgeon’s training.

Knowledge-driven context-aware system: A knowledge-driven context-aware system was developed, which analyze the surgical workflow of Thoracentesis procedure automatically. It is composed of three modules:

    • Knowledge module, where we implemented procedural knowledge on Thoracentesis in the form of a knowledge graph i.e. ontology and surgical workflow management through logical rules;
    • Computer vision module, where we implemented a) segmentation and tracking algorithms to track the surgical tools in surgeon’s hands and on the surgical table b) fiducial markers to recognize surgical tools;
    • Data monitoring and Graphical User Interface (GUI), “MRS (Medical Robotics section of NeuroEngineering and medicAl Robotics Laboratory) context-aware system”, where the system interacts with the clinicians and provides contextual information, e.g. instruments used in a particular phase of the surgical workflow.

Deep learning for surgical steps and other surgical context recognition: Deep learning algorithms, called CNN (Convolutional Neural Networks) and LSTM (Long Short-Term Memory) cells were used to recognize surgical steps of Robot-Assisted Partial Nephrectomy procedure, where we created the video datasets based on video annotations done at European Institute of Oncology, Milan. We also recognized several workflow context i.e. surgical phases, surgical tools and surgical anatomy based on the ontology after we predicted the results from deep-learning algorithms.

PhD Period

XXX cycle (November 2014 – November 2017)


PhD Candidate

Jacopo Buzzi

Advisors

Elena De Momi, Prof. Giancarlo Ferrigno

Title

Surgical gesture analysis in robotic teleoperation

Description

Jacopo_phd

The PhD project focuses on the analysis of the surgical movement during teleoperation. In the recent years, more and more surgical procedures have been enhanced by the use of teleoperated surgical robots. In this scenario, the surgeon interacts with a master controller, usually a multi degrees of freedom robot, to control an actuation unit (the slave) that directly manipulates the patient. The master controllers are usually designed to be able to reproduce all the movements the surgeon would do in free hand, nevertheless a sometimes extensive training phase is required to acquire familiarity with the teleoperation. The aim of this study is to understand if and how much the use of a master controller modifies the kinematic and musculoskeletal strategies adopted by the user, also trying to compare different controllers; in order to do so, the kinematic position of the user’s arm, and the muscular activation during specific tasks are studied. Thanks to this knowledge, it would be possible to potentially design better controllers, to create more specific training tasks and environment and to produce quantitative indexes to evaluate the surgeon’s ability in teleoperation.

Training under the effect of force fields in virtual teleoperated robotic surgery
In this work we introduce a virtual reality and haptic test bench to evaluate the effects of two novel force fields on the learning process. Multiple trajectory following tasks were performed by three groups: with visual feedback only, with a thresholded divergent force field (Limit-Push), and a mixed divergent-convergent force field (Limit-Trench).

PhD Period

XXX cycle (November 2014 – November 2017)


PhD Candidate

Sara Moccia

Advisors

Elena De Momi, Leonardo Mattos (IIT)

Collaboration

Istituto Italiano di Tecnologia

Title

Tissue classification in optical imaging for surgical data science applications

Description

 Surgical data science is evolving into a research field that aims to observe everything occurring within and around the treatment process to provide situation-aware data-driven assistance. In the context of endoscopic video analysis, the accurate classification of organs in the field of view of the camera proffers a technical challenge [Maier-Hein et al., Nature BME, 2017].

In this context, the objectives of this PhD project are:
1) Investigating, for the first time in the literature, the use of machine learning algorithms for (i) early-stage diagnosis in laryngoscopic videos in narrow-band imaging (Video 1) and (ii) liver transplantation assessment in RGB images acquired in the OR

 

2) Developing an uncertainty-aware framework for abdominal tissue classification with the first in-vivo multispectral imaging laparoscope (Video 1)

3) Presenting an innovative algorithm for automatic informative frame selection in endoscopic videos (Fig. 1)

4) Exploiting deep learning for fast and robust vessel segmentation (Fig. 2) for safety enhancement during neurosurgical procedure using an active handheld robotic tool

Fig. 2: Firts row: phantom vessels; second row: ground-truth segmentation; third row: obtained segmentation.

PhD Period

XXX cycle (May 2015 – May 2018)


PhD Candidate

Davide Scorza

Advisors

Elena De Momi, Luis Kabongo (Vicomtech)

Collaboration

Vicomtech-IK4 (Donostia/San Sebastian, Gipuzkoa – Spain)

Title

Automatic planning for intracranial electrode placement in SEEG

Description

Clinical Application and Current Limitations: Stereo-ElectrodeEncephaloGraphy (SEEG) is a minimally invasive technique which allows the exploration of brain’s activity in patients affected by focal epilepsy, helping the identification of the epileptogenic zone. The planning phase for electrode trajectories can be cumbersome and very time consuming, and actual commercial software do not provide any quantitative information regarding trajectory risks.

Purpose: The PhD thesis is focused on the development of an automated planner for intra-cranial electrodes insertion, based on the criteria that guide the intervention. The final aim is to assist surgeons during the electrodes’ trajectories planning phase and improve the results in terms of precision, safety and time consume.

Automated Planner: An automated planner has been developed as a 3D Slicer application to optimize the trajectories starting from an approximated input of the surgeon. The whole extension is composed by different modules

  • the electrode representation module is a simple module which allows the surgeon to represent the electrodes and add new models by defining their geometrical properties.
  • the image processing module to process the patient images before starting the optimization
  • the SEEG planning module, which allows electrode initialization and optimization by maximizing the distance from vessels and minimizing the insertion angle. In addition, the use of an atlas allows to keep the targeted brain zones during the exploration of feasible optimum trajectories. Finally, inter-electrode conflicts are checked to avoid collisions in the proposed trajectories.
  • the SEEG check trajectory module, that offers advanced tools to the surgeons to analyse the feasibility and viability of the proposed trajectories. Maximum Intensity Projection (MIP) of the vessels can be generated on the probe-eye-view plane, and an automatic segmentation method based on Gaussian Mixture Model and Markov Random Field assist the surgeon to understand the presence of vessel along the trajectory.

Figure 1 – on the left, the planning requirements that have been identified for SEEG intra-cranial electrode placement. In the centre The SEEG automated planner extension architecture that has been developed in 3D Slicer. On the right, three main constraints that are considered to guarantee the patient safety and the planning effectiveness (anatomy-driven exploration, vessel avoidance, perpendicular insertion angle).

On going work: the actual work relies on the extension of the algorithm proposed in the SEEG check trajectory module to 3D vascular segmentation. Different variations of the algorithm are being tested, with the aim to provide a reliable vascular segmentation that can be used for our planning purposes.

Figure 2 – preliminary results of the 3D vessels model obtained by an automated segmentation algorithm based on Gaussian mixture model and Markov random fields.

Finally, to increase the cognitive capacities of our system, we are collecting a large amount of data regarding successfully past cases and building a database that will be used to extract information to be applied to new patients.

Figure 3 – Schematic representation of a cognitive decision support system for SEEG trajectory planning. The surgeon will provide basic inputs that will be matched with past cases in the database, and finally adapted to the anatomy of the current patient.

PhD Period

XXXI cycle (Nov 2015 – Nov 2019)


PhD Candidate

Hang Su

Advisors

Prof. Giancarlo FerrignoElena De Momi

Title

Safety-enhanced Human-Robot Interaction Control of Redundant Robot for Teleoperated Minimally Invasive Surgery

Description

The PhD Thesis focuses on developing the Safety-enhanced control strategy for Human-robot Interaction in teleoperated Minimally Invasive Surgery (MIS). Although in the past decade, a lot of surgical systems has been developed to achieve greater accuracy and safety. In the future operating room, the robot should be more intelligent, flexible and human-friendly,  not only to let the surgeon focus on the main surgical task, but also be able to work friendly  with the nurse in the same workspace.  Let the work of nurses and surgeons easier and enjoyable.  The application scenario could be to ease the work for nurses sharing the same workspace. When there is a collision between nurse and robot arm in Fig. 1. The nurse could move the robot arm with hand without influence of the surgical tasks. As it is shown in Fig, 2, the user 2 could focus on the teleoperated MIS, and the user 1 could move the surgical robot arm by hand, simultaneously. Nonlinear adaptive controller is introduced to solve the uncertain disturbance from human-robot interaction, enhancing the precision and reliability of Surgical Robot systems. Prediction model and machine learning will also be applied to enhance the online performance of robot tele-operation so that the surgical robot can act intelligently in cooperation with nurse. Finally, experimental validation will be conducted to validate the performance of the developed real surgical robot system.

PhD Period

XXXI cycle (Nov 2015 – Nov 2018)


PhD Candidate

Alberto Favaro

Advisors

Elena De MomiGiancarlo Ferrigno; Feridnadno Rodriguez y Baena (Imperial College of London)

 Collaboration Imperial College London, U.K.
 Title Path planning for non-linear, non-holonomic systems

Description

The PhD thesis regards combinatorial and sampling-based methods for path planning intended for being used in the context of robot non-holonomicity. Further than this, the systems under study also show a non-linear characteristic in their three-dimensions steering capability, which represents a further constraint in the definition of feasible paths. The work uses the EDEN2020 steerable catheter as a case of study and defines surgical trajectories for brain tumor treatment in the context of Convention Enhanced Delivery.

PhD Period

XXXI cycle (November 2015 – November 2018).


PhD Candidate

Marco Vidotto

Advisors

Elena De Momi; Daniele Dini (Imperial college, London)

 Title Convection Enhanced Delivery model for brain tumour treatment

Description

Marcothesis

Convection-Enhanced Delivery is a therapeutic treatment consisting in the injection under positive pressure of drugs directly in the brain tumorous zone. It has recently emerged as a promising delivery method because it allows to circumvent the blood-brain barrier, which is a major impediment in systemic delivery of macro-molecular therapeutic agents, by directly injecting into the parenchymal space. The treatment success depends on the CED protocol, e.g., infusion site, infusate concentration and infusate rate. Computational and analytical models of infusion may be used to predict drug distribution and optimize CED protocols. The PhD project focuses on the development of a new CFD (Computational Fluid Dynamics) model which exploits two types of information which are related to two different scales: a micro (µm) and a macro (mm) scale. Imaging techniques as the FIB-SEM (Focused Ion Beam – Scanning Electron Microscope) allow to access the geometrical organization of groups of single axons which is directly related to fundamental hydraulic parameters such as conductivity and tortuosity. On the other hand, on a macro scale, DTI (Diffusion Tensor imaging) provides essential information both on diffusivity and white matter orientation. This study aims to funnel micro and macro scale parameters in a unique and comprehensive model.

PhD Period

XXXII cycle (November 2016 – November 2019).

PhD Candidate

Sara El Hadji

Advisors

Elena De Momi; Giuseppe Baselli

Collaborations Dr.Francesco Cardinale, Epilepsy neurosurgeon at the Ospedale Ca’ Granda, Niguarda, in Milan; Medtronic, Littleton, Boston, USA
 Title ART 3.5D – a novel algorithm, to label arteries and veins from 3D angiography

Description

Several neurosurgical procedures, such as Artero Venous Malformations (AVMs), aneurysm embolizations and StereoElectroEncephaloGraphy (SEEG) require accurate reconstruction of the cerebral vascular tree, as well as the classification of arteries and veins, to increase the safety of the intervention. This project aims at developing a novel approach attempting to recover the contrast dynamic information from standard Contrast Enhanced Cone Beam Computed Tomography (CE-CBCT) scans. The algorithm proposed by our team is called ART 3.5 D. It is a novel algorithm based on the post- processing of both the angiogram and the raw data of a standard Digital Subtraction Angiography from a CBCT (DSA- CBCT) allowing arteries and veins segmentation and labeling.

PhD Period

XXXII cycle (February 2016 – F ebruary 2019).



PhD Candidate

Marta Lorenzini

Advisors

Elena De Momi, Dr. Arash Ajoudani

Collaborations

Italian Institute of Technology

Title

Improvement and evaluation of human ergonomics and workplace safety in human-robot collaboration

Description

The introduction of robots in human environment has countless benefits. Robots can reduce fatigue and stress as well as prevent injuries by assisting humans in industrial scenarios or in daily activities. This PhD project focuses on the development of assistive robotic strategies to guide and help humans to perform their tasks in the most comfortable and ergonomic body configurations to avoid pain and consequent injuries. To achieve this objective, robots must monitor human states in real-time and the profile of the interaction forces they exchange with each other and the environment. On the basis of this dynamic data, which can be collected using suitable sensory systems, it is possible to estimate the physical loading of the humans in terms of joints overloading. From this point, two complementary directions will  be explored: implementing feedback modalities to alert humans about their physical loading (modalities which range from simple screens showing which joints are too much loaded, to mechanisms that prevent improper body configurations); and secondly, developing control methods to make the robots anticipate and then guide the movements of the humans during the task to minimize the overloading joint torques. To this respect, thresholds to categorize joints overloading and methods to evaluate human ergonomics in the specific domain of human-robot collaboration will be developed.

PhD Period

XXXII cycle (May 2017 – May 2020).


PhD Candidate

Maria Lazzaroni

Advisors

Elena De Momi, Jesus Ortiz

Collaborations

Italian Institute of Technology

Title

Human exoskeleton system

Description

An exoskeleton is a wearable device that generates appropriate assistive forces on its user, both for medical and industrial applications. In industrial manufacturing processes, various manual work tasks are still difficult to automate due to their complexity. These  activities can significantly load workers lumber spine, resulting in musculoskeletal disorders. The aim of this project is to develop a full-body powered exoskeleton to assist human effort performing manual handling tasks in industrial environments. In order to develop flexible systems, adaptable for different applications, a modular approach will be adopted. Modular exoskeletons for upper limbs, lower limbs and full-body will be designed, starting from the trunk exoskeleton developed in the European project Robo-Mate. The design of these advanced exoskeletons has to take into account several issues: sensors, actuators, control systems and exoskeleton structure itself. Moreover, the human-exoskeleton interface should be studied.

In this context, my contribution will focus on human body surface analysis in order to enhance user’s comfort and guarantee a good coupling between the user and the device. Another important aspect that I will consider is the forces distribution on the human body surface. To evaluate forces distribution along with the effectiveness of the device, appropriate tests, performance evaluations and simulations will be regularly conducted while maintaining contact with users in order to answer their specific requirements and expectations.

PhD Period

XXXII cycle (May 2017 – May 2020).


PhD Candidate

Andrea Cimolato

Advisors

Elena De Momi, Leonardo De MattosMatteo Laffranchi

Collaborations

Italian Institute of Technology

Title

EMG-driven control in lower limb prosthesis

Description

Background: Neurorobotics leads human-technology interaction in the field of rehabilitation and prosthesis devices. Recent technology innovations made possible to design the modern generation of smart prostheses able to behave accordingly to different environmental interaction and user locomotion recognition. Thanks to hierarchical controllers (Fig 1), semi-active and powered prosthesis can change their joint impedance adapting to user necessities, avoiding undesired energy expenditures. In spite of this, lower limb microprocessor controlled prostheses stands far behind from upper limb solutions, those are able to decode volitional motion intent through the employment of Electromyographic (EMG) signals.

Purpose: The objective of this PhD project is to design a novel control approach for lower limb neuroprosthesis based on EMG signals and other mechanical variables such joint angle position and velocities, linear accelerations and load. The designed control should permit the user to both voluntarily and automatically control the prosthesis depending on the incoming sensor signals.

On Going Work:

1) The first intent of this study will be to exploit the great potentialities resilient in musculoskeletal models for the design of novel EMG-driven control. The use of musculoskeletal model results particular handy in situation where biological internal variables are not directly accessible such as muscle length and activation (Fig 2). In this way, it is possible to compare the model results with the EMG measurements in order to derive a particular correlation that will be employed for the prosthesis control.

2) Second perspective of this is to study and to understand the possible employment of muscle synergies in the design of a neuro-driven Human Machine Interface. Further analysis will investigate how these synergies decay or modify with the amputation of the limb in final users.

PhD Period

XXXII cycle (May 2017 – May 2020).


This site uses cookies to enhance your experience. By continuing to the site you accept their use. More info in our cookies policy.     ACCEPT
503 Service Unavailable

Service Unavailable

The server is temporarily unable to service your request due to maintenance downtime or capacity problems. Please try again later.