Meeting Minutes

back

July, 25 2007

Long-term plan

Research Goal 1: Workflow-supported Ultrasound Diagnosis

• Context-specific support
  • Detect required Ultrasound (US) workflow for each form of diagnosis
  • Guided support of US diagnostician
• Individualized support
  • Recognition of US diagnosis preferences for each surgeon
    • For each US scan, perform a 3D scan of a broad area of interest
    • Extrapolate preferred views of the surgeon from 3D (research issue! Christian Wachinger, Oliver Kutter)
    • Tele-analysis: present points/images/viewpoints of interest customized to each diagnostician's preferences
  • Expert diagnostician:
    • Knowledge and skill transfer from expert diagnosticians to less-skilled practitioners by analysing and copying the diagnosis workflow of the expert
    • Analyze why certain diagnosticians are better with US than others
    • Eye-gaze (see work by Imperial College)
• Funding / Partner: Canon
• Canon research goal:
  • Automatic diagnosis system: recover different paths of diagnosis (also related to tele-analysis issue)
  • Using data-mining / machine learning: use diagnosis history to automatically assess 3D US volume
  • Let the system batch-process hundreds of datasets and diagnoses to identify suspicious patients à further assessment by a real physician

Research Goal 2: General Workflow Project

• Adaptive GUI
  • Future GUI of surgeries: integrated surgery environment with all necessary equipment, but only one unified system interface (e.g. only one display)
  • This future GUI (e.g. AR enhanced) should change dynamically, based on the workflow phase detected by our system
  • Abundant information streams should be monitored by the system; only relevant information should be presented to the surgeon; surgeon should be notified of irregularities in the data
  • Example procedures: coronary artery dissection (?), spine surgery (Joerg)
  • Proposed evaluation method: Shadow evaluation by expert (invented by Christoph and Sandro)
    • Surgeon A performs surgery using Siemens GUI system in already existing OR
    • Surgeon B (paid expert surgeon) is sitting in another room with no communication channel to the surgeon A in the OR
    • Surgeon B evaluates the performance A live, compared to his own assessment of the surgery given the additional data from the adaptive GUI
• Prediction of surgery duration
  • Better scheduling of out-of-OR workflow

Research Goal 3: Training

• Visualization of trainee vs. expert comparisons
• Cooperation with Mussak:
  • Training surgeries with phantoms and pigs will be performed
  • 3D tracking possible
  • Diploma/Master Thesis: 3D tracking of surgeon movements similar to delivery simulator by Tobias Blum
  • End of August and in December: attend some of Mussak's training sessions for students

Research Goal 4 (?): Skill Evaluation

• Performed by (too?) many other research groups
• Trainee skills
• Expert skills

Research Goal 5: Automatic Report Generation

• Analyze surgical reports
• Identify needs of an automatic system

Short-term plan / Upcoming projects

Ahmad's Diplomarbeit: Parameter-free recovery of surgical workflow

• Possible data / workflow issues
  • Feussner Actibelt data
  • Jörg Actibelt data
  • Fake task Actibelt data
• Motif discovery on a given task sequence with branching
• Automatic, parameter-free derivation of the high-level workflow topology
• Design and evaluation of offline and online workflow recovery for new instances

Month	Nicolas	Tobias	Ahmad
Jul 2007			Journal Paper
Aug 2007	Attend training sessions of Mussak Get data from Greg Hager		Data acquisition with Actibelt (Joerg's data or fake data, wait for Feussner data)
Sep 2007
Oct 2007	11.10.-13.10.: CURAC (registration deadline: 31.08.07) 25.10.-02.11.: MICCAI + holidays in Australia
Nov 2007
Dec 2007	Second training session of Mussak

Teaching

Offered DA / MSc / BSc Theses, IDP, SEP, Hiwi

• 3D tracking of surgeon movements:
  • Data: Mussak's surgeries
  • Objective: build a system that supports Mussak in his training
• Signal projects:
  • Eyegaze
  • RFID for tool detection (ask Armin)
  • Inertia / accelerometer (after Ahmad's DA)
• Mulptiple Sklerose project: perform machine learning on MS datasets

CAMP I and II lectures

• CAMP I: 2-3 lectures
  • Lecture 1: Motivation and methods, (Leipzig lecture: “computer-assisted surgery”, OR navigation systems, workflow recovery)
  • Lecture 2: DTW for synchronization
  • Lecture 3 (?): ??
• CAMP II: 2-3 lectures and MATLAB exercises as homework
  • MATLAB 1: DTW

CAMP Workflow Seminar

• Update website

Praktika

• Think of an entertaining/exciting way to attract students to a workflow-related Praktikum (hire most interested/capable students at the chair afterwards)

Ph.D. and Post-Docs

• Ph.D. with Pierre Jannin (because: (Nassir) "he couldn't find any good Ph.D. students in France and German students didn't want to go to France")