1 of 55

AI in surgical training

Mark Coleman

2 of 55

Mark Coleman

Consultant Surgeon

Clinical Director

Associate Professor

Colorectal Unit

University Hospitals Plymouth NHS Trust

3 of 55

Disclosures

  • Co-director Lapco International Consultancy 2017-present

  • National Clinical Lead Lapco 2008-2014

  • No influence on published work

4 of 55

Define AI

The capability of a computer to perform actions analogous to the human brain (McCarthy 1956)

5 of 55

Teenage sex

6 of 55

7 of 55

Becoming an expert surgeon

motivation

training

practice

persistence

awareness

Kind Thanks to Roland Valori and Robin Kennedy for this slide

Effective Health Care 1999;5(1)

persuasion

facilitation

consciously expert

unconsciously inexpert

Trainee

Trainer

SYSTEM

8 of 55

Becoming an expert surgeon

motivation

training

practice

persistence

awareness

Kind Thanks to Roland Valori and Robin Kennedy for this slide

Effective Health Care 1999;5(1)

persuasion

facilitation

consciously expert

unconsciously inexpert

Trainee

Trainer

SYSTEM

AI

9 of 55

AI uses in surgical training

Selection

Assessment

Pre-procedure training

Technical competency

Technical guidance

Technical intervention

Non-technical skills

Post procedure analysis

Sign off

10 of 55

AI in training - ?’s

Can AI measure performance objectively in surgery?

Are significant differences observed between novices and experts?

Does surgical Surgical performance affect clinical outcomes?

Can AI provide real time feedback?

Future possibilities of AI – image overlay – feedback - intervention

11 of 55

AI & training

What is its role now?

Future role

Cost

12 of 55

Priorities for AI in training

Objective

Automated

Realtime

Cost effective

13 of 55

Competency evaluation

Reliable

Validated

Quick

Intuitive

14 of 55

Objectives of surgical training

EFFICIENT

EFFECTIVE

SAFE

COST EFFICIENT

15 of 55

Basic operative AI metrics

Time to completion

Instrument path length

Errors – consequence (corrected/uncorrected), no consequence

16 of 55

Confirming Objective AI metrics in training

Construct validation – expert vs novice

Concurrent validation – expert evaluation

Reliable – reproducible between raters

Live feedback

17 of 55

Selected 4 steps performed by widest range of surgeon experience

Bladder mobilization

Seminal vesicle dissection

Anterior Vesico-urethral anastomosis

Rt lateral pelvic LN dissection

Residents, fellows and surgeons

>100 cases = expert

GEARS – global evaluative assessment of robotic skills – 3 experts

18 of 55

AI in RARP – DaVinci + DV logger 100 cases

  • Instrument time
  • Instrument path length
  • Velocity of instrument
  • Frequency of master controller clutch use
  • Energy use
  • Camera use
  • Third arm swap

19 of 55

results

20 of 55

results

21 of 55

results

22 of 55

Results – 3D instrument path tracings

23 of 55

DEVELOPING BASIC HAND-EYE COORDINATION SKILLS�FOR LAPAROSCOPIC SURGERY USING GAZE TRAINING�Mark Wilson*, Mark Coleman† and John McGrath – Department of Urology, Royal Devon and Exeter NHS Foundation Trust and *Department of Sports and Health Sciences, University of Exeter, Exeter; and †National Training Programme in Laparoscopic Colorectal Surgery (LAPCO), Plymouth, UK�Accepted for publication 12 January 2010 �© 2010 THE AUTHORS JOURNAL COMPILATION © 2010 BJU INTERNATIONAL��Novices see instruments, experts see the field�Gaze analysis can be used for training ��

24 of 55

Conclusions - 1

Can distinguish between experts and novices for real-time events and kinematics

25 of 55

Surgical competence

  • In surgery, competence is the ability to successfully apply professional knowledge, skills, and attitudes to new situations as well as to familiar tasks. A critical step in assessing surgical competency is developing methodology for competency evaluation and certification.

Time

Independence

Error

26 of 55

Is Competency Assessment at the Specialist Level Achievable? A Study for the National Training Programme in Laparoscopic Colorectal Surgery in England

  • Miskovic, Danilo MD, PhD, FRCS*; Ni, Melody PhD*; Wyles, Susannah M. MBBS, MSc, MRCS(Eng)*; Kennedy, Robin H. MD, FRCS; Francis, Nader K. PhD, FRCS; Parvaiz, Amjad FRCS, FRCS(Gen)§; Cunningham, Chris MD, FRCS; Rockall, Timothy A. MD, FRCS; Gudgeon, Andrew M. MBBS, FRCS**; Coleman, Mark G. MD, FRCS††; Hanna, George B. PhD, FRCS*
  • Author Information Annals of Surgery: March 2013 - Volume 257 - Issue 3 - p 476-482

27 of 55

GAS FORM EXAMPLE

GRADING OF STEPS

1- 6 SCALE

28 of 55

GAS Form data – raw – vs CUSUM

29 of 55

CUSUM plot – 3 trainees & average scores

30 of 55

GAS form scores – procedural steps

31 of 55

GAS form scores – LCS procedures

32 of 55

Competency AssessmentObservational clinical human reliability analysis (OCHRA) (Concurrent Validity)

Miskovic et al Surg Endosc 2012 Mar;26(3):796-803

33 of 55

Competency Assessment�Reliable prediction of competent performers (Concurrent validity)

Miskovic et al Surg Endosc 2012 Mar;26(3):796-803

34 of 55

Sign off and outcomes

Danilo Miskovic, Susannah M Wyles, Fiona Carter, Mark G Coleman, George B Hanna. Surgical endoscopy 2011, 25 (4) 1136-1142

35 of 55

Lapco sign off: risk factors for failure�

          • Incomplete video (OR=8.7)
          • Distal division vasc pedicle (OR=2.9)
          • Inadequate resection (OR=5.8)
          • SB Injury (OR=4.8)
          • Damage to mesentery (OR=3.4)
          • Unclear pedicle plane (OR=4.1)
          • Out of dissection plane (OR=3.8)
          • Poor anastomosis (OR=11.6)

36 of 55

Competency Assessment�Clinical Outcomes Related to LCAT Score (Predictive Validity)

37 of 55

Lapco Competency Assessment – Lessons Learned

  • Competency assessment is an integral part of a training programme
    • It identifies under-performers and difficult procedures
    • It ensures high operative standards are achieved

  • Educational research enables design and validation of assessment tools

  • Assessment of consultants is feasible and acceptable

  • High technical standards improve clinical outcomes

38 of 55

Does performance correlate to outcomes & can this be measured objectively?

39 of 55

40 of 55

  • 20 videos of lap bypass from different surgeons – self selected
  • Edited
  • Critical components of each procedure
  • Peer reviewed
  • Anonymous
  • Validated OSATS tool (objective structured assessment of technical skills)
  • Gentleness, tissue exposure, instrument handling, time and motion
  • Rated 1-5 (1 = resident, 5 = expert/master)

41 of 55

  • Outcomes
  • Any significant complication
  • Reoperation
  • 30-day mortality
  • Readmission
  • ED visits

42 of 55

  • 10 reviews of each video
  • Mean score
  • No significant rating bias

43 of 55

44 of 55

  • Scores 2.6-4.8
  • 4 quartiles
  • Bottom quartile mean score 2.9
  • Top quartile mean score 4.4
  • Surgeons in bottom quartile took 40% longer and perform fewer cases

45 of 55

46 of 55

Conclusions - 2

Surgical skills correlate to clinical outcomes……but do they longterm?

47 of 55

� J Clin Oncol 2010 10;28(2):272-8.� doi: 10.1200/JCO.2009.24.1448.

  • Complete mesocolic excision with central vascular ligation produces an oncologically superior specimen compared with standard surgery for carcinoma of the colon

48 of 55

49 of 55

50 of 55

New Lapco Digital Learning Platform

Aim: to create a versatile user-friendly digital training platform for surgeons

51 of 55

Protect- Commercially Sensitive. Copyright Lapco Consultancy Ltd 2021

  • Faculty (administrators) can build and manage their Lapco TT courses with automated course design and booking features.

  • Other types of courses can be integrated on the system.

  • Courses can be promoted across the user network.

52 of 55

Protect- Commercially Sensitive. Copyright Lapco Consultancy Ltd 2021

  • Practical, quick and practical assessment tools to record and evaluate trainee and trainer performance at all stages of the pathway.

  • Fully integrated communication between the trainee and assessor.

  • Dashboard to evaluate data and compare results.

  • Integrate (or develop) bespoke tools to assess training, clinical and economic outcomes.

53 of 55

Protect- Commercially Sensitive. Copyright Lapco Consultancy Ltd 2021

  • Trainers (attendees) receive course materials directly to their account to access at any time.

  • Attendees complete ‘Course Assessment form’ to automatically generate their certificate.

  • Course feedback recorded to maintain programme quality.

54 of 55

summary

AI can be used to evaluate performance

Performance correlates to clinical outcomes

AI can already provide real-time feedback

AI enabled robotic interfaces could guide and correct surgeons

55 of 55

Thank you!

mcoleman1@nhs.net

@lapcoleadmark

+447789873190