Coinciding with half a century of RANZCO, the 51st Annual Scientific Conference played host to lectures on a wide variety of topics. In part one of his report, LEWIS WILLIAMS details some of the many presentations on show.
The Royal Australian and New Zealand College of Ophthalmology (RANZCO) held its 51st Annual Scientific Conference at Sydney’s International Convention Centre (ICC), with the organisation also celebrating its 50th anniversary. The meeting was held in conjunction with the American Association for Paediatric Ophthalmology and Strabismus (AAPOS) and the Asia-Pacific Strabismus and Paediatric Ophthalmology Society (APSPOS).
The Congress commenced with the traditional Welcome Reception held on 8 November in Sydney’s Museum of Contemporary Art, followed by three and a half days of varying presentations and sessions. Many concurrent sessions were included to allow for the huge diversity of interests now necessary in modern ophthalmology.
A number of named lectures were carried over from previous events. The RANZCO Congress Opening Lecture was presented by UNSW astrophysicist Professor Lisa Harvey-Smith, the Norman McAlister Gregg Lecture was delivered by former CERA managing director and current Singapore-based ophthalmologist Professor Tien Y Wong, the Paediatrics Update Lecture was given by US ophthalmologist Professor David A Plager, the Retina Update Lecture was given by US ophthalmologist Dr Emily Y Chew, the Council Lecture was delivered by Australia’s Associate Professor Penelope Allen, the Cataract Update Lecture was presented by Singapore-based ophthalmologist Professor Soon-Phaik Chee, the Fred Hollows Lecture was delivered by Asia-Pacific-focused Fijian ophthalmologist and eye care advocate Dr Anasaini Cama, the Glaucoma Update Lecture was given by UK ophthalmologist Professor David Garway-Heath, the Dame Ida Mann Memorial Lecture was presented by the UK medical scientist and laser expert Professor John Marshall, and the final presentation of the Congress was the Oculoplastics Update Lecture given by Indian ophthalmologist Dr Santosh G Honavar.
At the graduation ceremony, RANZCO awarded its latest College Medal to Associate Professor Anne Brooks, who became the first female ophthalmologist to be presented with the award. She is a noted Melbourne ophthalmologist, educator extraordinaire, and head of the General Eye Clinic at the Royal Victorian Eye and Ear Hospital (RVEEH).
AI in ophthalmology: The Norman McAlister Gregg Lecture
Singapore’s Professor Tien Wong tackled this trendy topic, and his subtitle said it all; Concepts, Progress, Challenges, and Myths. Contrasting media coverage has suggested that AI is either completely ineffective at its task or capable of a transformative role in medicine.
Wong traced discussion of AI in medicine, or what passed for AI at the time, to a 1970 article by Schwartz in the NEJnlMed. Some 17 years later the same author posed the question: “Why Not?”.
While AI per se can be traced to the 1950s, it acquired the rider ‘machine learning’ in the 1980s and ‘deep learning’ by 2010. After traditional machine learning reached a plateau, deep learning is now almost synonymous with AI. In 2015, a dermatology paper reported a deep learning and neural network system that was trained on more than 750 classes of skin cancer. Its performance identifying cases was equivalent to that of at least two human specialists.
Turning to ophthalmology, Wong reported earlier work on diabetic retinopathy (DR) using deep learning and neural networks, which has already been approved by the US Food and Drug Administration as ready for clinical use. Applications for age-related macular degeneration (AMD) and diabetic macular oedema (DMO), as well as some other diseases, are also under investigation.
Early attempts with DR relied on pattern recognition of factors such as exudates, drusen, and cotton wool spots. That proved to be good (sensitivity 80-85%), but not good enough. A trial in Thailand, as well as a separate Google trial, now show results in excess of 99%.
The Singapore Eye Research Institute-developed SELENA early detection of DR system (Wong is the project’s principal investigator) has now been ‘trained’ on almost 500,000 images. The system’s versatility is being expanded through the use of additional images of glaucoma and AMD suspects. Its performance already approximates that of the Google-based system, despite its frequent use on non-ideal images and patient data.
Wong posed the obvious question: “Where does AI fit in DR screening?” The answer in Singapore is a triage approach, with the system’s output governing whether the patient is referred to an ophthalmologist, an optometrist, or ‘passed’. In other jurisdictions, AI and OCT data have been combined for analysis of other diseases. In one AMD study, the area under the curve (AUC – a plot of sensitivity [y-axis] against specificity [x-axis]) output was rated at 0.975. In a Google/Moorfields Eye Hospital study, an AI analysis preceded an examination by an optometrist with special, focused training. If necessary, a referral to an ophthalmologist followed.
Current data suggests that age, blood pressure, and RNFL status can be predicted accurately using a combination of a photo and AI. Deep learning of papilloedema has also shown accurate results (AUC 0.98).
Regardless of the promise, AI cannot fully compensate for poor input data. Furthermore, suspicion follows the current AI technology, whereby humans are unable to understand most of what happens inside the AI’s ‘magic box’, or clinically irrelevant or meaningless correlations with the target disease are offered. Making AI systems more accountable and understandable is a work in progress that still has some way to go.
Those factors, as well as predictable human suspicion of ‘black boxes’, means that physician acceptance will also need work. Wong concluded that AI is here to stay and will be disruptive, have a significant impact on health care, and inevitably affect ophthalmology but not replace it. Further progress and greater accuracy will depend on focusing systems to specific and narrow tasks. He also expects that AI will eventually assist in lowering the cost of health care.
Teaching and Learning Cataract Surgery
An afternoon session chaired by Melbourne-based ophthalmologist Dr Jacqueline Beltz devoted to the education and training issues surrounding cataract surgery was held in front of a considerable audience.
Beltz and colleagues run regular Melbourne courses in the fine art of ocular surgery, particularly cataract surgery, using VRmagic’s (Germany) Eyesi surgical simulators. Called the GENEYE VR Series, the courses target those interested in ophthalmology as a specialty, ophthalmology registrars, and RANZCO fellows interested in bettering their skills.
The session’s first presenter was Dr Danielle Buck, currently the chief ophthalmology registrar at the Royal Victorian Eye and Ear Hospital (RVEEH). She itemised the GENEYE programme’s aims as: enhancing surgical skills, the provision of a stress-free environment, and 24/7 access across wet, dry, and VR simulations of ocular surgery. It is proficiency-based and follows a structured, planned-sessions, lab-centric, four-month programme of six hours per week under a 1:4 supervision ratio. Transition to real eye surgery is based purely on merit.
A typical lesson starts with an introductory video that is followed with some theory, a live demonstration, and then some surgery on firmly mounted pig eyes. Assessments note suture skills and results, tissue handling and care, and wound tunnel and scleral tunnel formation. The dry lab uses the Japanese KITARO Surgical Training System, the Phillips Studio of Ophthalmic Simulated Surgery, and the KITARO dry model of capsulorhexis.
She described the Eyesi system as having a real phaco foot pedal and an operating microscope with realistic depth-of-field. Various studies have validated the equipment, often using posterior capsule rupture (PCR) rate as a criterion. One study demonstrated a 38% reduction in PCR after VR training.
An adequate result in three consecutive procedures is required to advance to the next stage of the training system. Even under close supervision, participants find it fun, supportive, and safe. Supervisor feedback is central to improving skill and decreasing the risk of surgical complications.
In charge of complications training is Dr Bernardo Soares, RVEEH’s Advanced Cataract Fellow. Generally, when complications are met during real surgery the training surgeon takes over. The most common issue remains posterior capsule tears (PCTs) or PCRs. PCTs are a challenge for both the trainee and supervisor, and either can perform a repair depending on the trainee’s skills and experience. A dropped crystalline lens nucleus remains the surgical aspect that engenders the lowest confidence level.
The unexpected need for an anterior vitrectomy, while less common now, is still the subject of ‘fire drills’ on a six-monthly basis. This is so staff become more familiar with the equipment, stress is reduced when dealing with the real thing, and decreasing the number of trips by attending staff in and out of the operating theatre. With progression, it is expected that trainees are more likely to repair their own missteps, gain confidence, and be able to undertake an anterior vitrectomy unsupervised.
Dr Tiffany Lo, an ophthalmology registrar at the RVEEH, detailed the qualities of an excellent trainer as: a good leader, a good communicator and able to teach good communication skills, resourceful, interested in training and further development, and professional in all education matters. Attributes include being approachable, being patient, being willing to let the trainee operate, and a willingness to offer encouragement as appropriate.
In a small survey Lo found that 67% of her trainee colleagues preferred trainers that asked their ‘charges’ if they needed any help. She reported the most difficult aspects of cataract surgery as being the capsulorhexis, nucleus cracking, segment removal, and irrigation and aspiration of the lens cortex. Despite the desire for hands-on operating experience, she stated that wet labs still had a role. She also believes that trainees need to be open to constant feedback, be flexible, and remain open to other useful input.
RVEEH and private practice ophthalmologist Dr Rahul Chakrabarti gave an overview of surgical assessment at his hospital. Among his numerous qualifications is a Master of Surgical Education, so he is well placed to lecture on the topic.
He claimed that the old Halstedian approach of surgical training, that being the apprenticeship model sometimes exemplified as the ‘see one, do one, teach one’ approach, was too subjective and lacked validity. Furthermore, defining surgical expertise was difficult and quantifying it even more so.
An objective alternative is preferred and to that end, the RVEEH uses a structured cataract surgery assessment tool to assess trainees and determine suitability for progression. He also noted that fewer adverse outcomes resulted from the approach. Validation of the transition from the old to the new surgical assessment system was still a work in progress. He also reported on the system being implemented at the Wilmer Eye Institute (Johns Hopkins Medicine, Baltimore) in which AI and deep learning are being applied to videos of trainees in action.
Experienced educator Dr Cathy Green discussed the issues of trainer and trainee feedback, which are usually based on reflections from individuals on both sides of the interaction. Effective feedback is based on learners’ insight and must be both timely and constructive if the trainee is to improve. It should involve appropriate input, be delivered in an appropriate setting and, importantly, be both useful and encouraging.
She prefers the so-called Pendleton Framework in which the learner is asked for feedback (e.g. “What went well and why?”) and likewise, the trainer is asked for similar feedback and the responses compared. The feedback is tailored to be relative to the stage of education reached. This can be found to be a good early warning system of emerging difficulties.
An aim is to understand why the trainee took the action they did – see, hear, think, wonder. A failure is usually when feedback gives mixed messages or leads to false impressions. She also stated that feedback on the trainer was important, although issues include determining how good it is and what criteria apply.
The session’s final presentation was delivered by session chair Beltz and was titled Mind Skills – striving for excellence. She gave the relevant types of expertise as medical, routine, and adaptive, all of which rely on technology, innovation, and changes.
Derived originally from military applications, learning mind sets relates to adaptive expertise, that being the degree of skill developed, which is valued over efficiency. The ability to be flexible enough to adapt to different tasks, often quickly, is also highly valued. Such an approach is required because while patient safety can often be protected by supervision and preparation done in labs is useful, when it comes to cataract surgery things can evolve very rapidly and the presentation might not conform to any previous experience or training.
As more complex tasks become the preserve of special surgeons, routine tasks will eventually be left to lesser qualified people such as optometrists and nurse practitioners.
Mind skills include: situation awareness, decision making, communications, teamwork, leadership, psychological flexibility, and the health and wellbeing of all involved.
Currently, the RVEEH is training registrars in mind skills – ACT (acceptance and commitment training). The core training is being done by special psychologists under a grant provided by Johnson and Johnson. Green reported that the ACT programme required significant ambition on the part of the participants. An expressed aim was to identify promptly, those less capable so that remediation or career changes can be enacted. The system aims to identify at least 75% of those with surgical difficulties in their first year of training so that rational decisions can be made sooner rather than later.
Paediatrics Update Lecture: IOLs in under two-year olds
This lecture was delivered by Professor David Plager, Professor of Ophthalmology, School of Medicine, Indiana University.
He reported that in the 1950s, a consensus of relevant specialists recommended against unilateral ocular surgery. The evolution of the automated vitrector in the 1970s, a revolution at the time, was described as a significant step forward. The period from 1973 to 1981 saw some authors revising their thinking and by 1994, some 46% of AAPOS members were using intraocular lenses (IOLs) in young patients. By 1996 IOLs, mainly posterior chamber IOLs, were the standard of care for over two-year olds in the US.
In 1997 a seven-centre study of IOL usage in younger cases was undertaken. By 2004, a 12-site Infant Aphakia Treatment Study (IATS) was underway for unilateral aphakia, in which the choice was an IOL or a CL. The study’s primary hypothesis was that by 4.5 years of age, IOLs gave better Visual Acuity (VA). A secondary hypothesis was that strabismus, patient stress, and eye patching aspects were also improved.
In a later study (2009 to 2014), Tan showed that in 114 eyes of 112 patients, the VA was the same for IOLs and CLs (decimal VA 0.9), with little real difference in practice. The main adverse response at one year and five years was visual-axis IOL opacification. The formation of a so-called Soemmering ring, an after-cataract opacity that occurs in those operated on when very young, required further surgery. Glaucoma was also found to be a risk, but the difference between CL and IOL cases was not statistically significant. Interestingly, 20% of the IOL group at the five-year mark were found be high myopes, a figure expected to increase with further ageing.
In extreme Rx cases, Plager suggested that if CLs prove to be satisfactory they at least be used initially.
In 2019, the 10-year data became available from Tan’s study. The results showed 25% had good VA, 50% had poor VA (6/60 or worse), and some strabismus cases persisted. Furthermore, additional surgeries were required in the IOL cases at about twice the rate for those fitted with CLs. Perhaps more importantly, about 47% of CL cases either had glaucoma or were glaucoma suspects, whereas only 36% of IOL cases were classified similarly.
Plager then raised the topic of an even younger cohort – the seven to 12-month olds. Data showed that group faired about the same, tending slightly worse, than the original IATS study. Those aged 12 to 24 months were similar or slightly better than the IATS group. Adverse events in the 12 to 24-month group were less common (24% versus 81%), although extra surgeries were fewer than in the IATS experience. This is possibly due to improvements in technique, biometry, and tools.
A later study, TAPS2 (Toddler Aphakia and Pseudophakia Study 2), demonstrated VA better than IATS (85% had 6/24 or better VA) and just 1% had VA worse than 6/60. Importantly, some cases had binocular vision with IOLs. Most adverse outcomes were glaucoma-related. IOL implantation in those under seven months of age offered no VA advantage (still 50% had worse VA than 6/60) and no glaucoma protection was imparted by IOL implantation.
However, in such a young age group there were more adverse events and extra surgeries required. The foregoing translates to no recommendation for IOL implantation in the <seven-months group, but IOLs might be advantageous in the older groups studied.