Page 122 - The Indian Optician Digital Edition September-October 2022
P. 122
nowadays only because of the application of AI.
AI can also be used in the surgical skill training. APPLICATION OF AI IS USEFUL
AI is also being successfully used to predict
the progression of ECU and eventual need of IN ADDRESSING ANTERIOR
capsulotomy. SEGMENT DISORDERS SUCH
POSTERIOR SEGMENT DISORDERS AS KERATOCONUS, INFECTIOUS
KERATITIS, REFRACTIVE SURGERY,
There is enormous role of AI in the diagnosis
and treatment of Posterior Segment Disorders. CORNEAL TRANSPLANTATION
AGE-RELATED MACULAR CATARACTS, ANGLE-CLOSURE
DEGENERATION (AMD) GLAUCOMA, CONTACT
LENS PRACTICE AND TELE
AI is useful in predicting the likelihood of
developing wet AMD from the dry AMD. AI OPHTHALMOLOGY
algorithms are utilised in the prediction of
likelihood of need of giving intravitreal anti-
VEGF injections. in neuro-ophthalmic diseases, detection of eye
movement disorders (Strabismus and Conjugate
DIABETIC RETINOPATHY (DR) Gaze Abnormalities, Nystagmus), classical
We take the fundus photographic images machine learning in ocular myasthenia gravis.
and subject them to AI to get the diagnosis LIMITATIONS OF AI
and staging of diabetic retinopathy. AI is useful
in the processing of OCT images, non-invasive The computational cost and training
fundus angiography. experience is high due to the forming of an
algorithm which indicates that AI might be
RETINAL VEIN OCCLUSION (RVO) useful for diseases with higher morbidity
Efforts have been made to utilise ML in RVO whereas it might be unavailable for rare
utilising CNN combined with patch-based diseases. AI could not wholly identify a disease
and image-based vote methods to recognise separated from a human’s intervention as
the fundus VRVO automatically with higher computers recognise structures and features
degree of success. mechanically. Thus, a small part of unusual
features could be missed. Also, if the relation
RETINOPATHY OF PREMATURITY (ROP) between the input and the expected output is
complicated, the machine would not be able to
AI has been tried to automatically identify
the ROP stage from plus to not-plus disease build a model.
with high accuracy. CONCLUSION
NEURO-OHTHALMOLOGY AND There are some algorithms formed in the
STRABISMUS field of Ophthalmology such as DR, ROP, AMD,
RVO, glaucoma, cataract, etc. Not every image
Unlike other ophthalmic sub-disciplines, could be identified precisely or missed which
neuro-ophthalmology has not gained much not only depend on computer techniques but
from the latest investigative instruments. Still also on the quality of the input images. The
there are distinct applications of AI in neuro- analysts should be trained well because of the
ophthalmology. These include, classification importance of the marking process as it is the
of optic nerve appearance (Papilledema, foundation of computer learning.
Pseudo-papilledema, and other Optic nerve
head abnormalities (ONH), ONH pallor, Thus, AI can proficiently conduct a task, but
glaucomatous versus non-glaucomatous optic at some level, human intervention is crucial
neuropathy, exploration of optic nerve function during the process.
118 | THE INDIAN OPTICIAN | SEPT-OCT 2022 CLINICAL
