Healthcare
❯
Treatment recommendation
❯Medical Assistants
❯AI solution to predict post-operative visual acuity for LASIK surgeries
Support system for optimization and personalization of drug therapy
For:
Public and private healthcare system, pharmaceutical companies
Scope:
This system is a full range of integrated solutions for the selection of the optimal
type of drug, its dose, and its combination with other drugs.
Goal:
Other
AI solution to predict post-operative visual acuity for LASIK surgeries
For:
Hospitals, patients undergoing LASIK surgeries.
Scope:
Predicting post-operative visual acuity for laser-assisted in SItu keratomileusis
(LASIK) surgeries from retrospective LASIK surgery data with patient follow-
ups.
Goal:
Automate a Business Process
Improving the knowledge base of prescriptions for drug and non-drug therapy and its use as a tool in support of medical professionals
For:
Doctors and patients
Scope:
Providing the medical professional with methods and means that would allow,
within the time allotted for the appointment of 0 patient with a known nosology,
to make a high-quality choice of drugs and to formulate a prescription
corresponding to good medical practices
Goal:
Improved Employee Efficiency
AI-based mapping of optical to multi-electrode catheter recordings for atrial fibrillation treatment
For:
Hospitals, cardiologists
Scope:
Predicting possible targets for atrial fibrillation ablation based on explanted human heart data of two modalities (multi-electrode mapping and near-infrared optical imaging)
Goal:
Improved Product Development / R&D
WebioMed clinical decision support system
For:
End-users (physician, nurse, laboratory technologist, pharmacist, patient)
Sales and marketing team
CDSS product development and maintenance team (system administrator,
system developer, system architect, project manager, and system
maintenance)
Scope:
Screening for cardiovascular disease risk prediction with machine and deep
learning methods
Goal:
Other
Integrated recommendation solution for prosthodontic treatments
For:
Dentist
Hospital
Scope:
In order to support complicated prosthetic treatments according to the patient's
condition, the artificial intelligence technology provides a comprehensive
analysis of the given information and situations to recommend various
prosthetic treatment methods and visualize them to support doctors and
patients.
Goal:
Improved Employee Efficiency
AI solution to predict post-operative visual acuity for LASIK surgeries
For:
Hospitals, patients undergoing LASIK surgeries.Goal:
Automate a Business ProcessProblem addressed
Given: Pre-operative examination results and demography information about
a patient.
Predict: Post-operative UCVA one day, one week and one month after surgery.
Scope of use case
Predicting post-operative visual acuity for laser-assisted in SItu keratomileusis
(LASIK) surgeries from retrospective LASIK surgery data with patient follow-
ups.
Description
Introduction to LASIK surgeries: Refractive surgeries for the
eyes are performed to correct (normalize) the refractive
state of the eye, to decrease or eliminate dependency on
glasses or contact lenses. This can include various methods
of surgical remodelling of the cornea or cataract surgery.
LASIK is a refractive eye surgery that uses a laser to correct
near-sightedness, farsightedness, and/or astigmatism. In
LASIK, a thin flap in the cornea is created using either a
microkeratome blade or a femto-second laser. The surgeon
folds back the flap, then removes some corneal tissue
underneath using a laser. The flap is then laid back in place,
covering the area where the corneal tissue was removed.
With near-sighted people, the goal of LASIK is to flatten the
steep cornea; with farsighted people, a steeper cornea is
desired. LASIK can also correct astigmatism by smoothing an
irregular cornea into a more normal shape. LASIK surgeries
61
are highly popular; over ten million LASIK procedures have
been performed in the United States alone in the past decade.
Motivation: While overall patient satisfaction rates after
primary LASIK surgery have been around 95 %, it may not
be recommended for everybody for two reasons: (1) high
cost with potentially no significant improvement for certain
types of patients, and (2) possible eye complications after the
surgery. LASIK surgeries cost approximately $2 000 USD per
surgery. An ability to predict post-operative UCVA can help
patients make an informed decision about investing their
money in undergoing a LASIK surgery or not. It can also help
surgeons recommend the most promising type of laser
surgery to the patients. How can we perform this prediction?
Further, while performing such surgeries, surgeons are
expected to set multiple parameters like suction time, flap
and hinge details, etc. These are often set using manually
designed rules. Can we design a data driven automated
method to suggest the best settings for a patient undergoing
a laser surgery of a certain type?
Problem definition: In this paper, we address the following
problem.
Given: Pre-operative examination results and demography
information about a patient
Predict: Post-operative UCVA after one day, one week and
one month after surgery.
Challenges: The problem is challenging because (1) a large
amount of data about such surgeries is not easily available;
(2) there are a lot of pre-operative measurements that can be
used as signals; and (3) data is sparse, i.e., there are a lot of
missing values.
Brief overview of our approach: We model the task as a
regression problem. We use domain knowledge to pre-
process data by transforming a few categorical features into
binary features. We also use average values to impute
missing values for numeric features. For categorical features,
we impute missing values using the most frequent value for
the feature. We evaluate multiple regression approaches.
Our experiments on a dataset of 791 surgeries provides an
RMSE of 0,102, 0,094 and 0,074 for the predicted post-
operative UCVA after one day, one week and one month after
surgery respectively.
Summary: We described a critical problem of predicting
post-operative UCVA for patients undergoing LASIK
surgeries. We modelled the task as a regression problem. We
explored the effectiveness of demographic, pre-operative
features and surgery settings for the prediction task. Using a
dataset of 791 LASIK surgeries performed on 404 patients
from 2013 and 2014, we tested the effectiveness of the
machine learning methods.