1.
Introduction
Seriously
I don’t trust a computer system when it comes to health related issues
including robots performing surgery. Human doctors could react quickly to an
emergency situation that was out of previous plans.
There
was a discussion about developing an AI system to assist doctors to screen
health issues. Especially waiting in an emergency a hospital room is too long
for a patient. In Canada if we’re sick outside of regular clinic hours, we
could go to any hospital’s emergency department and wait for a medical doctor.
However, the nurse would provide initial screening with a few questions and
taking temperature. Waiting sometimes could take a few hours.
The
AI system would help hospitals to pre-screen patients with symptoms; so on duty
doctors would make final decisions quickly.
2.
Implementation
Developers
simply enter the entire books of medicine in the system including symptoms.
This is a question of design a good database for quick processing.
This
system could be a web based system with patient, doctor, nurse, etc. logging in
different web pages or access system database.
3.
Hospital version
After
pre-screening patients by a nurse, the patient enters a waiting room with
temperature and blood pressure taken, if needed. The nurse would provide the
patient with a system ID, thus it could be used as identification in the AI system.
The
nurse could use the same AI system to enter the temperature reading, blood
pressure, and notes (currently taken any medication) via a web page. This would
eliminate this task by patients.
The
AI Doctor Assistant system would ask or survey the patient with a series of questions
by both voice and text on a computer. For example,
a.
Initial Questions
Patient
would enter the ID provided by the nurse to trigger a start of pre-screening
issues, if it was not done by a nurse. Patients would enter current medication,
if not entered by the nurse.
System
would show the following questions by voice and text on a screen
Why
did you come to our hospital?
-
Pain
-
Fever
-
Other health issues
b.
Pain
System
would provide 2 images of our body, so patients could click on the area of
pains. Depending on the location of pain, different series of questions would
be asked.
c.
Fever
Different
series of questions would be asked.
d.
Other health issues
Based
on the suggestion of an experienced medical doctor, this should list common
symptoms to make questionnaires short and effective
-
Rash (leads to different questions including location of rash)
-
Cough (leads to different questions)
-
Trouble to sleep (leads to different questions)
-
Etc.
e.
Decisions
The
system would store answers of the questionnaire from (a), (b), and (c) in database.
Based
on the questionnaires and nurse’s standard personal notes, the AI system would “guess”
a health problem and recommended medications in database. This information won’t
be presented to patients as on duty doctors would make final decisions.
Doctors
in the emergency room would open the system and viewed patient’s answers as
well as suggested health issues and medication by system.
4.
Medical clinic version
This
version is similar to the hospital version. The exception is the patient must
take his/her own temperature and blood pressure by themselves with equipment
provided by the medical clinic. An ID is provided to patient, if healthcare ID
was not used to identify a patient.
Based
on the questionnaires, the AI system would “guess” a health problem and recommended
medications in database. This information won’t be presented to patients as the
medical doctor would make final decisions.
The
medical doctor logins system and uses proper patient’s ID to view data and
information presented by the system.
5.
Web version for general public
There
are many under developed countries that are shortage of qualified medical
doctors. Therefore, ICO charity department wanted a web version to provide free
screening of sick people using web services via Internet. There would be many translated
versions in different languages.
This
version is similar to the medical clinic version, where answers of questionnaire
are stored in database as well as suggested medications and guessed health
issues.
Patients
must take their own temperatures. It’s not a common scenario for general public
to keep a blood pressure device at home.
This
web version only presents “guessed” health issues and suggested the users to
see a medical doctor, if serious problems.
System
would provide an ID of this session to users, thus an authorized medical doctor
could log in system and retrieve information related to this questionnaire.
It
is recommended that the web version would be available to general public after
a few years of use in hospital and medical clinics. This is to prove its
accuracy and effectiveness.
6.
Cost of system
By
using Windows 12 server, it could bring the cost of a system to under $20K USD,
thus any medical clinics could easily afford. Doctors in private medical clinics could perform screening by themselves, thus they wouldn't invest in an expensive system.
No comments:
Post a Comment