Natural Language Processing Uses in Healthcare Laura Pietromica - - PowerPoint PPT Presentation

▶

Apr 23, 2023 180 likes •417 views

Natural Language Processing Uses in Healthcare Laura Pietromica Customer Advisor & Consultant HIMSS Certified Consultant NLP Overview Agenda Potential in Healthcare Challenges NLP is expected to be a $22.3 billion

SLIDE 1

Natural Language Processing – Uses in Healthcare

Laura Pietromica

Customer Advisor & Consultant HIMSS Certified Consultant

SLIDE 2

Agenda

NLP Overview
Potential in Healthcare
Challenges

SLIDE 3

NLP is expected to be a $22.3 billion industry by 2025.

Source: Tractica, August 2017

https://www.tractica.com/newsroom/press-releases/natural-language-processing-market-to-reach-22-3-billion-by-2025/

SLIDE 4

NLP Overview

SLIDE 5

Natural Language

Vocabulary Text Language

SLIDE 6

Artificial Language

Java Python C#

SLIDE 7

Computer interacting with language?

Rules / Statistical Engine Understand Interpret Manipulate

SLIDE 8

Rules / Statistical Engine

BRIEF HISTORY: The patient is a 44-year-old cigarette smoker with a markedly positive family history of coronary artery disease who has an electrocardiogram showing an old anterior septal myocardial infarction. Nuclear study showed a fixed anterior defect. He has had episodes of chest pain and coronary angiography has been recommended. DESCRIPTION OF THE PROCEDURE: One percent lidocaine was infiltrated over the right femoral artery. A 6-French sheath was placed in the right femoral artery. Diagnostic coronary angiography was performed with a 6-French JL-4 and 6-French JR-4 diagnostic catheters. Following coronary angiography, a 6-French pigtail catheter was placed in the left ventricle where left ventriculography was performed with 36 cc of contrast injected at 12 cc per second. At the conclusion of the procedure, the catheter and sheath were removed and Angio-Seal plug was deployed. TECHNICAL FACTORS: Medications: Fentanyl 100 mcg and Versed 2 mg. Contrast: Isovue 130 cc. Fluoroscopy time: 2 minutes. HEMODYNAMICS: Opening aortic pressure 103/58. Following coronary angiography, the left ventricular pressure was 107/12. There was no aortic stenosis on left heart pullback. CORONARY ANGIOGRAPHY: There was no significant obstructive coronary artery disease in this right dominant system. There is a 10% to 20% stenosis in the distal left main. The left circumflex artery is a moderate-sized vessel filling a large first obtuse marginal and diminutive second obtuse marginal and there is no significant disease. The left anterior descending coronary artery is a large vessel which extends to the apex. It fills several small diagonal branches. There are no significant stenoses. There is a large ramus intermedius which fills the lateral wall. It, too, has no significant disease. The right coronary artery is a large dominant vessel filling a moderate-sixed posterior descending artery and two larger posterolateral branches. There are no significant obstructions in the right coronary artery. LEFT VENTRICULOGRAPHY: The left ventricular systolic function is noimal. There are no regional wall motion abnormalities. There is no mitral regurgitation. DIAGNOSTIC IMPRESSION:

1. There is no significant obstructive coronary artery disease in this right dominant system. There

is a plaque in the distal left main, but there is no significant obstruction.

2. Left ventricular systolic function is normal. There is no evidence of previous anterior wall

myocardial infarction. PLAN: I do not have a definite explanation f or the patient's abnormal electrocardiogram or abnormal nuclear study. Coronary spasm remains in the differential diagnosis and we have encouraged him to completely discontinue smoking, which he has accomplished. He certainly can take nitroglycerin

n an as-needed basis in the future. I would continue aspirin indefinitely.

SLIDE 9

Potential

SLIDE 10

“Why are you asking me this? You have my chart!”

Identify Key Concepts, Facts, Trends and Phrases Executive Summary

SLIDE 11

saykara

Siri

SLIDE 12

SLIDE 13

Analytics & Data Mining

QRISK

– Cardiovascular disease

HLA’s Horizon

– Oncology

IBM Watson

– Radiology – Drug discoveries – Clinical Trial Matching

SLIDE 14

Sean Captain, Can IBM Watson Do It All, Fast Company

A doctor reads about a half dozen medical research papers in a month…whereas Watson can read half a million in about 15 seconds.

SLIDE 15

Challenges

SLIDE 16

SLIDE 17

Language

Regional dialects
France or France
Phonetics
Celery / Salary
Hear / Here
Ice Cream / I Scream
Peas / Peace

?

SLIDE 18

Complexities

Structured vs Unstructured

– Data set – Notes section

Different Senses & Meaning

– Eye drops – Murmur appreciated – Mass at 3 o’clock

SLIDE 19

Review Process

Design / Modify Test Deploy Feedback

X ? 

SLIDE 20

AI Is Inventing Languages Humans Can’t Understand. Should We Stop It?

Mark Wilson, Fast Company

SLIDE 21

For More Information…

iCIMS & Health Language Analytics

Booth 63

Alcidion

Booth 29

SLIDE 22

Thank you

Laura Pietromica

Customer Advisor & Consultant HIMSS Certified Consultant

Laura.Pietromica@hyland.com +1.216.407.6769 Booth 51