Regulatory affairs, causal inference, safe and effective health care in machine learning for Bio-statistical services - Pubrica
                     AN OVERVIEW OF REGULATORY  
AFFAIRS, CAUSAL INFERENCE, 
SAFE  AND EFFECTIVE HEALTH 
CARE IN  MACHINE LEARNING 
FOR BIO-  STATISTICAL SERVICES
An Academic presentation by
Dr. Nancy Agens, Head, Technical Operations, 
Pubrica  Group: www.pubrica.com
Email: [email protected]
Today's 
Discussion
Outlin
In Brief  
e Introduction
Regulations for Safe and Effective Health Care Machine 
Learning  Limitations
Transfer Learning
Biomarkers in 
FDA  Conclusion
In-
Brief
Over the past few years, the magnitude of machine learning in the 
 field of healthcare delivery setting becomes plentiful and 
captivating.
Many regulatory sectors noticing these developments and the FDA has 
 been appealing to provide bet machine learning services with safe and 
 productive use. Despite having the limitations in software-driven  
products, FDA leads to giving a significant benefit of causal inference  
for the development of machine learning. FDA is giving suggestions to  
provide well equipped regulated products. Pubrica is here to help you  
with the regulated for Bio-statistical consulting services.
Introductio
n The significance of machine learning has evolved 
 globally, especially in th field of medical and  
healthcare sectors.
Many tools are significant for various purposes likes  
diagnosis, software tools for many clinical findings in 
 multiple areas.
The machine learning paves an easier way to c 
linical  Bio-statistical services using many software 
tools.
Contd
..
It creates an excellent standard on radiology and  
cardiology and improves the patient’s medical issues  
rapidly, more comfortable decision making in clinical trials.
All these maintained by drafting a set of regulations by  
various government sectors around the world.
Contd
..
CAUSALITY MATTERS 
IN  MEDICAL 
IMAGING
Regulations for  FDA is a regulatory organization there to perform 
 the quality of any medical or clinical testing  
Safe and equipment, medicines, or any food-related  
Effective  products.
Health Care  
FDA is looking to provide the best facilities in  
Machine health care sectors through machine-learning  
Learning artificial intelligence services for the s tatistical  p
1. F DA (food and rogramming services.
Drug   A
Though it is not an urgent need for ML-driven  
dministration) tools, there are few benefits of using ML-driven 
 tools in medical fields, says FDA
Contd
..
2. Applications
Instrumental usage
Machine implementation
Invitro reagents implantation technology
Diagnostic kit
Treatment for humans and animals.
Contd
..
3. F DA 
definition
The usage of ML can provide both physical 
 equipment and software tools.
This software device is known as SiMD 
 (software in a medical device).
International medical device regulators verify 
 these software-driven tools.
Contd
..
4. C hallenges in 
SiMD
Cybersecurity
Management of data 
 Collection of data
Protecting 
information
To create 
opportunities in 
patient’s care
Limitations For some reasons, the FDA does not regulate twoapplications of ML systems. They are
Clinical design support software(CDS)
Laboratory developed tests.
The actual reason for exempting these uses are CDS 
 provide instance decision making, which may 
affect  in the future.
On the other side Laboratory, developed tests can 
 access only one available health care.
FDA cannot regulate these type of software.
Contd..
Last year FDA released a paper after conducting a serious discussion with the  
regulatory members and proposed “Regulatory Framework for Modifications to
Artificial Intelligence/Machine Learning (AI/ML)-based Software as a Medical 
Device.”
For statistics in clinical research. It includes some premarket research products 
 approval procedures that would delay the ML process.
Many Bio-statistical firms raised few critics against it.
The objective of the proposal is to give access to real-world data using ML products 
 more efficiently with some regulatory barriers.
Contd.
.
The objective of the proposal is to give access to real-world data using ML  
products more efficiently with some regulatory barriers.
It also includes some real-world affirmations.
To overcome this, the FDA officials spoke to the public to create awareness about 
 the “approach of regulating algorithms”.
Regardless of all benefits and limitations, ML is facing challenges in the  
development of the safe and efficient product. Some of the challenges are
Contd.
.
ML identifications
ML predictions
ML recommendations
ML algorithms for diagnostic tools
To overcome this, Subbaswamy and Saria provide some potential remedies by 
 discussing the statistical foundations in the Bio-statistical analysis.
Data curation of individual patient’s health raises questions for request algorithms 
 to give a more specific context.
Transfer 
 The process of learning a task from the already-  
completed job through knowledge transfer is called 
Learnin  transfer learning.
g
However, this process is complicated.
The datasets can affect the algorithms, resulting in  
the false provisional services in h ealth care analysis
.
This process is not allowed in the medical sectors.
Biomarker In the process of validation of a biomedical tool, 
s  in FDA  biomarker validation is mandatory in the
 clinical research services.
There are so many parameters for qualifying a 
 biomarker.
The casual inference is a novel digital 
 biomarker validation.
An ML algorithm that detects the patient’s  
therapy benefits may not be relevant unless a 
 casual inference tool access in that 
biomarker.
Contd..
Some make a precise diagnosis and treatment recommendations to 
 understand the factors in ML algorithms.
The production of digital biomarkers facing more challenges to incentivizing 
 parties in health care sectors.
R&D validated provide significance in delivery of healthcare services.
Studies say that statistician’s tool kit has grown fast, and various technical 
 tools have a development for causal inference of machine learning in
b iomedical investigations and reviews.
Conclusion
Wrapping up, in a complex environment, the role of  
regulatory affairs in biomedical studies for machine 
 learning is essential.
One of the easiest ways to support the regulators is 
 the usage of biomarkers in h ealthcare tools.
These regulations help to provide better  
healthcare services under the guidance of pubrica.
Contact 
Us
UNITED KINGDOM
+44-1143520021
INDIA
+91-9884350006
EMAIL
[email protected]