How Digitization in Diabetes is Changing the Landscape?

CME INDIA Presentation by Dr. Sridhar GR, MD DM FACE FRCP, Endocrinologist and Diabetologist, EDC-Vishakhapatanam, AP, India.

Based on APDF Oration, Annual Conference, Ongole, Andhra Pradesh, 27th November 2022.

Digitosome

• ‘Data generated online and by digital technologies.’

AI driven wearable devices: Studies

• 37 studies
  • Main measure: related to blood glucose (41%).
  • Most assessed: prediction of glucose levels.
  • Less than 50% were commercially available.
  • Wrist-worn devices (>50% studies).

• Artificial intelligence methods used:
  • Machine learning.
  • Support vector analysis.
  • Random forest.

Development of databases

What and how much information?

• Important consideration.
• Depends on the purpose for which the information will be used.
• Have a coarse-grained information of the patient population?
• Use specific information for specific purpose?
• Trade-off between depth of information versus the time available for each patient.

What kind of interaction?

• If the physician herself enters the information, it is possible to observe and assess not just what but how the information is given.

The EDC Database

Rule based system

• Diagnosis and instructions are automatically generated
  • E.g.
    • Hypertension (BP levels).
    • To stop smoking.
    • Foot care.

Overview

• Use of existing data.
• Clinical analysis:
  • Traditional.
  • Non-traditional.
• Genetic data:
  • Novel uses.
• Other sources:
  • Clinical, biochemical.
  • Novel sources.

Clinical characterization: from EDC data

Biology and information come closer

Potential

Example

Harrison’s Textbook of Internal Medicine: 21st edition

• Contains a chapter on ‘Machine learning.’
• Artificial intelligence refers to ‘machines performing human-like cognitive functions (e.g., learning, understanding, reasoning and interaction). (8)
• This can at best be considered a starting point, because one must first define what is meant by ‘intelligence’ and by ‘artificial.’
• AI doesn’t aim at imitating humans, but seeks to be inspired by them.
• It may be considered a discipline that ‘uses the computer-processing capabilities of symbols to find generic methods for automating perceptual, cognitive and manipulative activities via algorithms.

What then is artificial intelligence (AI)?

• ‘The science and engineering of making intelligent machines, especially intelligent-computer programs.’ (11)
• The term is used interchangeably with machine learning (ML)
  • ML is an AI technique used to design and train software algorithms to learn from and act on data.

Data other than from EMRs

• The widespread use of electronic medical record system EMR. (9)
• Omics data from nucleotides, proteomics and metabolomics. (10)
• Food:
  • Photograph, analyze.
• Stress:
  • monitor and resolve.

Current constraints

• Constraint lay not in the availability of data, but in the ability to analyze and put it to appropriate use for improving clinical care.

Practical use of AI in clinical medicine

• AI can be profitably employed as an initial screening tool before referral to a specialist.
• This would lower the workload on the specialist, who can then carefully deal with those who have suspected disease requiring specific attention and treatment.

Avoid ‘junk in junk out’

Apportioning blame for errors

• It is difficult to apportion blame when an error occurs as a result of using AI in decision making:
• Who is to blame when there are so many sources of variability?
  • the provider of training data, the developer of the algorithm or the clinician who employed the system?

Role of AI (Artificial Intelligence)

• Artificial intelligence can improve care by automating and aiding in decision making.

Use of Deep Neural Network for Image Processing

• The most evolved area where neural network is applied in clinical medicine is, image processing: identification of diseased cells and tissues and recognizing abnormalities in the interior of eye from digital images.
• The first FDA approved application of using AI in patient care is for ‘automated identification of eye changes in diabetes.’

AI to identify conditions other than DR from retinal images

• Macular oedema.
• Glaucoma.
• Differentiating individuals with and without type 2 diabetes.
• Predict cardiovascular risk factors.

IoT and diabetes

• Internet of things will revolutionize diabetes management.

• A confluence of DL with human intervention has the ability to provide the best possible care: DL is not limited by fatigue, but humans can recognize ‘out of-set’ variations more easily. (4)
• Ultimately a holistic approach is needed where, apart from technology, one must ‘gather all the key components of clinical care’ if the potential of AI is to be fully met (5)

Future

• A simultaneous broad-based attention must be paid to ethics, regulation and practical applicability. (6)
• While it took more than 20 years for electronic medical records to evolve from an aspiration to a robust clinical platform, AI applications are expected to have a shorter gestation. – Sridhar, Venkat (7)

System Medicine

EDC Model (Endocrine & Diabetes Centre, Visakhapatnam, AP, India)

Link of the Oration

APDF oration 2022

https://drive.google.com/file/d/1FoJphOji7iv4YzDCENP2j4jgf6xw65sP/view

CME INDIA Tail-Piece

• Dr. G. R. Sridhar is one of the senior-most endocrinologists of the country with over 30 years of clinical, research and teaching experience.
• He was the President of the Endocrine Society of India (ESI), President of the Research Society for the Study of Diabetes in India (RSSDI).
• He was the founding editor of Indian Journal of Endocrinology and Metabolism and the Editor in Chief of International Journal of Diabetes in Developing Countries.
• His centre was among the world’s first endocrine centre to establish an in-house electronic medical record system nearly 30 years ago.
• Currently it has a live database of more than 82,000 subjects with endocrine diseases.

References:

1. J Med Internet Res 2022;24:e36010
2. N Engl J Med 375:205-7 (2016)
3. J Am Heart Assoc 9:e013924 DOI: 1161/JAHA,119.013924 (2020)
4. Curr Diab Reports 19:72. https://doi.org/10.1007/s11892-019-1189-3 (2019)
5. Lancet Digital health 2:e8-9 (2020)
6. Lancet 395:1579-86 (2020)
7. Sridhar, Venkat. Indian J Endocrinol Metab 2000; 4:70-80
8. OECD Science, Technology and Industry Working Papers 2020/05. https://dx.doi.org/10.1787/5f65ff7e-en (2020)
9. Sridhar GR et al Int J Diabètes Dev Ctries 31:48-50 (2016)
10. Sridhar GR et al Int J Diabètes Dev Ctries 33:183-5 (2013)
11. Hong. Endocrinol Metab 2020; 35:71-84



Discover CME INDIA





• Explore CME INDIA Repository
• Examine CME INDIA Case Study
• Read History Today in Medicine
• Register for Future CMEs

Get CME INDIA Highlights

Curated articles in your inbox.

SUBSCRIBE!

You have Successfully Subscribed!