2. Artificial Intelligence
Artificial intelligence (AI) refers to the
simulation of human intelligence in
machines that are programmed to think
like humans and mimic their actions. The
term may also be applied to any machine
that exhibits traits associated with a
human mind such as learning and
problem-solving.
3. AI for Molecular Imaging
• Artificial intelligence can play a vital role in the
field of medical imaging, especially in diagnostic
perspective.
• From the basic operations like image
registration, filtering, enhancement, calibration,
to advanced algorithm implementation for the
interpretation of data either by supervised or
unsupervised classification AI can greatly help
the physician/researcher.
4. AI for Molecular Imaging
• AI has ability to perform human cognitive
operation more efficiently by minimizing the
errors.
• AI performs imaging operation by integrating
statistical, distributed, computer aided
automation & detection, machine language,
neural network processing approaches for
analysis. Therefore we can get better accuracy
than traditional approaches used in past for
molecular imaging.
5. AI for Molecular Imaging
• Three integral components of the AI
applications are :
• Complex Algorithm,
• Extensive Computational Resources
• Big Data Set (in form of imagery)
6. AI for Molecular Imaging
• Machine learning and Deep Learning are
both types of AI.
• Machine learning is AI that can
automatically adapt with minimal human
interference.
• Deep Learning is a subset of machine
learning that uses artificial neural networks
to mimic the learning process of the
human brain.
7. AI for Molecular Imaging
• Deep learning algorithms categorise
1: supervised : (Consist of an input layer which
simply accepts the input data in its original
dimension, feature extraction layers with
repeating pattern of convolutional operators to
extract the underlying features of the input data
and progressively create higher-order
discriminative features & classification)
2: unsupervised: (In unsupervised algo, the
machine learns from the input dataset itself
without any labels through decoding the inherent
distinctive structures/patterns within the input
data.
8. AI for Molecular Imaging
Architectures for unsupervised deep
learning include
1: Autoencoders
It consists of three major components:
1: Encoder 2: Code /Embedding 2: decoder
2: Generative Adversarial Networks
(GANs).
1: Discriminator 2: Generator
12. AI for Molecular Imaging
Conclusion
• AI in molecular imaging context can minimize
the clinician’s efforts and furnish speedy, quick
and highly précised diagnostic results.
• AI has potential to figure out the human
shortcomings by minimizing the radiation/
chemical exposure risks & accurately handling
the entire image analysis procedure.
13. AI for Molecular Imaging
Conclusion
• AI-based approaches have exhibited
adequate accuracy and robust
performance to be employed as an
alternative or complementary resource to
conventional tools in clinical practice
• Failure of AI-based approaches can be
verified/corrected by existing conventional
tools.
14. AI for Molecular Imaging
Conclusion
The performance of AI algorithms depends
largely on the training data used for model
development.
• Data collection, a critical step in AI solutions, is
one of the major challenges faced by
developers. Though there are uncountable
numbers of clinical databases around the globe,
many of them are not valid or properly annotated
as required by learning systems.
15. References
• Arabi Hossein, Zaidi Habib (2020) Application of Artificial
Intelligence and Deep Learning in Molecular Imaging
and Radiotherapy. European Journal of Hybrid Imaging.
• Porenta Gerold(2019) Is there Value for Artificial
Intelligence Application in Molecular Imaging and
Nuclear Medicine? Journal of Nuclear Medicine.
2019;60(10):1347–1349. doi:
10.2967/jnumed.119.227702
• Sollini Martina (2020), Artificial Intelligence and Hybrid
Imaging : The Best Match for Personalized Medicine in
Oncology, European Journal of Hybrid Imaging 2020
Dec 9;4(1):24. doi: 10.1186/s41824-020-00094-8.