Page 224 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

214 Chapter 7 Early detection and diagnosis using deep learning

The human brain is a complex web of networks. Advances in
functional MRI, a type of imaging that measures brain activity
by detecting changes in blood flow, have aided in the linking
of connections within and between the networks in the brain.
Brain MRI is believed to have a potential role in the diagnostic
process as research has been conducted indicating that ADHD
results from a type of breakdown or disruption in this sophisti-
cated brain map. Although a lot of models primarily focus on the
single-scale approach, research has been conducted for the
advancement of the multiscale approach, which uses multiple
brain maps based on multiple parcellations. Brain parcellations
can be defined based on anatomical criteria, functional criteria,
or both, and hence, the brain can be studied to extreme extents
based on distinct brain parcellations. The success of this
approach evidently proves that the DL-based approach for
early-stage disease detection has potential even beyond ADHD.

4. Conclusion and further advancements

Contemporary developments in AI offer a thrilling chance to
advance healthcare. Nevertheless, the conversion of research
methods to active medical placement offers an original limit
for scientific and ML investigation. With vigorous potential,
medical assessment will be indispensable to guarantee that AI
structures are innocuous and operative, by means of clinically
valid presentation metrics, which goes out of procedures of
methodological correctness to contain how AI marks the excel-
lence of maintenance, the capriciousness of healthcare special-
ists, the efficacy and production of medical preparation, and
most prominently, persistent consequences. Self-governing
data sets that are illustrative of upcoming goal populaces should
be curated to allow the contrast of dissimilar algorithms, though
prudently appraising for symbols of potential bias and correctly
fitting to unintentional confounders. Creators of AI tools should
essentially be conscious of the probable inadvertent penalties of
the algorithms they will create and guarantee that algorithms
must be considered with the universal community in attention.
Additional effort to advance the effectiveness of algorithms must
be done, and comprehending the humanealgorithm communi-
cations is also necessary at the same time, which will be indis-
pensable to their upcoming implementation and security
reinforced by the expansion of considerate governing outlines.