Abstract
Advancement in the medical field promotes the diagnosis of disease through
automation methods and prediction of the brain tumor also plays an important role due
to the fact that millions of people are affected by brain tumor and the rate of affected
people is increasing every year randomly. Hence, in saving the lives of many
individuals, the early detection of the disease plays an important role. Using the MRI
Images, it’s easy to find the location and existence of the tumor. Expert manual
diagnosis is playing a vital role in detecting the information about the tumor and its
type. Though there are various models that can detect tumor location with the help of
ML models in the medical field, somewhere there is a lag in the success of these
models. Deep learning is one of the widely used approaches for the same. But the
black-box nature of these machine-learning models has somewhat limited their clinical
use. Explanations are essential for users to know, trust, and well manage these models.
The chapter proposes dual-weighted deep CNN classifiers for early prediction of the
presence of brain tumor along with the explanation-driven DL models such as Local
Interpretable Model-agnostic Explanations (LIME) and SHapley Additive explanation
(SHAP). The performance and accuracy of the planned model are assessed and relate
with the existing models and it is expected that it will produce high sensitivity as well
as specificity. It is also expected to perform well by means of precision and accuracy.