The traditional method of detecting brain tumors is through costly medical scans,
advanced equipment, and trained radiologists, making it time-consuming and inaccessible to
the masses. NeuroAid is a cross-platform mobile application designed using Flutter to
provide a machine-learning-based platform for brain tumor detection. Users can upload the
MRI scans and employ a TensorFlow Lite model to scan the same and identify whether brain
tumors are present and classify them into three categories: Glioma, Meningioma, and
Pituitary tumors.The software can be used both by medical doctors and patients seeking to
perform a preliminary self-screening. Equipped with real-time processing, user-friendly
design, and a chatbot offering instant advice, NeuroAid offers an inexpensive, effective, and
simple way of early diagnosis, aiding in medical decision-making.

1.1 Project Overview
NeuroAid is built as an independent and efficient ML-based brain tumor diagnostic
tool that scans MRI images to facilitate early diagnosis. Flutter-based mobile app allows
uploading MRI images, which are passed through a TensorFlow Lite model to identify the
type of tumor, i.e., Glioma, Meningioma, or Pituitary tumors. Backend operations, like secure
authentication of users, are performed via Firebase to ensure data secrecy and availability.
Front-end development is performed under Flutter for the purpose of offering an interactive
and user-friendly experience. There is also live support provided by a Gemini API-based
chatbot incorporated in the application, with the users being provided with immediate
information. Giving priority to accessibility and efficiency, NeuroAid is a cost-effective and
trustworthy device for the early detection of brain tumors.

1.2 Project Objective
The main goal of the NeuroAid project is developing a mobile application for the
early diagnosis and classification of brain tumors based on machine learning on a user-
friendly interface. Users will be able to upload their MRI scans through the app, and they will
be processed with an embedded TensorFlow Lite model specifically trained to identify
Glioma, Meningioma, and Pituitary tumors. The DL model not just detects the tumor but also
includes a confidence score, which notifies users about the results' accuracy. It is an
automated system for detection that tries to aid doctors, radiologists, and patients by
providing quick and precise initial diagnoses.

1.3 Module Description
User Authentication Module
The application utilizes Firebase Authentication for performing secure password and
email login with good security in addition to session management in a managed way. This
offers protection of the patients' confidentiality as well as safe access to their medical history
without being required to repeatedly log in.

Deep Learning-Based Brain Tumor Detection Module:
This module processes MRI scans using the DenseNet121 model to identify and
classify Glioma, Meningioma, and Pituitary tumors. It also enables rapid processing
on mobile devices for real-time results.The model is tuned using the Optuna
hyperparameter tuning technique and then converted into a TFLite model for efficient
deployment.

Chatbot Integration Module
This Module takes the help of Google Gemini API for real-time suggestions on tumor
detection in brain, MRI scans, and general health queries. It gives prompt recommendations
to guide users appropriately.

Frontend User Interface Module
This Module is built using Flutter, which provides Android and iOS users
with a responsive, fluid, and cross-platform experience. It is easy to navigate , having
a sleek interface, and supports easy traversing from MRI scan classification, chatbot
interaction, and setting up accounts.

Backend Firebase Module
This Module manages user authentication through Firebase Authentication with email
and password login. It provides secure access and session management, so users remain
logged in without re-authentication issues.

Software Description
Flutter Framwork
NeuroAid's user interface, developed using Flutter, provides native cross-platform
support for iOS and Android. It provides quick rendering, code sharing, UI features of today,
and third-party integration. Navigation, uploading an MRI scan, displaying results, and
interaction with a chatbot are all enabled.

DenseNet121 Model
DenseNet121 is a deep CNN that enhances feature reuse and reduces parameters. Pre-trained
on ImageNet, it efficiently classifies Glioma, Meningioma, and Pituitary tumors from MRI
scans. Optuna hyperparameter tuning optimizes performance, and the model is converted to
TFLite for real-time mobile deployment on Android and iOS. It provides fast, accurate
diagnoses with a confidence score, aiding medical decision-making and reducing reliance on
professionals.

Firebase Authentication
The backend of NeuroAid provides user authentication via Firebase Authentication
for secure login, session handling, and data protection.

Chatbot – Google Gemini API Integration
Google Gemini API-based NeuroAid chatbot offers real-time support by answering
questions related to brain tumors and giving instant context-specific answers. It is web-based,
passing user input to Gemini API for rapid and informative responses to MRI scan reports.