Transforming Healthcare with Intelligence: AI‑Driven Diagnostics

and the Evolution of Personalized Medicine

K. Sathiyamurthi, A. Sundar Raj, N. Ragavi and M. Jeevadharshini

Department of Biomedical Engineering, E.G.S. Pillay Engineering, College, Nagapattinam, Tamil Nadu, India

Keywords: Google’s Teachable Machine, AI, MRI, Brain Tumour, Diagnostic Tools.

Abstract: Brain tumors are an important health challenges globally, with substantial consequences for individuals and

health-care systems. This Automated Brain Tumour Detection proposes to make use of Googles "Teachable

Machine" Tool To lead to a breakthrough in medical imaging and diagnostics based on machine learning and

artificial intelligence. After training the "Teachable Machine" with various MRI images (tumour and non-

tumor), this model can be converted to a user interface for doctors to submit MRI and get automatic answers

regarding the presence of tumours in the scan instantaneously. The potential impact of this research is

enormous in sufficiently improving the efficiency and accuracy of brain tumour diagnosis. The process remains

time-consuming to interpret the MRI image and there is high chance of error based on the human subjectivity

unlike the proposed automated detection technique which will allow early intervention and improved outcome

of the patient. In addition, by using the "Teachable Machine" platform provided by Google, they are affordable

and use a small amount of processing power and do not require any specialized hardware. The democratization

of sophisticated technical tools can be a boon to healthcare systems around the world.

1 INTRODUCTION

Manual interpretation of medical imaging for brain

tumor diagnosis is difficult and time-consuming,

highlighting the importance of developing new

methods to enhance detection efficiency and

accuracy. To address this urgent need, this project

aims to promote brain health through an automated

brain tumor detection system using Google's

"Teachable Machine" tool. The fusion of medical

imaging with machine learning is setting the stage for

a tectonic shift in diagnostic paradigms. The proposed

project aims to leverage the power of machine

learning algorithms to significantly impact the

diagnostic process by achieving fast and accurate

detection of brain tumors from magnetic resonance

imaging (MRI) scans. This is not just impressive in

terms of how it could improve upclinical workflows,

however. Emphasizing the importance of early

intervention in brain tumorDetection, it will enhance

lesion detection and all the related aspects.

Automating the detection process allows healthcare

practitioners to make the diagnostics process faster,

development of more efficient treatment strategies,

and improve health results. Additionally, the user-

friendly and readily available nature of Google

“Teachable Machine” empowers clinicians from all

specialties to apply machine learning to health care,

paving the way for innovation and collaboration in

various health care environments. The purpose of

this project is to automate the detection of brain

tumors, which should be a major advancement in

brain health. The project's goal is to reduce disparities

in healthcare by making it easier to detect tumors. To

ensure high detection accuracy as an early detection

system, as misdiagnosis and missed diagnosis must

be avoided; the project aims to help with tuning those

machine learning models so that they perform with

utmost efficiency. An intuitive interface will be

created to simplify the process, enabling healthcare

professionals to upload medical images and receive

quick and dependable tumor detection conclusions.

This project is grounded in work to research

automated tumor detection, towards better patient

outcomes and brain health.

2 EXISTING TECHNIQUES

Traditional methods like MRI, CT, and PET scans

have been pivotal in brain tumor detection, but they

686

Sathiyamurthi, K., Raj, A. S., Ragavi, N. and Jeevadharshini, M.

Transforming Healthcare with Intelligence: AI-Driven Diagnostics and the Evolution of Personalized Medicine.

DOI: 10.5220/0013903900004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 3, pages

686-692

ISBN: 978-989-758-777-1

often necessitate specialized interpretation and may

not be universally accessible. Machine learning

algorithms, especially Convolutional Neural

Networks (CNNs), have emerged as powerful tools in

medical image analysis. These algorithms can learn

intricate patterns from images, aiding in the

automated detection of abnormalities. Deep learning

approaches, such as deep neural networks, further

enhance the capabilities of these algorithms by

enabling them to extract hierarchical features from

images. Figure 1 shows the MRI scanner. Figure 2

shows the MRI scan results.

Figure 1: MRI scanner.

Figure 3 shows the Typical CT Scanner and Figure 4 and

5 shows the CAT scan results and PET scanner. Figure 6

shows the PET scan results.

Figure 2: Mri Scan Results.

Figure 3: Typical CT scanner.

Figure 4: CAT scan results.

Figure 5: PET scanner.

Figure 6: PET scan results.

3 PROPOSED TECHNIQUES

The proposed system in "Advancing Brain Health:

Automated Brain Tumour Detection Using Google's

Teachable Machine Tool" utilizes Google's

Teachable Machine to enhance the process of brain

tumor detection. The System combines deep learning

models with medical imaging data to facilitate brain

tumor detection. The new system combines

Teachable Machine's simple user interface and

powerful training capability to allow intuitive image

analysis and fast tumor detection results to be

Transforming Healthcare with Intelligence: AI-Driven Diagnostics and the Evolution of Personalized Medicine

687

obtained by all healthcare professionals. Its large

structure utilizes deep learning methods to provide

unbiased and stable predictions for incentive of early

recognition and better treatment of patients. The

system retains the ability to learn through multiple

feedback loops, constantly perusing and improving

performance in line with the new technology in

medical imaging. This is a really exciting step

forward in brain health as a whole.

3.1 Hardware Descriptions

3.1.1 High-Performance Computing (HPC)

System

To deal with this large volume of medical imaging

data, a powerful HPC system is necessary. The

hardware requirements will need a multi-core CPU,

proper RAM (Random Access Memory) and GPU

(Graphics Processing Unit) to efficiently accelerate

the machine learning algorithms that drive tumor

detection.

3.1.2 Medical Imaging Equipment

Ironically, MRI (Magnetic Resonance Imaging) and

CT (Computed Tomography) scanners are critical to

acquiring images of our brain in detail. These

imaging modalities provide the critical data input that

is used both for training and testing the automated

tumor detection model. Figure 7 shows Example of

MRI.

Figure 7: Example of MRI images that have been collected

for data acquisition.

3.1.3 Data Storage Solution

Both MRI and CT scanning create massive amounts

of medical imaging data. Therefore, a robust data

storage solution is required. These may include high-

capacity hard disk drives (HDDs) or solid-state drives

(SSDs) and may also include a dedicated backup

system to guarantee data integrity and accessibility.

3.1.4 Peripheral Devices

HPC systems require input peripherals, like

keyboard, mouse and monitors, to operate and

visualize the results obtained from tumor detection

algorithms. In addition, high-resolution displays are

useful for the precise examination of medical

images.

3.1.5 Network Infrastructure

Moreover, a strong network architecture is necessary

to ensure that there is connectivity between the HPC

system, medical imaging devices and other project

elements. Use of high-speed Ethernet connections or

dedicated fiber-optic links may be necessary for

expeditious transfer of large datasets. Figure 8 shows

the Network Infrastructure.

Figure 8: Network infrastructure.

3.1.6 Power Backup and Conditioning

To maintain data integrity and ensure continued

operation during power failures or fluctuations,

Uninterruptible Power Supply (UPS) systems are

crucial.

3.1.7 Webcam

The major hardware component used in the project is

the webcam to take visual data in real time during

neurological checks. Such integration improves

diagnostic capabilities for healthcare professionals by

providing information, supplementing existing

imaging modalities, which may facilitate timely

detection and intervention.

3.2 Software Descriptions

The software portion of the project includes different

software tools and applications necessary for training,

testing and deploying machine learning models to

detect brain tumors. The following is a complete

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

688

exposition of the software pieces that make up the

project.

3.2.1 Google's Teachable Machine Tool

This is trained on Google’s Teachable machine tool

which is the basic software platform for building and

deploying machine learning models without

extensive programming skill. It also allows users to

easily build their own image classification models

based on convolutional neural networks (CNNs) with

minimal effort. Figure 9 shows the Teachable

Machine. Figure 10 shows the Steps of preparing an

image processing project.

Figure 9: Getting started with teachable machine.

Figure 10: Steps of preparing an image processing project.

Figure 11: Different types of projects that can be

undertaken with TM.

Use Python for scripting and implementing your

custom algorithms into the Google Teachable

Machine Tool, Data Preprocessing. Figure 11 shows

Different types of projects that can be undertaken

with TM.

3.2.3 Data Preprocessing Tools

Medical imaging data from MRI and CT scans are

pre-processed for enhancing quality and feature

extraction before training the machine learning

models. This may include image resizing,

normalization, denoising, and segmentation using

SimpleITK or DICOM (Digital Imaging and

Communications in Medicine) libraries.

Figure 12: Axial T2 - MRI image of a normal brain.

Figure 13: Saggital T1 - MRI image of a normal brain.

3.2.4 Model Evaluation and Validation

Tools

Show using software tools how trained machine

learning models are evaluated using metrics including

accuracy, sensitivity, specificity, and area under the

ROC curve. By using cross-validation techniques and

confusion matrices, we can evaluate the model's

generalization capability and bias. Figure 12 shows

Axial T2 – MRI Image. Figure 13 MRI normal brain.

3.2.5 User Interface (UI) Design and

Visualization Tools

Interactive interfaces provide a means for healthcare

professionals to engage with the automated tumor

detection system, uploading medical images for real-

time detection initiation and result visualization.

Transforming Healthcare with Intelligence: AI-Driven Diagnostics and the Evolution of Personalized Medicine

689

Figure 14: Various customizable parameters in teachable

machine.

4 RESULTS AND DISCUSSIONS

So far, the project has yielded promising results,

indicating that machine learning techniques have the

potential to transform brain health care. The

proposed model had been trained and validated with

a large dataset of medical imaging scans, and had

achieved high levels of accuracy and sensitivity in

detecting brain tumors through improved processing

techniques. By incorporating Google's Teachable

Machine Tool, the process flow becomes simple,

allowing healthcare workers to upload images and

receive fast, accurate diagnoses. Nevertheless, there

are aspects like dataset bias, interpretability of learnt

features, and deployment in real world that need

exploration and refinement. Certainly, this project is

a cornerstone towards the application of artificial

intelligence in brain health care and highlights the

need for interdisciplinary intersection of technology

and healthcare. Figure 15 shows the CASE 1 of the

machine correctly determining he diagnosis of a

tumor by using an MRI scan and Figure 16

shows

CASE-2 of the machine correctly determining the