batch 14.ppt aitam useful ppt for students

Aitam

1. (Approved by AICTE ,Permanently Affiliated to JNTUV , Vizianagaram)
K.KOTTURU,TEKKALI-532201
DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

2. Dugana Lakshmi Prasanna -20A51A0520
Borra Yeshoda Venkata Bhavani -20A51A0512
Pallela Kashyapu -20A51A0529
Yabaji Kiran -20A51A0549
Presented by:
SriTippana.Chalapathi Rao
MTech. Sr.Asst.Professor
Dept of CSE
Under the
esteemed
Guidance of
BRAIN TUMOR DETECTION USING HYBRID DEEP
LEARNING MODELS

3. 01 ABSTRACT
02 INTRODUCTION
03
LITERATURE SURVEY04
PROBLEM IDENTIFICATION
Contents
05 EXISTING SYSTEM AND PROPOSED SYSTEM
06
07
08
09
OVERVIEW AND METHODOLOGY
REQUIREMENTS AND EVALUATION
REFERENCES
PLAN OF ACTION

4. ABSTRACT
Brainisoneofthevitalorgansinthehumanbody.Brain
relateddiagnosisdemandsutmostcareandaminuteerrorin
judgmentmaybedisastrous.Braintumorisoneofthemost
life-threateningdiseasesatitshighestgrade.Generally,
Magneticresonanceimaging(MRI)iswidelyusedimaging
techniquetoassessthetumorsinbrain,liver…etc.Thelarge
amountofdataproducedbyMRIpreventsmanual
classificationinareasonabletime,limitingtheuseofprecise
quantitativemeasurementsintheclinicalpractice.So,itis
notanoptimisticmethodtousetodetectbraintumors.
Hencetrustedandautomaticclassificationschemesare
essentialtopreventthedeathrateofhuman.Onusingdeep
learningmodelslikevgg19,Resnet101,denseNet121Hence
anMRIbraintumorimagewillbeclassifiedaseithera
High-GradeGliomas(HGG)orLow-GradeGliomas(LGG).
Finally,performanceanalysisofthesemodelswillbe
compared.

5. Braintumorsareoneofthedeadliestformsofcancer.Early
detectionandtreatmentcanimproveoutcomesandsavelives.
Theimportanceofearlydetection
Wehavedevelopedahybriddeeplearningmethodthat
combinesseveraltechniquestoimproveaccuracyandreduce
falsepositives.
Ourapproach
Braintumorscanbedifficulttodetectbecausetheyoftenhave
irregularshapesandcanbeinsensitiveareasofthebrain.
ChallengesinBrainTumorDetection
Deeplearningisapowerfultoolforanalyzingmedicalimagesand
detectingabnormalitiesthatmightnotbevisibletothehumaneye.
Theroleofdeeplearning
Introduction

6. ProblemIdentification
•Discoverthecriticalchallengesin
braintumordetectionandthe
limitationsoftraditionalmethods.
•Learnhowhybriddeeplearning
methodshaverevolutionizedthe
accuracyandefficiencyoftumor
identification.

7. Title Authors
Publication
Year Summary
"A MRI based CNN
Approach for Brain
Tumor Image
Detection"
Chattopadhyay, A., & Maitra,
M.
2022
The major goal of this study is to create a
convolutional neural network-based
algorithm for segmenting brain tumors
from 2D magnetic resonance imaging
(MRI) data. This algorithm will be used
in combination with conventional
classifiers and deep learning techniques.
The objective is to develop a very precise
automatic tumor detection technique for
use in medical diagnosis.
The Precision, recall, and F1 score are
performance indicators used to assess the
algorithm's accuracy. These measures
assess how accurately the system can
detect tumors in MRI images while
minimizing false positives or false
negatives.
Literature Survey

8. “Adaptive fuzzy deformable
fusion and optimized CNN
with ensemble classification
for automated brain tumor
diagnosis"
Murthy, M. Y. B.,
Koteswararao, A., & Babu,
M. S
2020
This study's goal is to create a
brand-new brain tumor
classification model based on
sophisticated segmentation and
classification techniques. This
model combines the use of fuzzy
deformable fusion, optimized
convolutional neural networks,
and ensemble classifiers to
accurately identify brain tumors
from MRI images.
The main advantage of this
method is that it can provide more
accurate results than manual
identification by medical
professionals. Additionally, the
use of fuzzy deformable fusion
and optimized convolutional
neural networks allows for a more
efficient segmentation and
classification process.

9. Time-ConsumingProcedures
Learnabouttheextensivetimeandeffort
requiredtoanalyzecomplexmedicalimages
andtheneedformoreefficientsolutions.
ManualDetection
Discoverthecurrentrelianceon
manualinterpretationofmedical
imagesandthedrawbacksof
humanerrorandsubjectivity.
TraditionalAlgorithms
Explorethelimitationsof
conventionalimageprocessingand
machinelearningalgorithmsin
identifyingbraintumorseffectively.
ExistingSystem

10. Hybriddeeplearning
DiscoverthepowerofcombiningVGG19
andResNet101deeplearningmodelsto
enhancebraintumordetectionaccuracy,
precision,andspeed.
AutomatedDiagnosis
Learnhowourproposedsystem
automatestheprocessoftumor
identification,reducinghumaneffort
andminimizingdiagnosticerrors.
ProposedSystem

11. Preprocessing Wepreprocessthemedicalimagesto
enhancecontrastandreducenoise.
Segmentation
Weuseasegmentationalgorithmtoidentify
thetumorregionofinterest.
Classification
WeuseanensembleofCNNsandauto
encoderstoclassifythetumorasbenignor
malignant.
Overview of Hybrid Deep Learning Methods

12. Highqualitymedicalimages
Accuratemedicalimage
analysisrequireshigh-quality
images.Weneedgood
resolutionandcontrastto
identifytumorregions.
ExpertiseinMedical
Imaging
Developingahybriddeep
learningmethod requires
expertiseinbothdeeplearning
andmedicalimaging.
Powerful computing
resources
Deeplearningalgorithms
requirealotofcomputing
power.WeneedpowerfulCPUs
orGPUstotrainandrunour
models.
RequirementsforImplementingHybridDeep
LearningMethodsforBrainTumorDetection
Magneticresonanceimaging
OnceMRIshowsthatthereisa
tumorinthebrain,themost
commonwaytodeterminethe
typeofbraintumoristolookat
theresultsfromasampleof
tissueafterabiopsyorsurgery.

13. SOURCE REQUIREMENTS
Operating system Windows 11-64bit
Programming languages Python
Tools JupyterNotebook, Visual studio code
Libraries Tensorflow, seaborn, pandas, numpy,
cv2, sklearn
Software Requirements

14. Webelievethatourmethodhas
thepotentialtobeusedinclinical
settingstoimprovepatient
outcomes.
ClinicalApplications
Weaimtomakeourmethod
moreaccessibletohealthcare
providersbydevelopinguser-
friendlysoftware.
ImprovedAccessibility
Weplantoexploretheuseof
reinforcementlearninginbrain
tumordetectiontoimprove
decision-making.
FutureResearch
ADVANTAGES

15. DATASET
Training
data
Testing
data

21. References
OnthePerformanceofDeepTransferLearningNetworksforBrainTumorDetectionUsingMRImages
https://ieeexplore.ieee.org/abstract/document/9785791
AHybridDeepLearningModelforBrainTumorClassification
https://doi.org/10.3390/e24060799
Two-phasemulti-modelautomaticbraintumordiagnosissystemfrommagneticresonanceimagesusingconvolutional
neuralnetworks
https://link.springer.com/article/10.1186/s13640-018-0332-4
AVGGNet-BasedDeepLearningFrameworkforBrainTumorDetectiononMRIImages
https://ieeexplore.ieee.org/abstract/document/9515947
Braintumordetectionandmulti-classificationusingadvanceddeeplearningtechniques
https://doi.org/10.1002/jemt.23688
A.B.Hamida,Histogramequalization-basedtechniquesforcontrastenhancementofMRIbraingliomatumorimages:
comparativestudy,in20184thInternationalConferenceonAdvancedTechnologiesforSignalandImageProcessing
(ATSIP),IEEE,2018,pp.1.6.510

22. PLAN OF ACTION
12-10 2023 17-11 2023 10-12-2023 19-01-2024 10-02-2024 15-03-2024
Abstract
Literature survey
Problem
identification
Algorithm
Implementation
Results
Analysis

23. Conclusion
•Inconclusion,thisstudydemonstratestheeffectivenessofdeeplearning
models,specificallyVGG19,ResNet101,andDenseNet121,intheautomatic
classificationofMRIbraintumorimagesintoHigh-GradeGliomas(HGG)
andLow-GradeGliomas(LGG).Theresearchaddressesthecriticalneedfor
accurateandefficientdiagnosisofbraintumors,consideringthelife-
threateningnatureoftheseconditionsandthechallengesposedbythelarge
volumeofMRIdata.Throughrigorousperformanceanalysis,eachmodel's
abilitytoaccuratelyclassifybraintumorimageswasevaluated,sheddinglight
ontheirrespectivestrengthsandweaknesses.Thefindingsofthisstudyoffer
valuableinsightsintothepotentialapplicationsofdeeplearninginmedical
imaging,highlightingopportunitiesforfurtheroptimizationandrefinementof
thesemodelstoenhancediagnosticaccuracyandultimatelyimprovepatient
outcomes.