Please enable JavaScript.
Coggle requires JavaScript to display documents.
Cancer & AI Diagnosis Tools, Pancreatic Cancer, Cervical Cancer,…
Cancer & AI Diagnosis Tools
Diagnoses
Applications
Image Processing
workflow for training neural networks
1.The input data is required
it needs to be preprocessed or the training will not end in an accurate model
training methods
Levenberg Marquadt Method
parameters
number of neurons
layers which determine the complexity of the mathematical model
Overfitting
activation
function
but usually the sigmoid and
the linear functions are feasible
learning rate
this not necessary to obtain an
accurate model
number of echoes
gradient
validation checks
Artificial neural networks
Neural Networks
Convolution Function
Images
Feature Extraction
Geometry
area,
slope, perimeter, centroid, irregularity index, equivalent
diameter, convex area and solidity.
Statistical
contrast
entropy
RMS
mean and standard devia
tion
Texture
Gray scale level
Tamura
Color
Color Moments
Histogram
Average RGB
by using MATLAB
grey or coloured
The main drawback is that if the
image is resized or its format needs to be changed, from
certain information may be lost
and either more images are required to obtain a good model,
or the model itself meets certain criteria set by the user and
RGB to gray
Layers
input layer
2.convolutional layer
convolution function between two matrices
filter
highlight certain features of an image, or entirely block some off
Dense Layer
The pooling layer
complementary layer
Often 3 × 3 pooling layers
1stbneural network is perceptron
key features
ultrasound imaging with a colonoscopy
thyriod cancer
CNN AND SVMs classify
2 or 3 thyroid nodules are cancerous
too much thyroid hormone and cause hyperthyroidism
malignancy, including hypo-echogenicity, absence of a
halo, micro-calcifications, solidity, intranodular flow
Uses ultrasound imaging
Feature cropping, SVM, GoogLeNet
CNN models trained to classify benign vs. malignant nodules
Types of thyroid cancer
Papillary carcinoma
Affects people under 40 particularly women
Accounts for 8 cases in 10
Follicular carcinoma
tends to affect middle-aged adults, particularly
women
accounts for up to 1 in 10 cases
Medullary thyroid carcinoma
less than 1 in 10 cases
, it can run in families
Anaplastic thyroid carcinoma
the rarest and most seri
ous type, accounting for around 1 in 50 cases
it usually
affects people over the age of 60
How thyroid cancer can be diagnosed
Thyroid Ultra Sound
Laboratory tests
Thyroid Cytology
CT and MRI Scan
Colorectal Cancer (CRC)
Colonoscopy images analyzed for polyp detection
AI improves detection accuracy to 85%-99%
Method Used
CNN
Histopathological analysis
Fourier Transform Spectroscopy
Causes
related to lifestyle and old age
corre
lation between red and processed meat and CRC
alcohol consumption over a large period of time
hereditary
Symtoms
vomiting blood
wight loss and fatigue
even blood in the stools
Operations to train a network
noise removal, image resizing and segmentation.
2100 images used for colonoscopy training
Lung Cancer
AI helps detect nodules in CT scans and ultrasound images
ANN and CNN models trained on large datasets
• Al-assisted diagnosis
reduces false positives
improves early-stage lung cancer detection.
Pancreatic Cancer
• Survival prediction models
• Machine learning techniques
improve risk assessment and treatment planning.
chronic pancreatitis using endoscopic ultrasound images
Cervical Cancer
• CNNs and deep learning techniques
analyze medical imaging data
to detect cancerous lesions
• AI helps optimize radiation therapy plans by identifying organs at risk.
• Image-based classification
using AI
enhances the efficiency of Pap smear tests
Challenges and Future Directions
Challenges
• Data Availability
Large, high-quality datasets are required for training AI models.
• Interpretability
AI models often function as "black boxes," making it difficult to understand how decisions are made
• Integration into Clinical Practice
AI tools must be integrated into existing medical workflows while ensuring compliance with regulations
Future Directions
Future advancements in AI technology will likely address these challenges
Enhanced data sharing
Explainable AI models
Improved computational power will further refine diagnostic accuracy and reduce errors in cancer detection.
Analysing Factors