Lung cancer Detection with AIML in Healthcare sector.PPT
jayasankaanushan199
83 views
31 slides
Feb 28, 2025
Slide 1 of 31
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
About This Presentation
🚀 AI & Machine Learning in Lung Cancer Detection 🏥đź§
This slides delves into the role of AI and Machine Learning in revolutionizing lung cancer detection within the healthcare sector. It specifically explores the widely used Convolutional Neural Network (CNN) architecture, breaking down...
Slide Content
Lung cancer
Detection
with AI/ML in
Healthcare
sector.
By Jayashanka Anushan
Detection
with AI/ML in
Healthcare
sector.
By Jayashanka Anushan
Lung cancer is a type of cancer
that starts when abnormal cells
grow in an uncontrolled way in
the lungs. It is a serious health
issue that can cause severe
harm and death.
(World Health Organization, 2023)
What is Lung Cancer? What it is looks
like.
that starts when abnormal cells
grow in an uncontrolled way in
the lungs. It is a serious health
issue that can cause severe
harm and death.
(World Health Organization, 2023)
What is Lung Cancer? What it is looks
like.
Cause of lung
cancer...
•Smoking is the leading cause of lung cancer,
responsible for approximately 85% of all
cases.
•Lung cancer is often diagnosed at advanced
stages when treatment options are limited.
•Screening high risk individuals has the
potential to allow early detection and to
dramatically improve survival rates.
•Primary prevention (such as tobacco control
measures and reducing exposure to
environmental risk factors) can reduce the
incidence of lung cancer and save lives.
cancer...
•Smoking is the leading cause of lung cancer,
responsible for approximately 85% of all
cases.
•Lung cancer is often diagnosed at advanced
stages when treatment options are limited.
•Screening high risk individuals has the
potential to allow early detection and to
dramatically improve survival rates.
•Primary prevention (such as tobacco control
measures and reducing exposure to
environmental risk factors) can reduce the
incidence of lung cancer and save lives.
Latest Statics...
Lung Cancer Facts and Statistics –
Worldwide 2022
•Globally, Lung cancer is the world’s
leading dominant cancer with the
highest incidence and mortalities
compared to other cancer types
(accounting for 12.4% of all cancer
cases, and 18.7% of all cancer deaths).
•By 2022, about 2.4 million new cases
occurred and 1.8 million people died
from lung cancer worldwide.
Lung Cancer Facts and Statistics –
Abu Dhabi 2018 - 2019
•Ranked the second most deadly
cancer in the Emirate of Abu Dhabi,
accounting for 9.4% of all cancer
deaths which is the second highest
cancer mortality following breast
cancer (2018)
•The sixth most common diagnosed
cancer in the Emirate, accounting for
3.9% of all cancer cases (2019)
(Abu Dhabi Public Centre - www.adphc.gov.ae)
Lung Cancer Facts and Statistics –
Worldwide 2022
•Globally, Lung cancer is the world’s
leading dominant cancer with the
highest incidence and mortalities
compared to other cancer types
(accounting for 12.4% of all cancer
cases, and 18.7% of all cancer deaths).
•By 2022, about 2.4 million new cases
occurred and 1.8 million people died
from lung cancer worldwide.
Lung Cancer Facts and Statistics –
Abu Dhabi 2018 - 2019
•Ranked the second most deadly
cancer in the Emirate of Abu Dhabi,
accounting for 9.4% of all cancer
deaths which is the second highest
cancer mortality following breast
cancer (2018)
•The sixth most common diagnosed
cancer in the Emirate, accounting for
3.9% of all cancer cases (2019)
(Abu Dhabi Public Centre - www.adphc.gov.ae)
How to identify
lung cancer?
Imaging tests Sputum cytology
Biopsies
•X-rays – Chest x-rays
•CT scans
•Positron Emission
Tomography (PET) scan
•The mucus that is
coughed up from the
lungs.
•Remove a sample of tissue
for testing in a lab.
lung cancer?
Imaging tests Sputum cytology
Biopsies
•X-rays – Chest x-rays
•CT scans
•Positron Emission
Tomography (PET) scan
•The mucus that is
coughed up from the
lungs.
•Remove a sample of tissue
for testing in a lab.
Use of AI/ML
for lung
cancer
detection.
•Deep Learning - convolutional neural networks
(CNNs)
1. These models are trained on thousands of annotated images
to detect patterns and abnormalities associated with lung
cancer, such as nodules, tumors, and other suspicious lesions.
2. AI systems can identify subtle differences in images that may
be challenging for human radiologists to spot, even in early-
stage cancers.
•Automated Nodule Detection
1. AI-based systems automatically detect these nodules, segment
them, and classify them based on size, shape, and density,
which helps identify cancer at earlier stages.
2. AI models can flag potential malignancies for further
evaluation, reducing human error and improving the accuracy
of diagnoses.
(Shah, et al., 2023)
for lung
cancer
detection.
•Deep Learning - convolutional neural networks
(CNNs)
1. These models are trained on thousands of annotated images
to detect patterns and abnormalities associated with lung
cancer, such as nodules, tumors, and other suspicious lesions.
2. AI systems can identify subtle differences in images that may
be challenging for human radiologists to spot, even in early-
stage cancers.
•Automated Nodule Detection
1. AI-based systems automatically detect these nodules, segment
them, and classify them based on size, shape, and density,
which helps identify cancer at earlier stages.
2. AI models can flag potential malignancies for further
evaluation, reducing human error and improving the accuracy
of diagnoses.
(Shah, et al., 2023)
Deep Learning - Neural Network
•Connections - links between neurons
•Weights and Biases - measure the
strength and influence of connection
•Propagation Functions - mechanism to
process and transfer data across layers.
•Learning Rule - Method that adjust
weight and biases over time to improve
accuracy.
Neural Network is a deep learning network architecture, this mimic the complex functions
in human brain. This model consist of nodes or neurons that process data and learn the
patterns. Neurons - input signal (Geeksforgeeks, 2024)
•Connections - links between neurons
•Weights and Biases - measure the
strength and influence of connection
•Propagation Functions - mechanism to
process and transfer data across layers.
•Learning Rule - Method that adjust
weight and biases over time to improve
accuracy.
Neural Network is a deep learning network architecture, this mimic the complex functions
in human brain. This model consist of nodes or neurons that process data and learn the
patterns. Neurons - input signal (Geeksforgeeks, 2024)
CNN Model (convolutional neural
networks)
A Convolutional Neural Network (CNN) is a type of Deep learning neural network
architecture use in compiuter vision. This model use for image clasification. This is a
extended version of artificial neural network (ANN) which is used to extract the feature
from grid (matrix dataset).
CNN Architecture
Convolutional Neural Network
consists of multiple layers like
the input layer, Convolutional
layer, Pooling layer, and fully
connected layers.
networks)
A Convolutional Neural Network (CNN) is a type of Deep learning neural network
architecture use in compiuter vision. This model use for image clasification. This is a
extended version of artificial neural network (ANN) which is used to extract the feature
from grid (matrix dataset).
CNN Architecture
Convolutional Neural Network
consists of multiple layers like
the input layer, Convolutional
layer, Pooling layer, and fully
connected layers.
Key features of CNN architecture
These layers apply convolutional operations to the input image using filters (kernels).
Each filter extracts specific features (edges, textures, patterns) from the image by
reducing the complexity of the input while retaining important features.
1. Convolutional Layers
Key Properties:
•Filters/Kernels: Learnable weights that are
used to slide over the input data.
•Local Receptive Field: The region of the
input data a filter sees at a time.
•Stride: Controls how the filter moves over
the input data.
•Padding: Adds extra pixels to the input to
ensure the filter fits well along the edges.
These layers apply convolutional operations to the input image using filters (kernels).
Each filter extracts specific features (edges, textures, patterns) from the image by
reducing the complexity of the input while retaining important features.
1. Convolutional Layers
Key Properties:
•Filters/Kernels: Learnable weights that are
used to slide over the input data.
•Local Receptive Field: The region of the
input data a filter sees at a time.
•Stride: Controls how the filter moves over
the input data.
•Padding: Adds extra pixels to the input to
ensure the filter fits well along the edges.
After each convolution, an activation function like ReLU (Rectified Linear Unit) is applied
to introduce non-linearity. This allows the model to learn complex patterns in the data.
2. Activation Function (ReLU)
ReLU:
ReLU: Replaces all negative values with zero, helping with sparsity and
reducing computational complexity.
to introduce non-linearity. This allows the model to learn complex patterns in the data.
2. Activation Function (ReLU)
ReLU:
ReLU: Replaces all negative values with zero, helping with sparsity and
reducing computational complexity.
Pooling layers reduce the spatial dimensions (height and width) of the feature maps
produced by convolutional layers, which helps to reduce the computational load, control
overfitting, and provide a form of translation invariance.
3. Pooling Layers (Subsampling or Downsampling)
Types:
•Max Pooling: Selects the maximum value in each region of the feature map.
•Average Pooling: Computes the average value of each region.
•Sum Pooling: Computes the sum of all values in each region of the feature map.
Pooling Window:
•A region of the feature
map over which pooling is
performed.
produced by convolutional layers, which helps to reduce the computational load, control
overfitting, and provide a form of translation invariance.
3. Pooling Layers (Subsampling or Downsampling)
Types:
•Max Pooling: Selects the maximum value in each region of the feature map.
•Average Pooling: Computes the average value of each region.
•Sum Pooling: Computes the sum of all values in each region of the feature map.
Pooling Window:
•A region of the feature
map over which pooling is
performed.
After multiple convolutional and pooling layers, CNNs use fully connected layers to
interpret the high-level features and make final predictions.
4. Fully Connected Layers (Dense Layers)
•The final fully connected layer typically uses a softmax or sigmoid activation function
to output probabilities for classification tasks.
interpret the high-level features and make final predictions.
4. Fully Connected Layers (Dense Layers)
•The final fully connected layer typically uses a softmax or sigmoid activation function
to output probabilities for classification tasks.
Normalization layers, such as Batch Normalization, help stabilize and speed up the
training process. They normalize the output of a previous activation layer by adjusting
the mean and variance.
5. Normalization Layers (Batch Normalization)
Benefit:
Helps prevent issues like
vanishing/exploding gradients and
reduces overfitting.
training process. They normalize the output of a previous activation layer by adjusting
the mean and variance.
5. Normalization Layers (Batch Normalization)
Benefit:
Helps prevent issues like
vanishing/exploding gradients and
reduces overfitting.
Dropout is a regularization technique where random units (neurons) in the network are
"dropped" or deactivated during training. This prevents the network from overfitting by
forcing it to rely on multiple paths for learning.
6. Dropout
"dropped" or deactivated during training. This prevents the network from overfitting by
forcing it to rely on multiple paths for learning.
6. Dropout
In CNNs, the same filter (weights) is applied to different parts of the image, meaning
that the learned features are translation-invariant. This significantly reduces the number
of parameters, making the model more efficient.
7. Weight Sharing
that the learned features are translation-invariant. This significantly reduces the number
of parameters, making the model more efficient.
7. Weight Sharing
CNNs can be trained end-to-end, meaning that the feature extraction and classification
tasks are learned simultaneously during training, eliminating the need for manual
feature engineering.
8. End-to-End Learning
tasks are learned simultaneously during training, eliminating the need for manual
feature engineering.
8. End-to-End Learning
In CNNs, lower layers typically learn simple features like edges and textures, while
higher layers combine these features to learn more complex structures like shapes or
objects.
9. Hierarchical Feature Learning
higher layers combine these features to learn more complex structures like shapes or
objects.
9. Hierarchical Feature Learning
CNNs exploit the spatial relationships in data (e.g., in images), maintaining the spatial
structure of the data throughout the layers. The lower layers capture fine-grained
details, and the higher layers capture more abstract features.
10. Spatial Hierarchy
structure of the data throughout the layers. The lower layers capture fine-grained
details, and the higher layers capture more abstract features.
10. Spatial Hierarchy
Real world applicate
where use AI/ ML
where use AI/ ML
Spire Health Tag (Breathing
Monitor)
•This is a wearable device and monitoring
breathing pattern. Specially designed for
lung cancers. Detect abnormal breathing
pattern that could be indicative of
respiratory conditions, including lung
cancer. There is a mobile app analyze data
and give recommendations.
•Uses a smart fabric sensor to track
breathing patterns, heart rate, and
activity levels. It uses machine learning to
analyze changes in these patterns and
send alerts when it detects irregularities.
(Spirehealth, 2024)
Monitor)
•This is a wearable device and monitoring
breathing pattern. Specially designed for
lung cancers. Detect abnormal breathing
pattern that could be indicative of
respiratory conditions, including lung
cancer. There is a mobile app analyze data
and give recommendations.
•Uses a smart fabric sensor to track
breathing patterns, heart rate, and
activity levels. It uses machine learning to
analyze changes in these patterns and
send alerts when it detects irregularities.
(Spirehealth, 2024)
Monarch™ Platform by Auris Health
•This is a robotic-assisted, minimally invasive
system which is enables physicians to navigate
deep into the lungs using a combination of
robotics, flexible endoscopy, and advanced
imaging.
•Use a hand held controller which have a camera at
the end. Physicians can control the robotically
controlled endoscope deep into the lung for see
the real nodules for examine. Real time vision
feedback.
•System create a 3D map of using a pre-procedure
CT scan. It then uses electromagnetic tracking to
navigate the target lesion.
•When reach to the destination, biopsy tool use to
collect tissue sample.
(Businesswire, 2018)
•This is a robotic-assisted, minimally invasive
system which is enables physicians to navigate
deep into the lungs using a combination of
robotics, flexible endoscopy, and advanced
imaging.
•Use a hand held controller which have a camera at
the end. Physicians can control the robotically
controlled endoscope deep into the lung for see
the real nodules for examine. Real time vision
feedback.
•System create a 3D map of using a pre-procedure
CT scan. It then uses electromagnetic tracking to
navigate the target lesion.
•When reach to the destination, biopsy tool use to
collect tissue sample.
(Businesswire, 2018)
Novel wearable vest for long-term
lung monitoring
•The WELMO vest, developed by CSEM, is a
wearable device for long-term lung
monitoring.
•It uses electrical impedance tomography
and has 18 embedded sensors that can
record chest sounds and thoracic electrical
bioimpedance.
•It is designed for monitoring respiratory
diseases.
(csem.ch, 2022)
lung monitoring
•The WELMO vest, developed by CSEM, is a
wearable device for long-term lung
monitoring.
•It uses electrical impedance tomography
and has 18 embedded sensors that can
record chest sounds and thoracic electrical
bioimpedance.
•It is designed for monitoring respiratory
diseases.
(csem.ch, 2022)
E-nose (Electronic Nose) for Lung
Cancer Detection
•The e-nose is a device that uses an array of
sensors to detect volatile organic compounds
(VOCs) in the breath, which can act as
biomarkers for diseases like lung cancer.
•Use Metal Oxide Semiconductor (MOS) Sensors,
Conducting Polymer (CP) Sensors, Quartz
Crystal Microbalance (QCM) Sensors etc. The
data collected is processed using machine
learning algorithms to identify the chemical
signatures associated with lung cancer.
(Parker, 2024)
Cancer Detection
•The e-nose is a device that uses an array of
sensors to detect volatile organic compounds
(VOCs) in the breath, which can act as
biomarkers for diseases like lung cancer.
•Use Metal Oxide Semiconductor (MOS) Sensors,
Conducting Polymer (CP) Sensors, Quartz
Crystal Microbalance (QCM) Sensors etc. The
data collected is processed using machine
learning algorithms to identify the chemical
signatures associated with lung cancer.
(Parker, 2024)
Challenges in
AI Development
for cancer
detection
Data Quality Interpretability
Security
AI systems heavily rely on quality data for
training and decision-making.
Understanding how AI systems arrive at their
decisions is crucial for trust and accountability.
Risks of AI systems being exploited or
manipulated for malicious purposes.
AI Development
for cancer
detection
Data Quality Interpretability
Security
AI systems heavily rely on quality data for
training and decision-making.
Understanding how AI systems arrive at their
decisions is crucial for trust and accountability.
Risks of AI systems being exploited or
manipulated for malicious purposes.
Future
Trends
•Wearable AI/ML for Lung Cancer Detection: No
dedicated device exists yet, but integrating wearables,
IoT, AI, and ML could enable real-time monitoring of
biomarkers (e.g., gas analysis, oxygen levels, breathing
patterns) for early detection. A compact, portable
device could be embedded in mobiles or smartwatches.
•Multi-Sensor Integration: Future devices may combine
gas sensors, breath analyzers, and heart rate monitors
for continuous health tracking, detecting lung cancer or
other serious conditions.
Trends
•Wearable AI/ML for Lung Cancer Detection: No
dedicated device exists yet, but integrating wearables,
IoT, AI, and ML could enable real-time monitoring of
biomarkers (e.g., gas analysis, oxygen levels, breathing
patterns) for early detection. A compact, portable
device could be embedded in mobiles or smartwatches.
•Multi-Sensor Integration: Future devices may combine
gas sensors, breath analyzers, and heart rate monitors
for continuous health tracking, detecting lung cancer or
other serious conditions.
Conclusion...
•Lung cancer is a leading cause of death; early
detection improves survival.
•AI/ML, particularly CNNs, enhance lung cancer
detection by analyzing medical images and
reducing human error.
•Wearable devices and IoT enable continuous
monitoring of lung health.
•Real-time monitoring of biomarkers aids in early
cancer detection.
•Future advancements will integrate multiple
sensors into compact devices like smartwatches.
•AI/ML and wearables offer promising solutions,
despite challenges with data quality and security.
•Lung cancer is a leading cause of death; early
detection improves survival.
•AI/ML, particularly CNNs, enhance lung cancer
detection by analyzing medical images and
reducing human error.
•Wearable devices and IoT enable continuous
monitoring of lung health.
•Real-time monitoring of biomarkers aids in early
cancer detection.
•Future advancements will integrate multiple
sensors into compact devices like smartwatches.
•AI/ML and wearables offer promising solutions,
despite challenges with data quality and security.
Q&A
Reference
Businesswire, 2018. Auris Health Unveils the FDA-Cleared Monarch Platform, Ushering in a New Era of Medical
Intervention. [Online]
Available at: https://www.businesswire.com/news/home/20180323005162/en/Auris-Health-Unveils-the-FDA-
Cleared-Monarch-Platform-Ushering-in-a-New-Era-of-Medical-Intervention
[Accessed 11 02 2025].
csem.ch, 2022. Novel wearable vest for long-term lung monitoring. [Online]
Available at: https://www.csem.ch/en/news/welmo_novel-wearable-vest-for-long-term-lung-monitoring/
[Accessed 18 02 2025].
Geeksforgeeks, 2024. What is a Neural Network?. [Online]
Available at: https://www.geeksforgeeks.org/neural-networks-a-beginners-guide/
[Accessed 13 02 2025].
Mayoclinic, 2024. Lung cancer. [Online]
Available at: https://www.mayoclinic.org/diseases-conditions/lung-cancer/diagnosis-treatment/drc-
20374627#:~:text=Tests%20might%20include%20X%2Dray,looked%20at%20under%20a%20microscope.
[Accessed 11 02 2025].
Businesswire, 2018. Auris Health Unveils the FDA-Cleared Monarch Platform, Ushering in a New Era of Medical
Intervention. [Online]
Available at: https://www.businesswire.com/news/home/20180323005162/en/Auris-Health-Unveils-the-FDA-
Cleared-Monarch-Platform-Ushering-in-a-New-Era-of-Medical-Intervention
[Accessed 11 02 2025].
csem.ch, 2022. Novel wearable vest for long-term lung monitoring. [Online]
Available at: https://www.csem.ch/en/news/welmo_novel-wearable-vest-for-long-term-lung-monitoring/
[Accessed 18 02 2025].
Geeksforgeeks, 2024. What is a Neural Network?. [Online]
Available at: https://www.geeksforgeeks.org/neural-networks-a-beginners-guide/
[Accessed 13 02 2025].
Mayoclinic, 2024. Lung cancer. [Online]
Available at: https://www.mayoclinic.org/diseases-conditions/lung-cancer/diagnosis-treatment/drc-
20374627#:~:text=Tests%20might%20include%20X%2Dray,looked%20at%20under%20a%20microscope.
[Accessed 11 02 2025].
Reference
Parker, J., 2024. E-nose: the test that can “sniff” out lung cancer. [Online]
Available at: https://www.oncology-central.com/e-nose-the-test-that-can-sniff-out-lung-cancer/
[Accessed 18 02 2025].
Shah, A. A., Malik, H. A. M., Muhammad, A. & Butt, A. A. &. Z. A., 2023. Deep learning ensemble 2D CNN approach
towards the detection of lung cancer. [Online]
Available at: https://www.nature.com/articles/s41598-023-29656-z
[Accessed 11 02 2025].
Spirehealth, 2024. Remote Patient Monitoring for Chronic Respiratory Disease. [Online]
Available at: https://www.spirehealth.com/
[Accessed 11 02 2025].
World Health Organization, 2023. Lung cancer. [Online]
Available at: https://www.who.int/news-room/fact-sheets/detail/lung-cancer
[Accessed 10 02 2025].
Parker, J., 2024. E-nose: the test that can “sniff” out lung cancer. [Online]
Available at: https://www.oncology-central.com/e-nose-the-test-that-can-sniff-out-lung-cancer/
[Accessed 18 02 2025].
Shah, A. A., Malik, H. A. M., Muhammad, A. & Butt, A. A. &. Z. A., 2023. Deep learning ensemble 2D CNN approach
towards the detection of lung cancer. [Online]
Available at: https://www.nature.com/articles/s41598-023-29656-z
[Accessed 11 02 2025].
Spirehealth, 2024. Remote Patient Monitoring for Chronic Respiratory Disease. [Online]
Available at: https://www.spirehealth.com/
[Accessed 11 02 2025].
World Health Organization, 2023. Lung cancer. [Online]
Available at: https://www.who.int/news-room/fact-sheets/detail/lung-cancer
[Accessed 10 02 2025].
Thankyou
Tags
Download
Download Slideshow Get the original presentation fileQuick Actions
Statistics
- Views 83
- Slides 31
- Age 278 days
Related Slideshows
36
Clinical approach Dyspnoea A simple and practical approach.pptx
ShajahanPS
25 views
238
Cancer Awareness therapy for public by Dr Kanhu Charan Patro
kanhucpatro
24 views
41
Prof Satyadas Memorial oration Kozhikode.pptx
ShajahanPS
20 views
26
Viral Conjunctivitis and it;s managment.pptx
ZaidAzhar
34 views
30
Essential Thrombocythemia 15 Years of Experience at the Hematology Department, Algies, Algeria.pdf
sbelakehal
30 views
10
Hypertension sign symptoms cmdt style with regime
androiddabest
31 views