AI in Medical Diagnosis - Module 1

Question 1

Advances in the last decade have largely been due to three factors

Answer 1

1. Maturation of deep learning
2. Compute power via GPUs
3. Open-sourcing of large labeled datasets

Question 2

Clinical tasks suitable for Computer Vision

Answer 2

Screening
Diagnosis
Detecting conditions
Outcome prediction
Segmentation of pathologies
Monitoring disease
Clinical research

Question 3

Object classification

Answer 3

Refers to identifying the type of object in an image

Question 4

Object localization

Answer 4

Refers to the localization of an object in a image

Question 5

Object detection in computer vision

Answer 5

Refers to identifying the type of object and location of the object

Question 6

What is the key technique that leveraged CV

Answer 6

Convolutional neural networks

Question 7

Characteristic of the mechanism of a CNN

Answer 7

Hardcodes translational invariance

Question 8

Traslational invariance is

Answer 8

The system produce the same response regardless of how the input is shifted

Question 9

Image registration

Answer 9

identifying corresponding points across similar

images

Question 10

image retrieval

Answer 10

finding similar images

Question 11

Data points in CV

Answer 11

Images

Question 12

Data labels in CV

Answer 12

Object classes

Question 13

Why transfer learning has been critical in medical AI

Answer 13

Given the sparcity and access difficulties of medical data

Question 14

3 Strategies to reduce costs associated with collecting and labeling data

Answer 14

1. Data augmentation
2. Generative adversarial networks
3. Crowd sourcing

Question 15

Self supervised learning means

Answer 15

Implicit labels are extracted from data points and used to train algorithms

Question 16

Federated learning means

Answer 16

Centralized algorithms can be trained on distributed data

Question 17

Multimodal learning

Answer 17

Combining vision with other modalities

Question 18

Useful in detecting adverse clinical events

Answer 18

Activity recognition

Live scene undestanding

Question 19

Challenges in DL based computer vision with medical imagery

Answer 19

Massive images
Variance on technique of digitalizing images
3D images

Question 20

Multiple instance learning (MIL)

Answer 20

Is a type of supervised learning. Instead of receiving a set of instances which are individually labeled, the learner receives a set of labeled bags, each containing many instances.

Question 21

How to deal with non standarized data collection

Answer 21

Integrating computer vision in existing physical systems

Question 22

CV in gastroenterology

Answer 22

Esophageal cancer detection
Lesion detection (gastric cancer)
Lesion diagnosis (h pilory)

Question 23

CV in radiology

Answer 23

Analysis of brain (stroke)
Nodule detection
Region segmentation (ventricular or coronary)

Question 24

New uses of CV in radiology

Answer 24

Image reconstruction and enhancement
Automated report generation
Temporal tracking

Question 25

CV in cardiology

Answer 25

Coronary artery segmentation

Echocardiography (hypertrophic cardiomyopathy, cardiac amyloid and pulmonary hypertension)

Question 26

CV in Pathology potentials

Answer 26

1. Overcome limitations of human vision and cognition
2. Develop new signatures of disease and therapy
3. Combination of pathology with radiological, genomic and proteomic measures to improve prognosis

Question 27

CV in dermatology

Answer 27

Lesion specific differential diagnostics
Finding concerning lesions against bening
Track lesion over time

Question 28

CV in Opthalmology

Answer 28

Diagnosis of diabetic retinopathy
Macular edema
Glaucoma
Non ocular complications

Question 29

CV in surgical video

Answer 29

1. Increase performance through real time context awareness
2. Skills assesments
3. Training

Question 30

GOALS

Answer 30

Global

Operative Assessment of Laparoscopic Skills criteria

Question 31

Ambient intelligence

Answer 31

Computer vision coupled with sensors
and video streams enables a number of safety applications in clinical
and home settings, enabling healthcare providers to scale their
ability to monitor patients.

Question 32

Key considerations when applying ML in healthcare

Answer 32

1. Assessment of data
2. Planning model limitations
3. Community participation
4. Trust building

Question 33

Methods to remove individual level bias

Answer 33

Expert discussion

Label adjudication

Question 34

Method to remove population level bias

Answer 34

Missing data supplements
Distributional shifts
Multi institutional evaluation
Multi task learning (eg. multicancer detection rather than one type of cancer)
Assess demographic performance w saliency maps

Question 35

Model limitations planning

Answer 35

Developing confidence intervals

Explainability

Question 36

To increase community participation

Answer 36

Side by side deployment with existing workflows

Question 37

AI interpretability

Answer 37

why specific factors about

the patient or environment lead them to their predictions

Question 38

Federated learning

Answer 38

Federated Learning enables mobile phones to collaboratively learn a shared prediction model while keeping all the training data on device

Question 39

inherent explainability

Answer 39

For machine learning models for which the input data are of limited complexity and clearly understandable (eg. linear regression)

Question 40

post-hoc explainability

Answer 40

the data and models are too
complex and high-dimensional to be easily understood; they cannot be explained by a simple relationship dissect the model’s decision making procedure

Question 41

Heat map or Saliency maps

Answer 41

highlight how much each
region of the image contributed to a given decision and are illustrative because they provide a simple means of understanding some of the limitations of post-hoc explainability techniques

Question 42

adversarial attack

Answer 42

when precise modifications are made to the input that

substantially alter the model’s predictions