The utilization of artificial intelligence applications to improve breast cancer detection and prognosis

Article Figures & Data

Download figure
Open in new tab
Download powerpoint
Figure 1
- Conventional/Artificial neural networks (CNN/ANN) composition

Table 1

- Strengths and drawbacks of artificial intelligence (AI)-based applications in radiology.

Strengths	Drawbacks
Enhance analysis: Automated pathology screening, detection, and characterization.	Time and cost consuming in training and testing AI model.
Accurate classification: Categorize image based on abnormality (benign and malignant).	Ethical and legal issues: Ethics of data (How should we use, label and protect data?) Ethics of algorithm and trained model (How does the AI model make decisions? How can we diminish the risk of patient harm from privacy breaches? Who is responsible for mistakes resulting from the use of the AI model?) Ethics of practice (monitor and verify AI-driven autonomy)
Extract additional required information from previous detected pathology.	Biased predicted outcomes due to incomplete and/or unrepresentative data.
Offer a second opinion which increases confidence of the diagnosis.	Lack of interpretability can lead to a lack of trust and acceptance of AI models by physicians and patients.
Minimize interindividual variability, bias and time.	Lack of strong evidence and regulations to support the use of an AI model.
Minimize interindividual variability, bias and time.	Lack of standardized benchmarks which make it difficult to validate the performance of an AI model.