Page 112 - AIH-2-2

Artificial Intelligence in Health AI in early breast cancer diagnosis: A review

Figure 3. Flowchart of the screening process for eligible articles included in this study

Table 3. Summary of computer‑aided detection studies in early detection of breast cancer
Technique(s) References Year Approach(es) Inference Limitations
employed
Feature-based 17 2022 Multi-scale Performs better w.r.t accuracy and Needs to be validated on the external
convolutional recall than the existing segmentation dataset.
network and classification.
Region-based 18 2021 Depth-wise Outperforms when compared with • Handling small label datasets by
convolutional ShuffelNet, MobileNet, SVM, K-NN, transfer learning
neural network RF, and other studies. • True‑positive rate needs to
improve, and the ability to detect
architectural distortions track from
DBT slices
Feature-based 19 2023 Deep convolutional Results indicated that the proposed • No data was collected about
neural networks EDL-based AI system can enhance modifiable risk factors, breast
the accuracy and sensitivity of cancer family history, and BRCA
radiologists. gene test results
• The performance of the DL model
can be impacted by the quality of
diagnosis.
(Cont’d...)

Volume 2 Issue 2 (2025) 106 doi: 10.36922/aih.4197