The impact of pre-processing techniques on deep learning breast image segmentation
Jéssica Catarino, Nuno Cruz Garcia, Sara Silva, João Santinha · 2025
Read The impact of pre-processing techniques on deep learning breast image segmentation by Jéssica Catarino, Nuno Cruz Garcia, Sara Silva, João Santinha (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12789679.txt.
Description
Jéssica Catarino and co-authors ask how much of a medical AI's skill is decided before the model ever sees the image. Deep learning pipelines begin with pre-processing, resizing, normalizing, orienting, yet its effect on breast image segmentation is barely studied. Their 2025 paper runs one standard U-Net model over two public breast datasets, mammography and MRI, while varying the preparation systematically.
Two pipelines compete. One applies the generic habits of image analysis; the other is domain-specific, handling breast metadata carefully to preserve anatomy. Each technique, intensity normalization, spacing harmonization, resizing and padding, orientation standardization, is isolated and scored for its effect. The authors identify which preparation choices genuinely help segmentation in this domain and which are ritual. The study is a first step, they concede, toward treating pre-processing as an experimental variable rather than a house style.
Catalog record
| CREATOR | Jéssica Catarino, Nuno Cruz Garcia, Sara Silva, João Santinha |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |