Simulating mismatch between calibration and target population in AI for mammography the retrospective VAIB study
(2025) In npj Digital Medicine 8(1).- Abstract
AI cancer detection models require calibration to attain the desired balance between cancer detection rate (CDR) and false positive rate. In this study, we simulate the impact of six types of mismatches between the calibration population and the clinical target population, by creating purposefully non-representative datasets to calibrate AI for clinical settings. Mismatching the acquisition year between healthy and cancer-diagnosed screening participants led to a distortion in CDR between −3% to +19%. Mismatching age led to a distortion in CDR between −0.2% to +27%. Mismatching breast density distribution led to a distortion in CDR between +1% to 16%. Mismatching mammography vendors lead to a distortion in CDR between −32% to + 33%.... (More)
AI cancer detection models require calibration to attain the desired balance between cancer detection rate (CDR) and false positive rate. In this study, we simulate the impact of six types of mismatches between the calibration population and the clinical target population, by creating purposefully non-representative datasets to calibrate AI for clinical settings. Mismatching the acquisition year between healthy and cancer-diagnosed screening participants led to a distortion in CDR between −3% to +19%. Mismatching age led to a distortion in CDR between −0.2% to +27%. Mismatching breast density distribution led to a distortion in CDR between +1% to 16%. Mismatching mammography vendors lead to a distortion in CDR between −32% to + 33%. Mismatches between calibration population and target clinical population lead to clinically important deviations. It is vital for safe clinical AI integration to ensure that important aspects of the calibration population are representative of the target population.
(Less)
- author
- Schurz, Haiko
; Solander, Klara
; Åström, Davida
; Cossío, Fernando
; Choi, Taeyang
; Dustler, Magnus
LU
; Lundström, Claes ; Gustafsson, Håkan ; Zackrisson, Sophia LU and Strand, Fredrik
- organization
- publishing date
- 2025-12
- type
- Contribution to journal
- publication status
- published
- subject
- in
- npj Digital Medicine
- volume
- 8
- issue
- 1
- article number
- 259
- publisher
- Nature Publishing Group
- external identifiers
-
- scopus:105004411904
- pmid:40341801
- ISSN
- 2398-6352
- DOI
- 10.1038/s41746-025-01623-0
- language
- English
- LU publication?
- yes
- id
- 315cb069-9623-4271-8bd4-0e9dcd135fda
- date added to LUP
- 2025-07-14 09:05:22
- date last changed
- 2025-07-15 03:08:36
@article{315cb069-9623-4271-8bd4-0e9dcd135fda, abstract = {{<p>AI cancer detection models require calibration to attain the desired balance between cancer detection rate (CDR) and false positive rate. In this study, we simulate the impact of six types of mismatches between the calibration population and the clinical target population, by creating purposefully non-representative datasets to calibrate AI for clinical settings. Mismatching the acquisition year between healthy and cancer-diagnosed screening participants led to a distortion in CDR between −3% to +19%. Mismatching age led to a distortion in CDR between −0.2% to +27%. Mismatching breast density distribution led to a distortion in CDR between +1% to 16%. Mismatching mammography vendors lead to a distortion in CDR between −32% to + 33%. Mismatches between calibration population and target clinical population lead to clinically important deviations. It is vital for safe clinical AI integration to ensure that important aspects of the calibration population are representative of the target population.</p>}}, author = {{Schurz, Haiko and Solander, Klara and Åström, Davida and Cossío, Fernando and Choi, Taeyang and Dustler, Magnus and Lundström, Claes and Gustafsson, Håkan and Zackrisson, Sophia and Strand, Fredrik}}, issn = {{2398-6352}}, language = {{eng}}, number = {{1}}, publisher = {{Nature Publishing Group}}, series = {{npj Digital Medicine}}, title = {{Simulating mismatch between calibration and target population in AI for mammography the retrospective VAIB study}}, url = {{http://dx.doi.org/10.1038/s41746-025-01623-0}}, doi = {{10.1038/s41746-025-01623-0}}, volume = {{8}}, year = {{2025}}, }