OMAMA-DB: THE OREGON-MASSACHUSETTS MAMMOGRAPHY DATABASE

Under the guidance of 

Professor Daniel Haehn 

March 31st 2026

SPIE Journal of Medical Imaging (JMI) Special Issue “Peer-Validated Resources for Medical Imaging: Datasets, Software, and Benchmarks

Submitted to :

Motivation

What is OMAMA-DB?

- Public mammography dataset

- Hosted on Harvard Dataverse

- Includes:
                   2D Mammograms
                   3D Tomosynthesis

- Total  of 231,080 images

2D

Available Data Formats in OMAMA-DB

3D

4-Views

Dataset Comparision

Dataset Creation

Outlier Detection (ODM)

Web-Based Annotation Tool

white/0.00, red/0.20, orange/0.40, yellow/0.60, green/0.80, and blue/1.00

Web-Based Annotation Tool

white/0.00, red/0.20, orange/0.40, yellow/0.60, green/0.80, and blue/1.00

Deepsight

Applied to:
191,547 2D
67,534 3D

Deepsight Study vs Image level

deepsight2.sh -i /path/to/combined/ -o /path/to/OUT/combined/ -cl
/path/to/combined caselist.txt

deepsight2.sh -i /path/to/single/01 -o /path/to/OUT/single1/ -cl
/path/to/single01 caselist.txt --additional params --checks to ignore
SAC-30 SAC-40 SAC-50 SAC-60 FAC-200

Deepsight Study vs Image level

Deepsight Study vs Image level

Left : (0.6889+0.557)/2 = 0.623

Right: (0.0709+0.1253)/2 = 0.0981

Download Stats

APPLICATIONS AND USE CASES

Binary Cancer Classification Using MedGemma

Synthetic Classifier