Type of Data

Sample

Objective

Model Design

Results

Therapeutic Area

Paper

Mammography

Mammography images

45,000 images

Detect malign solid lesions and prevent overtreatment in false positives [2]

CNN

AUC of 0.90

Oncology

[89]

Mammography

667 benign and 333 malignant

Mammography diagnosis of early malignant breast

Stacked AE

Accuracy of 0.89

Oncology

[96]

Digital Mammography images and the biopsy result of the lesions [2]

1000 malignant masses and 600 cysts images and their biopsy [2]

Discriminate benign cysts from malignant masses

CNN

AUC of 0.80

Oncology

[97]

Mammography images

840 images of mammograms from 210 different patients

Breast arterial calcification on mammograms classifier to evaluate the risk of coronary disease [2]

CNN

Misclassfied cases of 6%

Cardiovascular

[101]

Digital mammograms

661 from 444 patients

Computer automated estimation of breast percentage density [2]

CNN

AUC of 0.981

Oncology

[151]

Mammography images

Mammograms from 604

women

Segment areas of dense fibroglandular tissue in the breast [2]

CNN

Accuracy of 0.66

Oncology

[116]

Digital mammograms images

29,107 left mediolateral oblique, right mediolateral oblique, left cranial-caudal and right cranial-caudal

mammograms images

Probability of cancer on mammograms [2]

CNN

AUC of 0.90

Oncology

[121]

Ultrasound

Image of the heart 2D

400 images with five different heart diseases and 80 normal echocardiogram images

Segment left ventricle images with greater precision

Deep belief

networks

Hammoude distance of 0.80

Cardiovascular

[152]

Ultrasound imaging

306 malignant and 136 benign tumors images

CAD system to detect and differentiate breast lesions with ultrasound

CNNs inspired in AlexNet, U-Net and LeNet

Best F-measure of 0.91 and 0.89 depending on the data

Oncology

[2] [24]

Transesophageal ultrasound volume and 3D geometry of the aortic valve images

3795 volumes from the aortic valves from 150 patients

Diagnose, stratification and treatment planning for patients with aortic valve pathologies

Marginal space deep learning

Position error of 1.66 mms and mean corner distance error of 3.29 mms

Cardiovascular

[2] [84]