Skip to main content

Table 4 The models that achieved the highest accuracy for classifying diseases, along with their strengths

From: Deep learning approaches for classification tasks in medical X-ray, MRI, and ultrasound images: a scoping review

Organ

Disease

Ref.

Model

Acc.%

Strength

Access.

Dataset size

Year

Lungs

Covid-19, Pneumonia, Pediatric Pneumonia

[40]

EfficientNet + VGG16 + ELM + explainable AI

99.1

Computational efficiency of EfficientNet and increased model interpretability through XAI

public

12,723

2022

[44]

DenseNet-201

99.1

High computational & memory efficiency

public

16,000

2022

Pulmonary Tuberculosis

[96]

Efficient Xception CNN

99.29

High efficiency + fast training + good generalizability

public

5600

2023

Brain

Alzheimer Disease, Dementia

[65]

CNN

98.63

Synthetic oversampling technique

public

6400

2022

Brain Tumor, Glioma Tumors

[77]

CNN with Multi-Branch Network + Inception block

99.3

Interpretability and visual explanation of the results through Grad-CAM + Computational efficiency of EfficientNet

public

2870

2023

ASD, DOC

[102]

3D EfficientNet-B3

92.7

Addressing class imbalance via synthetic oversampling technique

private

151

2024

Cortical Tubers

[103]

TSCCNN + InceptionV3 + ResNet50

95

Multi-scale extraction of features + efficiency on complex tasks

private

6318

2020

Mammary Gland

Breast Cancer

[87]

DeepBraestCancerNet (Custom CNN)

99.35

customization + flexibility in solution design

public

1030

2023

Breast Cancer

[91]

EfficientNet-B7 and Explainable AI

99.14

Computational efficiency of EfficientNet and increased model interpretability through XAI

public

780

2024

Thoracic Cavity

Thoracic Diseases

[71]

DenseNet

99.58

High computational & memory efficiency

public

984

2018

Liver

Liver lesions/Liver tumor

[98]

USC-Enet (based on EfcientNet-B0)

95.6

Computational efficiency of EfficientNet and increased model interpretability through XAI

private

2168

2023

Thyroid Gland

Thyroid Nodules

[83]

ResNet18

98.4

Low computational costs

private

508

2022

Bones

Bone Fracture, Femur Fracture

[93]

Hybrid SFNet + Canny + Grey

99.12

Integration of Canny Edge algorithm for feature localization + high efficiency + low computational cost

public

34,000

2022

Spine

Disc Herniation

[104]

AlexNet

87.75

Feature extraction + classification performed in a common structure and the use of real private patient data

private

2329

2019

Uterus

Fetal Malposition

[105]

Fet-Net (CNN)

97.68

Customization + flexibility in solution design

private

144

2023

Prostate Gland

prostate cancer

[107]

CNN

89.9

Customization + flexibility in solution design

private

21

2017