Prediction Method for Common Diseases Based on Chest X-Ray Images

Jiangfeng Wang; Lijun Liu; Qingsong Huang; Li Liu; Xiaodong Fu

doi:10.3788/LOP202259.1817001

Journals >Laser & Optoelectronics Progress >Volume 59 >Issue 18 >Page 1817001 > Article

Laser & Optoelectronics Progress
Vol. 59, Issue 18, 1817001 (2022)

Prediction Method for Common Diseases Based on Chest X-Ray Images

Jiangfeng Wang¹, Lijun Liu^1、2、*, Qingsong Huang¹, Li Liu¹, and Xiaodong Fu¹

Author Affiliations

¹School of Information Engineering and Automation, Kunming University of science and technology, Kunming 650500, Yunnan , China

²School of Information, Yunnan University, Kunming 650091, Yunnan , China

show less

DOI: 10.3788/LOP202259.1817001 Cite this Article Set citation alerts

Jiangfeng Wang, Lijun Liu, Qingsong Huang, Li Liu, Xiaodong Fu. Prediction Method for Common Diseases Based on Chest X-Ray Images[J]. Laser & Optoelectronics Progress, 2022, 59(18): 1817001 Copy Citation Text

show less

Fig. 1. General framework of DFFM

Download full size

Fig. 2. Channel attention sub module

Download full size

Fig. 3. Spatial attention sub module

Download full size

Fig. 4. Distribution of different tags in the MIMIC-CXR

Download full size

Fig. 5. ROC curves and results on MIMIC-CXR dataset

Download full size

Fig. 6. Comparison of AUC values of various diseases in different documents of MIMIC-CXR dataset

Download full size

Fig. 7. Heat maps of lesion area generated by Grad-CAM

Download full size

Pathlolgy	Rajpurkar et al	Zhang C M	DFFM
Atelectasis	0.700	0.789	0.820
Cardiomegaly	0.814	0.900	0.871
Effusion	0.759	0.838	0.808
Infiltration	0.661	0.708	0.906
Tumor	0.693	0.839	0.795
Nodules	0.669	0.798	0.839
Pneumonia	0.658	0.736	0.788
Penumothorax	0.799	0.877	0.721
Consolidation	0.703	0.763	0.890
Edema	0.805	0.851	0.899
Emphysema	0.833	0.938	0.916
Fibrosis	0.786	0.832	0.707
Pleural Thickening	0.684	0.804	0.830
Fracture	0.871	0.944	0.877
Average	0.745	0.830	0.833

Table 1. Comparison of AUC values of different algorithms on ChestX-ray14 dataset

Pathlolgy	ResNet	DenseNet	Rubin et al	Wang et al	DFFM
Atelectasis	0.695	0.687	0.766	0.826	0.822
Cardiomegaly	0.742	0.727	0.840	0.879	0.864
Lung Opacity	0.767	0.732	0.632	0.906	0.834
Edema	0.901	0.880	0.734	0.885	0.917
Pleural Effusion	0.726	0.646	0.757	0.725	0.812
Infiltration	0.510	0.468	0.761	0.632	0.788
Fracture	0.735	0.617	0.815	0.643	0.771
Lung Lesion	0.701	0.704	0.748	0.775	0.788
Support Devices	0.782	0.780	0.692	0.775	0.877
No Finding	0.878	0.872	0.758	0.942	0.909
Consolidation	0.632	0.615	0.568	0.705	0.888
Pleural Thickening	0.617	0.609	0.687	0.871	0.763
Pneumonia	0.859	0.839	0.625	0.833	0.923
Pneumothorax	0.808	0.891	0.706	0.729	0.895
Average	0.739	0.719	0.721	0.794	0.847

Table 2. Comparison of AUC values of different algorithms on MIMIC-CXR dataset

Jiangfeng Wang, Lijun Liu, Qingsong Huang, Li Liu, Xiaodong Fu. Prediction Method for Common Diseases Based on Chest X-Ray Images[J]. Laser & Optoelectronics Progress, 2022, 59(18): 1817001

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information