Table 4 The age information is integrated into the proposed framework.

From: A 178-clinical-center experiment of integrating AI solutions for lung pathology diagnosis

Visual lung aging match age

Visual lung aging does not match age

Abnormality

AUC

Sensitivity

Abnormality

AUC

Sensitivity

Effusion

0.89 [0.86–0.92]

0.97 [0.94–0.99]

Effusion

0.84 [0.80–0.88]

0.97 [0.93–0.99]

Infiltrates

0.70 [0.68–0.71]

0.86 [0.84–0.88]

Infiltrates

0.66 [0.64–0.68]

0.87 [0.85–0.89]

Cyst

0.87 [0.72–1.00]

0.88 [0.53–0.98]

Cyst

0.64 [0.23–1.00]

1.00 [0.34–1.00]

Pulmonary mass

0.70 [0.68–0.71]

0.87 [0.85–0.89]

Pulmonary mass

0.63 [0.60–0.65]

0.86 [0.83–0.88]

Opacity

0.77 [0.75–0.78]

1.00 [1.00–1.00]

Opacity

0.72 [0.70–0.74]

1.00 [1.00–1.00]

Atelectasis

0.83 [0.74–0.93]

0.96 [0.82–0.99]

Atelectasis

0.85 [0.77–0.93]

1.00 [0.91–1.00]

Pneumothorax

0.87 [0.81–0.93]

0.98 [0.92–0.99]

Pneumothorax

0.93 [0.88–0.98]

0.98 [0.88–1.00]

Pneumonia

0.81 [0.79–0.84]

0.92 [0.89–0.94]

Pneumonia

0.87 [0.85–0.89]

0.98 [0.96–0.99]

Tuberculosis

0.79 [0.73–0.85]

0.91 [0.83–0.96]

Tuberculosis

0.77 [0.67–0.88]

0.96 [0.82–0.99]

Fibrosis

0.71 [0.68–0.74]

0.89 [0.85–0.91]

Fibrosis

0.55 [0.51–0.58]

0.79 [0.74–0.83]

Hernia

0.58 [0.43–0.74]

0.79 [0.52–0.92]

Hernia

0.46 [0.28–0.65]

0.78 [0.45–0.94]

Cardiomegaly

0.70 [0.68–0.72]

0.88 [0.86–0.90]

Cardiomegaly

0.60 [0.57–0.62]

0.85 [0.82–0.90]

Wid. mediastinum

0.74 [0.64–0.85]

0.93 [0.77–0.98]

Wid. mediastinum

0.35 [0.22–0.48]

0.64 [0.39–0.84]

Hilar enlargement

0.87 [0.76–0.97]

1.00 [0.83–1.00]

Hilar enlargement

0.72 [0.58–0.85]

0.84 [0.62–0.94]

Scoliosis

0.65 [0.60–0.70]

0.83 [0.76–0.86]

Scoliosis

0.60 [0.52–0.68]

0.88 [0.76–0.94]

Bone fracture

0.77 [0.70–0.85]

0.92 [0.82–0.97]

Bone fracture

0.73 [0.62–0.84]

0.96 [0.81–0.99]

Consolidate fracture

0.66 [0.58–0.73]

0.83 [0.71–0.90]

Consolidate fracture

0.74 [0.65–0.82]

1.00 [0.91–1.00]

  1. A separate logistic regression model was implemented to predict visual lung aging from the age feature. The model performance was of 0.79 and 0.70 AUC and accuracy. This table presents the framework results in terms of the area under the receiving operator curve and sensitivity against the patients whose visual lung aging was correctly/incorrectly predicted from their age. The higher number for each abnormality is highlighted in bold.
Back to article page