European Radiology: The importance of CT image standardization in radiomic prognostic analysis of lung cancer

According to statistics, lung cancer is one of the leading causes of
cancer-related death worldwide.
Although complete resection is a potential curative treatment for early-stage non-small cell lung cancer (NSCLC), recurrence
occurs in 30-55% of patients.
Even patients at the same stage of pathology have different risks of recurrence and death, so it is difficult to make an accurate prognostic assessment
clinically.
Although the pathologic stage is the most important predictor of prognosis after surgical resection of NSCLC, studies have shown that tumor morphology may also be related
to patient prognosis.

Recent radiomics studies have attempted to predict prognosis
using the quantitative radiomic signature of computed tomography (CT) images.
However, because the quality of medical images varies depending on the reconstruction algorithm, the quantitative radiological features extracted from these images also vary
.
Especially for CT images, radiomic features vary
depending on acquisition parameters such as voxel size and reconstructed nuclei.

Recently, a study published in the journal European Radiology evaluated the performance of standardized CT images in improving the prognosis prediction of radiomics models, providing data support
for further standardizing the use of radiomics and improving the accuracy of prediction.

A total of 106 patients with
NSCLC were included in this review.
For each patient, 851 radiological features
were extracted from standardized and unstandardized CT images.
After feature selection, a random forest model
is constructed with selected radiological and clinical features.
These models were then externally validated
in a test set of 79 NSCLC patients.

Models using standardized CT images produced better diagnostic performance than models using unstandardized CT images (AUC of 0.
802 vs 0.
702, p = 0.
01), which was in the lung Patients with adenocarcinoma performed particularly well (AUC 0.
880 vs 0.
720, p < 0.
01).

Figure the histogram
of selected three features with standardized CT images.
The blue bars represent patients with a 3-year recurrence-free survival (RFS) rate
.
The red bars represent patients who relapse or die at 3 years

This study found that CT image standardization can identify more radiological features, which improves the predictive performance of prognostic models, and proves the importance of
CT image standardization for radiology feature extraction.
In addition, standardized prognostic prediction performance is improved
even when the same features are used for the prognostic model.

Original source:

Doohyun Park,Daejoong Oh,MyungHoon Lee,et al.
Importance of CT image normalization in radiomics analysis: prediction of 3-year recurrence-free survival in non-small cell lung cancer.
DOI:10.
1007/s00330-022-08869-2