European Radiology: Application of Convolutional Neural Networks in Automatic Detection of Brain Metastases

Because the presence or absence of brain metastases is related to staging and treatment planning, accurate diagnosis of brain metastases is critical for determining appropriate treatment strategies .
For example, whole-brain radiation therapy is the standard treatment for multiple metastases; however, in patients with smaller and fewer lesions, stereotactic radiation therapy is considered .
Therefore, it is crucial to accurately assess the number, size, and location of lesions before surgery .

Because the presence or absence of brain metastases is related to staging and treatment planning, accurate diagnosis of brain metastases is critical for determining appropriate treatment strategies .

For example, whole-brain radiation therapy is the standard treatment for multiple metastases; however, in patients with smaller and fewer lesions, stereotactic radiation therapy is considered .
Therefore, it is crucial to accurately assess the number, size, and location of lesions before surgery .

In recent years, automatic detection of brain metastases using deep learning techniques (such as neural network) algorithms has been introduced into clinical applications .

Several studies have demonstrated the advantages of using various convolutional neural networks (CNNs) to automatically detect brain metastases .
On contrast- enhanced T1-weighted imaging, enhanced blood vessels may be misdiagnosed as metastatic tumors .
Therefore, imaging methods of vascular inhibition have been introduced to reduce the occurrence of false positives (FPs) .

In recent years, automatic detection of brain metastases using deep learning techniques (such as neural network) algorithms has been introduced into clinical applications .
Several studies have demonstrated the advantages of using various convolutional neural networks (CNNs) to automatically detect brain metastases .
On contrast- enhanced T1-weighted imaging, enhanced blood vessels may be misdiagnosed as metastatic tumors .
Therefore, imaging methods of angiosuppression have been introduced to reduce the occurrence of false-positive (FPs) vessels .
Some scholars have proposed a three-dimensional MR sequence-VISIBLE technology for the detection of brain metastases .
VISIBLE can simultaneously acquire images with vascular suppression (hereinafter referred to as "black images") and images without vascular suppression ("bright images") .

Recently, a study published in the journal European Radiology developed an automated model for the detection of brain metastases using CNN and VISIBLE and compared its diagnostic performance with previous observer tests for rapid and accurate quantification of brain metastases.

, qualitative and positioning provided technical support .
 

Recently, a study published in the journal European Radiology developed an automated model for the detection of brain metastases using CNN and VISIBLE and compared its diagnostic performance with previous observer tests for rapid and accurate quantification of brain metastases.
, qualitative and positioning provided technical support .
 

This retrospective study included patients with clinical suspicion of brain metastases using March 2016 and July 2019 .
Images with and without vessel suppression were selected for training an existing CNN (DeepMedic) .
Diagnostic performance was assessed using sensitivity and false positive results per case (FPs/case) .
This study compared the diagnostic performance of the CNN model with that of 12 radiologists .
 

Fifty patients (30 males and 20 females; age range 29-86 years; mean 63.

3±12.
8 years; total 165 metastatic cases) who were clinically diagnosed with brain metastases during follow-up were used for training in

Fifty patients (30 males and 20 females; age range 29-86 years; mean 63.

Figure 60-year-old woman with lung cancer and multiple brain metastases .
In black (a) and bright (b) images, a metastasis (arrow) in the right frontal lobe is shown .
Its size is about 2 mm .
Our CNN model accurately detected this lesion (arrow) in black (c) and bright (d) images of a  60 -year-old woman with lung cancer and multiple brain metastatic lung cancer lesions .
In black (a) and bright (b) images, a metastasis (arrow) in the right frontal lobe is shown .
Its size is about 2 mm .
Our CNN model accurately detected this lesion (arrow) in black (c) and light (d) images 

In conclusion, the model created by VISIBLE and CNN in this study for diagnosing brain metastases showed higher sensitivity compared to radiologists .

In conclusion, the model created by VISIBLE and CNN in this study for diagnosing brain metastases showed higher sensitivity compared to radiologists .

Original source :

Yoshitomo Kikuchi , Osamu Togao , Kazufumi Kikuchi , et al.

A deep convolutional neural network-based automatic detection of brain metastases with and without blood vessel suppression .
DOI: 10.
1007/s00330-021-08427-2

Yoshitomo Kikuchi Osamu Togao Kazufumi Kikuchi , et al.
A deep convolutional neural network-based automatic detection of brain metastases with and without blood vessel suppression .
10.
1007/s00330-021-08427-2 10.
1007/s00330-021-08427-2 Leave a comment here