Detection of Acute Infarct on DWI Images

Purpose

The purpose of this proposal is to automatically detect acute infarcts on diffusion-weighted images (DWI) on brain MRI. Another goal is to simplify the language while reporting the lesions with diffusion restriction.

Tag(s)

 

Panel

Neuroradiology Panel

Define-AI ID

19020008

Originator

Achint Singh
Lead Achint Singh 

Panel Chair

Alexander Norbash

Panel Reviewers

Neuroradiology Panel

License

Creative Commons 4.0 
Status Public Commenting
RadElement Set RDES97
                               

Clinical Implementation


Value Proposition

Diffusion refers to the random microscopic movement of the water molecules in the tissues. DWI provides a quantitative measure of this molecular motion in three dimensional space. DWI is the most sensitive MR imaging sequence for the detection of acute infarct. Diffusion restriction is defined as increased signal on the diffusion weighted images and low signal on the ADC images. Diffusion restriction is usually seen in hyperacute infarct within the first 3 hours however, can be seen as early as 11 minutes. Low signal is usually seen on the ADC map up to 7-10 days after which the signal starts to increase, which is known as pseudonormalization. 

Diffusion restriction can also be seen in other conditions such as 

  1. Abscess

  2. CJD

  3. Herpes or other encephalitis

  4. Lymphoma

  5. GBM

  6. Epidermoid cyst

  7. DAI

  8. Hemorrhage

  9. Seizures

  10. Drug toxicity


Narrative(s)


A 30-year-old male comes with non-specific symptoms of left arm weakness, possible seizures, and altered mental state. CT head is negative for hemorrhage. CTA did not show large vessel occlusion. tPA was not given due to non-specificity of symptoms and a fast MRI was obtained which shows an acute infarct. Automatic detection of this infarct could be helpful in this patient for timely tPA administration.


Workflow Description


The patient receives a CT scan for suspected signs of stroke. If there is no hemorrhage on the CT then the patient receives a stroke protocol MRI with DWI images along with an ADC map. The algorithm receives the MR dataset. The algorithm tries to determine if there is true diffusion restriction based on the signal on DWI and ADC images. If diffusion restriction is present, then an alert should notify the user. 

The algorithm also tries to characterize the abnormality as described below in the output section. The algorithm is integrated with the dictation system to automatically populate the results. The representative image is also saved on the PACS system.


Considerations for Dataset Development



Procedures

MR brain (axial DWI and ADC)

Sex at Birth

Male and Female

Age

All Ages

DWI

Hyperintense signal in comparison to the adjacent gray matter (True and False)

ADC

low signal in comparison to the adjacent region (True or False)

Technical Specifications


Inputs

 

DICOM Study

Procedure

MR, Head

Views

Axial

Data Type

DICOM

Modality

MR

Body Region

Head

Anatomic Focus

Brain

Scenario

axial DWI and ADC




Primary Outputs



Diffusion Restriction

RadElement ID

RDE591

Definition

Determine if high signal on DWI is true diffusion restriction or T2 shine through by analyzing ADC signal

Data Type

Categorical

Value Set

  • No
  • Yes
  • Unknown

Units

N/A



Shape of Area with Diffusion Restriction

RadElement ID

RDE592

Definition

Categorize the shape of the area with diffusion restriction

Data Type

Categorical

Value Set

  • No
  • Yes
  • Unknown

Units

N/A


Distribution of Diffusion Restriction

RadElement ID

RDE593

Definition

Categorize the distribution of the diffusion restriction in the lesion

Data Type

Categorical

Value Set

  • Whole lesion
  • Central
  • Peripheral
  • Rim

Units

N/A


Count of lesions with diffusion restriction

RadElement ID

RDE589

Definition

Count the number of lesion(s) with diffusion restriction

Data Type

Numeric

Value Set

N/A

Units

Number of lesions



Lesion location with diffusion restriction

RadElement ID

RDE597

Definition

Location of the lesion(s) with diffusion restriction

Data Type

Categorical

Value Set

0 - Frontal

1 - Parietal

2 - Occipital

3 - Temporal

4 - Deep gray Nuclei (Basal Ganglia/Thalami)

5 - White matter

5A: Periventricular WM

5B: Deep WM

5C: Subcortical WM

5D: Corticospinal tract

5E: Cerebellar Peduncles

6 - Brainstem

6A: Midbrain

6B: Pons

6C: Medulla

7 - Cerebellum


Units

N/A

 

Laterality of lesion

RadElement ID

RDE598

Definition

Define the side of the lesion

Data Type

Categorical

Value Set

  • Right
  • Left
  • Midline

Units

N/A

 

2D Lesion Size

RadElement ID

RDE587

Definition

Size of the lesion in two dimensions

Data Type

Numeric

Value Set

N/A

Units

mm2

 

Lesion Volume

RadElement ID

RDE588

Definition

Volume of the lesion

Data Type

Numeric

Value Set

N/A

Units

mm3



Signal Characteristics 

RadElement ID

RDE594

Definition

Signal characteristics of the lesion

Data Type

Categorical

Value Set

  • Homogeneous
  • Heterogeneous

Units

N/A


Future Development Ideas


  • Analyze the associated features and clinical features with diffusion restriction so that the model is able to give reasonable differential diagnoses
  • Associated susceptibility artifacts: Primary hemorrhage
  • Peripheral diffusion restriction : PML
  • Solid ring enhancement with diffusion restriction in the solid portion : GBM
  • Ring enhancement with diffusion restriction in the center: Abscess
  • Focal gyriform diffusion restriction: Status epilepticus
  • Peripheral cortical areas with rapidly progressive dementia: CJD