Subdural Hematoma on MRI

Value Proposition

Subdural hematomas and collections are very common and can be traumatic or atraumatic; they predominately occur in the elderly and in patients with bleeding diatheses, although they can occur in patients of any demographic. They can also be seen in patients with brain infections, inflammatory diseases, and intracranial hypotension. They compress the adjacent brain and can cause midline shift and central downward brain herniation, leading to increased intracranial pressure (ICP), brainstem compression and death if untreated. Subdural hematomas and collections are critical to identify and manage as quickly as possible, since they may bleed, rebleed, and increase in size. Identifying them quickly can significantly improve patient outcomes. In emergency situations, automated identification and classification of subdural hematomas or collections would assist case prioritization and interpretation. In busy private practices, outpatient MRI scans that potentially demonstrate subdural hematomas/collections may not be reviewed for days after completion, delaying management and potentially leading to poor patient outcomes. Automated identification and classification would also provide value to non-specialists, especially those who read images in emergency situations.

Narrative

An 88-year-old woman is seen by her primary care physician on a Friday afternoon; she complains of a worsening headache over the past 4 days. She doesn’t remember any head trauma or falling down, but she did bump her head on a car door a few days back. She has no focal neurological complaints. Her neurological exam is normal. Her physician sends her for an outpatient MRI that afternoon at the closest imaging center.

Workflow Description

Patient receives a MRI scan of their head; the clinical indications can be numerous. Algorithm receives entire MRI dataset. If the algorithm can determine a result, return: size, location (R/L; frontal/parietal/temporal/occipital/parafalcine/cerebellar), intensity of the subdural hematoma/collection. A reliability metric, assessing the accuracy of the size measurement, would also be helpful. An alert should notify the user.
Additional considerations are as follows: Algorithm executes after exam is verified on PACS in the backend. Algorithm optimally integrates on PACS and dictation or reporting software. The user is able to automatically populate the report or manually input the results. Indicator image may save to PACS as part of the medical record.

Inputs

DICOM Study

Primary Outputs

Hematoma Shape/Configuration

Hematoma Laterality

Hematoma Location

Hematoma Width

Hematoma Volume

Hematoma T1 Intensity

Hematoma T2 Intensity

Hematoma Septations