Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical Care
Introduction Unruptured cerebral aneurysms (UCAs) have a relatively low prevalence of approximately 3%, but detection can prevent devastating consequences of subarachnoid hemorrhage. Here, we assess the performance of a machine‐learning (ML) algorithm to identify UCAs and determine whether routine u...
Main Authors: | , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Wiley
2023-03-01
|
Series: | Stroke: Vascular and Interventional Neurology |
Online Access: | https://www.ahajournals.org/doi/10.1161/SVIN.03.suppl_1.245 |
_version_ | 1797803589516656640 |
---|---|
author | Iman J Ali Sergio A Salazar‐Marioni Rania Abdelkhaleq Arash Niktabe Luca Giancardo Christopher J Love Dan Paz Orin Bibas Omri Segev Sunil A Sheth |
author_facet | Iman J Ali Sergio A Salazar‐Marioni Rania Abdelkhaleq Arash Niktabe Luca Giancardo Christopher J Love Dan Paz Orin Bibas Omri Segev Sunil A Sheth |
author_sort | Iman J Ali |
collection | DOAJ |
description | Introduction Unruptured cerebral aneurysms (UCAs) have a relatively low prevalence of approximately 3%, but detection can prevent devastating consequences of subarachnoid hemorrhage. Here, we assess the performance of a machine‐learning (ML) algorithm to identify UCAs and determine whether routine use of the algorithm would have improved detection rates and patient care. Methods From a prospectively maintained multi‐center registry across 8 certified stroke centers (1 comprehensive, 7 primary), we identified patients who underwent CT angiogram for evaluation of stroke from 3/14/21 – 11/31/21. An FDA‐cleared convolutional deep neural network (Viz ANEURYSM, Viz.ai, Inc.) trained to identify UCAs at least 4mm analyzed the images. Ground truth was provided by independent expert neuroradiology read. The primary outcome was rate of UCAs detected by the ML algorithm but not detected or addressed in the clinical radiology report or clinical notes, which was determined by two independent researchers. Results Among 1191 CT angiogram scans performed during the study period, 49 were flagged by the ML algorithm as possibly demonstrating an UCA, of which 26 cases were confirmed as true positive (PPV 53%).The most common locations included posterior communicating artery (22%), followed by MCA bifurcation (19%). Of these cases, 9 (35%) were not noted in the clinical radiology report or clinical notes, with a median size of 4.2 mm [IQR 3–7.5 mm], and 22 (85%) were not referred for follow up, with median size of 5 mm [IQR 3.7‐11.3 mm]. Of the 22 cases not referred for follow up, 13 (59%) had been noted in the radiology report. 46% (6/13) of the detected but not referred cases had a diameter greater than 10mm. Conclusions UCAs of sizes and intra‐dural locations that may warrant treatment are frequently missed or not followed up in routine clinical care. An ML algorithm that flags studies and notifies clinicians may minimize missed treatment opportunities. |
first_indexed | 2024-03-13T05:23:07Z |
format | Article |
id | doaj.art-386bb27d6bff431896b2f67bc88f33a2 |
institution | Directory Open Access Journal |
issn | 2694-5746 |
language | English |
last_indexed | 2024-03-13T05:23:07Z |
publishDate | 2023-03-01 |
publisher | Wiley |
record_format | Article |
series | Stroke: Vascular and Interventional Neurology |
spelling | doaj.art-386bb27d6bff431896b2f67bc88f33a22023-06-15T10:40:48ZengWileyStroke: Vascular and Interventional Neurology2694-57462023-03-013S110.1161/SVIN.03.suppl_1.245Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical CareIman J Ali0Sergio A Salazar‐Marioni1Rania Abdelkhaleq2Arash Niktabe3Luca Giancardo4Christopher J Love5Dan Paz6Orin Bibas7Omri Segev8Sunil A Sheth9Department of Neurology UTHealth McGovern Medical School Houston Texas United States of AmericaDepartment of Neurology UTHealth McGovern Medical School Houston Texas United States of AmericaDepartment of Neurology UTHealth McGovern Medical School Houston Texas United States of AmericaDepartment of Neurology UTHealth McGovern Medical School Houston Texas United States of AmericaUTHealth School of Biomedical Informatics Houston Texas United States of AmericaViz.ai San Francisco California United States of AmericaViz.ai San Francisco California United States of AmericaViz.ai San Francisco California United States of AmericaViz.ai San Francisco California United States of AmericaDepartment of Neurology UTHealth McGovern Medical School Houston Texas United States of AmericaIntroduction Unruptured cerebral aneurysms (UCAs) have a relatively low prevalence of approximately 3%, but detection can prevent devastating consequences of subarachnoid hemorrhage. Here, we assess the performance of a machine‐learning (ML) algorithm to identify UCAs and determine whether routine use of the algorithm would have improved detection rates and patient care. Methods From a prospectively maintained multi‐center registry across 8 certified stroke centers (1 comprehensive, 7 primary), we identified patients who underwent CT angiogram for evaluation of stroke from 3/14/21 – 11/31/21. An FDA‐cleared convolutional deep neural network (Viz ANEURYSM, Viz.ai, Inc.) trained to identify UCAs at least 4mm analyzed the images. Ground truth was provided by independent expert neuroradiology read. The primary outcome was rate of UCAs detected by the ML algorithm but not detected or addressed in the clinical radiology report or clinical notes, which was determined by two independent researchers. Results Among 1191 CT angiogram scans performed during the study period, 49 were flagged by the ML algorithm as possibly demonstrating an UCA, of which 26 cases were confirmed as true positive (PPV 53%).The most common locations included posterior communicating artery (22%), followed by MCA bifurcation (19%). Of these cases, 9 (35%) were not noted in the clinical radiology report or clinical notes, with a median size of 4.2 mm [IQR 3–7.5 mm], and 22 (85%) were not referred for follow up, with median size of 5 mm [IQR 3.7‐11.3 mm]. Of the 22 cases not referred for follow up, 13 (59%) had been noted in the radiology report. 46% (6/13) of the detected but not referred cases had a diameter greater than 10mm. Conclusions UCAs of sizes and intra‐dural locations that may warrant treatment are frequently missed or not followed up in routine clinical care. An ML algorithm that flags studies and notifies clinicians may minimize missed treatment opportunities.https://www.ahajournals.org/doi/10.1161/SVIN.03.suppl_1.245 |
spellingShingle | Iman J Ali Sergio A Salazar‐Marioni Rania Abdelkhaleq Arash Niktabe Luca Giancardo Christopher J Love Dan Paz Orin Bibas Omri Segev Sunil A Sheth Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical Care Stroke: Vascular and Interventional Neurology |
title | Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical Care |
title_full | Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical Care |
title_fullStr | Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical Care |
title_full_unstemmed | Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical Care |
title_short | Abstract Number ‐ 245: Machine Learning‐Enabled Detection of Unruptured Cerebral Aneurysms Improves Detection Rates and Clinical Care |
title_sort | abstract number 245 machine learning enabled detection of unruptured cerebral aneurysms improves detection rates and clinical care |
url | https://www.ahajournals.org/doi/10.1161/SVIN.03.suppl_1.245 |
work_keys_str_mv | AT imanjali abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT sergioasalazarmarioni abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT raniaabdelkhaleq abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT arashniktabe abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT lucagiancardo abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT christopherjlove abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT danpaz abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT orinbibas abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT omrisegev abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare AT sunilasheth abstractnumber245machinelearningenableddetectionofunrupturedcerebralaneurysmsimprovesdetectionratesandclinicalcare |