Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2503.10887 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916652387401728 |
|---|---|
| author | Mondal, Parmita Bhurwani, Mohammad Mahdi Shiraz Nagesh, Swetadri Vasan Setlur Lai, Pui Man Rosalind Davies, Jason Levy, Elad Vakharia, Kunal Levitt, Michael R Siddiqui, Adnan H Ionita, Ciprian N |
| author_facet | Mondal, Parmita Bhurwani, Mohammad Mahdi Shiraz Nagesh, Swetadri Vasan Setlur Lai, Pui Man Rosalind Davies, Jason Levy, Elad Vakharia, Kunal Levitt, Michael R Siddiqui, Adnan H Ionita, Ciprian N |
| contents | Bias from contrast injection variability is a significant obstacle to accurate intracranial aneurysm occlusion prediction using quantitative angiography and deep neural networks . This study explores bias removal and explainable AI for outcome prediction. This study used angiograms from 458 patients with flow diverters treated IAs with six month follow up defining occlusion status. We minimized injection variability by deconvolving the parent artery input to isolate the impulse response of aneurysms, then reconvolving it with a standardized injection curve. A deep neural network trained on these QA derived biomarkers predicted six month occlusion. Local Interpretable Model Agnostic Explanations identified the key imaging features influencing the model, ensuring transparency and clinical relevance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_10887 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Minimizing Human-Induced Variability in Quantitative Angiography for Robust and Explainable AI-Based Occlusion Prediction Mondal, Parmita Bhurwani, Mohammad Mahdi Shiraz Nagesh, Swetadri Vasan Setlur Lai, Pui Man Rosalind Davies, Jason Levy, Elad Vakharia, Kunal Levitt, Michael R Siddiqui, Adnan H Ionita, Ciprian N Medical Physics Bias from contrast injection variability is a significant obstacle to accurate intracranial aneurysm occlusion prediction using quantitative angiography and deep neural networks . This study explores bias removal and explainable AI for outcome prediction. This study used angiograms from 458 patients with flow diverters treated IAs with six month follow up defining occlusion status. We minimized injection variability by deconvolving the parent artery input to isolate the impulse response of aneurysms, then reconvolving it with a standardized injection curve. A deep neural network trained on these QA derived biomarkers predicted six month occlusion. Local Interpretable Model Agnostic Explanations identified the key imaging features influencing the model, ensuring transparency and clinical relevance. |
| title | Minimizing Human-Induced Variability in Quantitative Angiography for Robust and Explainable AI-Based Occlusion Prediction |
| topic | Medical Physics |
| url | https://arxiv.org/abs/2503.10887 |