25 Mumu Aktar Computer Vision News Stroke, a major global health concern, often requires endovascular thrombectomy for effective treatment. The success of this therapy, however, depends on factors like collateral circulation, a radiologic surrogate predicting revascularization response (Figure below). Currently, collateral circulation assessment relies on visual inspection by a radiologist which suffers from inter and intrarater variability leading to inefficient and time-consuming results. Mumu Aktar defended successfully her Ph.D. dissertation in Computer Science at Concordia University. Congrats Doctor Mumu! Her research, supervised by Dr. Marta Kersten-Oertel and Dr. Hassan Rivaz, focused on developing machine learning methods to improve treatment decision-making in stroke. From left to right, an example of good, intermediate, and poor collaterals on contrast-enhanced CTA. The blue arrow indicates the occlusion on the MCA. Mumu's research involved developing computer-aided methods to enhance collateral circulation assessment in ischemic stroke. Using deep learning aligned with radiologists' criteria, the approaches offer a more robust evaluation than traditional methods. Overcoming the challenge of small and imbalanced datasets in ischemic stroke, the research represents an advancement in improving the accuracy and reliability of collateral scoring to contribute to better treatment decision-making for stroke patients. Over the course of the Ph.D., Mumu developed several computer-aided decision support algorithms for collateral evaluation using a 4D CTA dataset, considering two key phases: (1) 2D images from 3D MIPs of the 4D CTA and (2) NCCT extracted from the 4D CTA before the contrast agent.
RkJQdWJsaXNoZXIy NTc3NzU=