Unmet Need
The typical workflow for reporting quantitative data in Picture Archive and Communication System (PACS), is redundant, subjective, time-consuming, and hard to record. Automated radiological reports describing consistent lesion features such as location, contrast, and volumes, and related effects is a time-saver, particularly in acute conditions and in those that require quantitative information at real time, such as acute brain strokes. In addition, automated reports produce text-structured information that would, in future, reduce the challenges of natural language processing (NLP) and other artificial intelligence (AI) applications in medical analysis.
In addition to the stroke volume, other clinically relevant indices for acute treatment are extracted from brain images. The Alberta Stroke Program Early CT Score (ASPECTS) is a visual evaluation system to assess the extent and location of ischemic core in patients with acute strokes. Due to its relative simplicity of assessment, ASPECTS gained popularity and was also adapted to diffusion weighted MRIs. However, the capability of ASPECTS for selecting patients’ treatment is debatable. A plausible reason might be the arbitrariness in human visual evaluation, especially when done by readers with less experience. There remains a clinical unmet need for a robust approach that guides treatment selection for patients suffering from acute strokes.
Technology Overview:
Johns Hopkins researchers have generated a machine learning approach to generate automated radiological reports and the Alberta Stroke Program Early CT Score (ASPECTS) comparable to expert readings on diffusion weighted MRIs of patients with acute stroke. This approach outputs in real time the proportion of diverse brain structures and vascular regions affected by an infarct, calculates ASPECTS, and provides results that are comprehensive, showing the extracted features involved in the machine learning classification.
Stage of Development:
Testing and development complete with user-friendly software work on-going.
Publication:
Liu, CF., Li, J., Kim, G. et al. Automatic comprehensive aspects reports in clinical acute stroke MRIs. Sci Rep 13, 3784 (2023).
https://doi.org/10.1038/s41598-023-30242-6
Liu, CF., Zhao, Y., Yedavalli, V. et al. Automatic comprehensive radiological reports for clinical acute stroke MRIs. Commun Med 3, 95 (2023)
https://doi.org/10.1038/s43856-023-00327-4
