Learning Objectives:
To understand that deep learning computer-aided detection (DL-CAD) software can address some limitations of the MRI prostate cancer diagnosis pathway
To know the tasks where DL software can be helpful including accelerating data acquisitions, tumour detections and lesion/gland outlining
To realize that harm from false positives (alarms) and false negatives (false reassurance) of DL software are like those of practising radiologists
SOP for using AI software:
Account for clinical history, PSA & DRE (F/H, LUTS, infection, ADT, finasteride)
Image quality check (SNR & rectal gas impacts on T2W/DWI)
Check gland & TZ outlines - lesions may be outside outline (be aware at the base posteriorly)
Review ROIs identified (likely true detections)
Review heatmaps without defined ROIs (removed false detections)
Common false detections - infection, stromal nodules, central zone (asymmetric); wrap-in artefacts
Review AI blind spots (small lesions, base, subcapsular regions, outside gland) - potential misses
Classify lesions using bpMRI (± use DCE if available); Level of Suspicion (LoS) score & assign PI-RADS & LIKERT scores
Adjust tolerance to false detections by care priority use history/PSAD/risk factors/patient preferences
Export Dicom RT objects for fusion Bx and/or RT treatment planning
Autogenerate report
Негізгі бет Reading Prostate MRI with AI Software Assistance
Пікірлер