Hi,
Thanks for releasing MedASR. I tested it in comparison with the Whisper model, and the results were weaker compared to Whisper. I’d like to get your advice on the best way to use it.
I suspect that the input audio file may have a high level of noise, since Whisper has noise cancellation or has been trained on many noisy audio samples. I used greedy decoding with MedASR. This is the result of our test:
result= uhD. looking at patient ID 47 to9 {dash} 03. Brain MRI from May 42025. soo this image shows masses in the right frontal lobe about 3.2 cm wide, raally irregular borders. Signal intensity looks like a high-grade glioma, may be glioblastoma multiform. Because of that heterogeneous enhancement pattern seeing edema around the toumors pushing the midline over about 4 mm left. No bleeding in them. Tumors are heting the cortex going into subcortical white matter, not invading basal ganglia.Vterestes are a bit squed no hydrocephalus. T2 weighted image show pretumoral edema pretty huge. No other tumors in cerebellum or brainstem uptake. She has a pituitary gland look good. Go to the pole gout to pull the march 2025 is sc to growth, could be CNS lymmphoma lylymphoma, but enhancement makes it less likely, soo primary brain tumors, right frontal lobe, probably high-grade glioma. Need neurosurgery to check it Biopsy for short diagnosis. MRI again in six weeks, sooner if symptoms get bad, may be radiation oncology to think I got the midline shift in here.</s>