Inspiration
Diagnosis by radiologists could be faulty, henceforth they need years' experience and tuning, which remains still susceptible to erroneous diagnosis incurring increased healthcare cost or patient problems - the "sensitivity-specificity" trade-off
What it does
The future is to develop cost-effective algorithms that will automate the diagnosis of medical images by intelligent systems and yet at the same time perform better than the current gold standard (i.e., diagnosis by radiologists)
How I built it
Try to explore current state-of-the-art in automated medical imaging diagnosis and available software packages / modules as a kick-off start. Next step is to try out various algorithms and validate them and measure their performance while tuning the optimal parameters
Challenges I ran into
Learning the most useful packages / modules proven for medical imaging and how they work; Obtaining data (of images) that have a stamped gold standard diagnosis to compare with
Accomplishments that I am proud of
I have a Master II & a PhD in medical imaging and telemedicine (bio-medical engg.)
What I learned
DSP, different medical imaging techniques (e.g., MRI-NMR, ultrasound/elastography, radionuclear imaging, microscopic imaging), model validation, cross-validation, parameter optimisation, statistical methods, artificial intelligence project on "smart bedrooms" for the elderly
What's next for Medical_Imaging.ai
- Establish a roadmap for 2020
- Estimates of resources, timelines, deliverables, budget
Log in or sign up for Devpost to join the conversation.