DeepScan

Problem: - Radiology (X-rays, CT, MRI scans) is a routine part of healthcare used to identify pathologies, and devise treatments. -The average time between a CT scan and the radiologist’s final diagnosis report is 5.9 hours. -15.1%–27.4% of brain tumors are misdiagnosed -Increased delays and misdiagnosis in reporting are associated with increased cost, morbidity, and mortality. -Upswing in delayed radiology cases due to increased pressure on services e.g. NHS, which was exacerbated due to the pandemic.

Solution: A python-based program interlinked with EMR in HMS and radiology servers of the hospital that runs an automated brain tumor detection and classification. This would be an algorithm that recognizes the presence or absence of a brain tumor by combining several procedures via deep machine learning.

Impact: - Revolutionary; it’ll shorten the time for referrals and start treatment early hence better prognosis. - Precise; efficient measurement of tumor extent with the percentage confidence of the result - Bridging the gap; decreases the delay between detection and diagnosis of tumor saving us valuable time - Efficient treatment; suggests the best treatment modality based on the extent of the tumor and comorbid - Burden; ultimately helps decrease the burden on the healthcare system

Feasibility & financials: - Bare minimum cost and easy to use since it involves the hospital’s pre-developed EMR and radiology servers - No manual entry as API automatically sends each picture to our web platform once imaging is done A detailed financial plan is attached to the demo video and business plan presentation.

What is new about your solution?: This idea has been theorized but not implemented anywhere, most MRI apps and programs focus on 3D imaging but not the diagnosis. Currently, the applications and software focus on 3D building of better images that can be available remotely yet neither of them have focused on automated detection of tumors or diseases. Researches on this topic are present but a working program has not been implemented till now.

What you have built at the hackathon - text explanation + code (e.g. GitHub link): We have built a brain tumor detection model using convolutional neural networks, a deep learning framework. Where brain MRI images will be input, the model will predict or classify if there is a brain tumor in the specific brain MRI. The YouTube link has the complete video of our code, GitHub link will be provided if requested by the organizers specifically. https://youtu.be/o53_fjXsnj0

What you had before the hackathon, please mention open source as well: This submission was created after the hackathon and after considering the challenges most suitable to us.

What comes next and what you wish to achieve: We would like an incubator to take us on board and help refine our idea, business model We would then look to implement this idea by establishing partnerships with hospitals and diagnostic centers Later on, we will look to add more features: -Aid in trauma management by identification of hemorrhage -Expand to include all modalities like X-Ray, CT, MRI, or PET scan -Expand to fields besides neurology that uses radiological imaging -Treatment regimen tailored to diagnosis -Differential Diagnosis

Video: https://www.youtube.com/embed/oY3ifff3UZc