AutoCom: Brain Tumor Detector

CATEGORY: Health

Log in or register to follow or vote for this project.

Developing next generation MRI Viewers for automatic detection and analysis of Brain Tumors, empowering the current diagnosis with reduced manual errors and increased efficiency while saving time and money.

Standings & Awards

41 out of 1313 in Asia
21 out of 284 in Health
34 out of 573 in Pilot
43 out of 992 in Charitable
113 out of 4003 Overall

3

VOTES

Neurons

Empowering common man with fast, economical and reliable automatic diagnostic method for Brain Tumors

Every year thousands of people die around the globe as a result of different brain tumors. Some due to human incapability because of large variations in size, location and form of the Brain Tumors, while some due to human errors because of increasing number of Neuro-patients leading to a huge manual workload on small Radiology group.

This inspired us to develop a tool which can assist radiologist by automatically detecting Brain Tumors in MRI images and thus help in Saving Time, Saving Money and Saving Radiologist for more complex and expertise requiring cases.

Current Situation:

Idea/ Flow
To come up with next generation MRI viewers which can assist the radiologists by automatically detecting Tumor, if present and generate report based on the tumor found.

Input : Like a normal Dicom Viewer, it loads ‘n’ patient cases with 20 slices(MRI images) each of T2, T1,T1 post contrast,etc sequences.
Output : Divide the loaded ‘n’ number of cases into three categories and generate report for each patient case automatically.
The three categories are :

Prototype : We have been working on this Idea from summer’11 and have developed a Prototype for testing the above made algorithm. Testing dataset so far, consist of 120 patients (each patient data consists of 20 images each of 3 sequences T1, T2 and T1 post contrast) and is gradually expanded. Out of 120, 65 were Normal and 55 abnormal (30 Tumor containing).

Achievements and Results:

Prototype has been tested on 120 patients dataset (each patient data consists of 20 images each of 3 sequences T1, T2 and T1 post contrast) from two MRI centres of New Delhi. Out of 120, 65 were Normal and 55 abnormal (30 Tumor containing). The AutoCom prototype showed excellent results differentiating normal-abnormal as well as identifying Tumor and generating report, when found.
Algorithm published as "Automatic Detection of Brain Abnormalities and Tumor Segmentation in MRI Sequences" at Twenty-sixth International Conference Image and Vision Computing New Zealand (IVCNZ 2011 ) , Auckland, New Zealand, Dec 2011
Awarded Summer Undergraduate Research Award, SURA , IIT Delhi
AutoCom Secured 3rd Position at National Innovation Award, Techtop-2012 at Trivandrum, Kerela
AutoCom Secured 2nd position in IKES-2012 organized by IBM and ACM IIT Delhi Student Chapter
Semi-Finalist during Dell Social Innovation Award, 2012.

Link to AutoCom Webpage : www.cse.iitd.ac.in/~cs5090255/autocom/index.html

Support Your Idea Optional (5 - 7 minutes for three uploads)

Supporting Document:

The paper has been presented and published in the International Conference on Image and Vision Computing New Zealand, IVCNZ-2011

Roadmap to Success Optional (1 - 3 minutes to upload)

Roadmap to Success:

roadmap_to_success.pdf

Sponsors, Investors, and Supporters

FIVE PROJECT QUESTIONS Required (60 - 90 minutes)

1. What is your innovation?

AutoCom is a next generation MRI Viewer for automatic detection and report generation of Brain Tumors. It has huge impact on thousands of people suffering from Brain Tumors by empowering the current diagnosis with reduced manual errors, reduced delays due to unavailability of an expert radiologist or time taking manual process and increased efficiency while saving money. It also eases pressure on overloaded Radiology group of the world due to ever increasing number of patients by saving time.

2. Who gains the most?

The spectrum of people benefited by this next generation Dicom Viewer, AutoCom is spanned through the entire globe. Thousands of people in different parts of the world suffering from Brain Tumor would get an efficient, error less diagnosis. Also as being automatic, it helps common man in reducing delays in treatment due to unavailability of radiologists, especially in developing countries. It also helps the overloaded Radiology group of the world by reducing manual work and saving time.

3. Who pays?

Ever increasing number of Brain Tumor cases and huge workload on Radiology Group of the world make the need of AutoCom inevitable. It would affect entire class of MRI centers and research institutes in different regions of world who would be the customers of AutoCom. Although, initially, rather than monetary, we want to have collaboration with them so that we can better understand the diagnostic flow that a radiologist follows and properties of tumor with access to a huge database for testing.

4. What is your success?

Working from summer'11 on the backend algorithm, we have already developed initial prototype of AutoCom which has been tested on 120 cases and efficiently separates normal and abnormal cases. By next year, we would be further refining and testing it for final integration in an open source Dicom viewer. In 3 years, we would be increasing the intelligence of this AutoCom to identify Brain Tumors efficiently. By 5 years,we would integrate tools for auto-analysis of Brain's CSF,white and grey matter

5. How will you do it?

The development of AutoCom requires knowledge of both medical image analysis techniques and of brain anatomy and thus collaboration is important. In our initial development of backend algorithm, we collaborated with local MRI centers for obtaining dataset and understanding tumor properties. Now we will work with MRI centers and Research labs in Brain anatomy and Medical image analysis in different parts of world for mentoring and collaboration to develop automatic brain tumor detector.