In traditional methods, cataracts are mainly detected through fundus images, where image acquisition is costlier and needs experts to handle the fundus cameras. The AI based solution with low-cost imaging devices can make it more accessible and inexpensive
A team of researchers from the Indian Institute of Technology Jodhpur has found that eye images acquired by low-cost near-infrared (NIR) cameras can aid in low design costs, ease of use, and practical solutions for cataract detection. Known as MTCD, the proposed multitask deep learning algorithm is inexpensive and results in very high levels of accuracy.
This research presents a deep learning-based cataract detection method that involves iris segmentation and multitask network classification. The proposed segmentation algorithm efficiently and effectively detects non-ideal eye boundaries. It is also cost-effective as low-cost NIR cameras are used in place of costly ophthalmoscopes. The proposed method can be used in rural settings where the availability of doctors is limited.
Highlights of the research are:
- Process is automated using multitask deep learning AI algorithm.
- Proposed cataract detection method uses eye images captured in a nearinfrared domain.
- The method is computationally inexpensive and yields high accuracy.
In traditional methods, cataracts are mainly detected through fundus images, where image acquisition is costlier and needs experts to handle the fundus cameras. The AI based solution with low-cost imaging devices can make it more accessible and inexpensive.
The research was conceptualised by Dr Mayank Vatsa and Dr Richa Singh from the Image Analysis and Biometrics (IAB) Lab at IIT Jodhpur. They were supported by various UG and Ph.D. students at the lab – Mahapara Khurshid, Yasmeena Akhter, Rohit Keshari, Pavani Tripathi, and Aditya Lakra.
Speaking about the research, Dr Singh, Professor, Department of Computer Science and Engineering, IIT Jodhpur, said, “Currently, a large number of patients with cataracts have to visit secondary and tertiary care centres. The availability of such a solution can assist doctors at the primary health care centres in helping such patients.”.
Dr Vatsa, Professor, Department of Computer Science and Engineering, IIT Jodhpur, further added, “We are extending this research to include both cataract and diabetic retinopathy in the solution and have collaborated with multiple hospitals in the country for domain expertise, data collection, and validation of the solution.”.
IHub-Drishti, TIH at IIT Jodhpur, has recently funded the next stage of this research. The researchers plan to undertake an extensive data collection exercise for building an ophthalmology databank with different kinds of devices. The second part improves the approach and creates an explainable and robust AI algorithm for cataract detection. IIT Jodhpur has collaborated with the Postgraduate Institute of Medical Education and Research, PGIMER Chandigarh, to further expand this research.