Researchers at the Indian Institute of Technology (IIT) Guwahati have successfully developed a groundbreaking framework called OsteoHRNet, utilizing Deep Learning (DL) techniques, to automatically evaluate the severity of Knee Osteoarthritis (OA) from X-ray images. This innovative AI-based model has the potential to detect and assess the level of the disease, enabling remote assistance to medical professionals for more precise diagnoses. The IIT Guwahati team has been diligently working on enhancing the automatic detection of knee osteoarthritis from X-ray images to aid in clinical evaluations. Their approach involves the development of an AI-based model specifically designed to assess the severity of Knee OA.

Unlike conventional deep learning models, this unique AI-based model employs an efficient Deep Convolutional Neural Network (CNN) algorithm, primarily used in image recognition tasks. By utilizing this model, knee OA severity can be predicted according to the Kellgren and Lawrence (KL) grading scale, approved by the World Health Organization (WHO), which ranges from grade 0 (low severity) to grade 4 (high severity). To capture the multi-scale features of knee X-rays, the model is built upon the High-Resolution Network (HRNet), one of the most recent and effective deep learning models.

Dr. Palash Ghosh, Assistant Professor in the Department of Mathematics at IIT Guwahati, explains, "Compared to other approaches, our model can pinpoint the area that is medically most significant in determining the severity level of knee osteoarthritis. This capability assists medical practitioners in accurately detecting the disease at an early stage."

Knee Osteoarthritis is a prevalent musculoskeletal disorder worldwide, affecting 28 percent of the population in India. Unfortunately, there is currently no definitive cure for Knee OA, except for total joint replacement in advanced stages. Consequently, early diagnosis plays a crucial role in pain management and corrective interventions. While MRI and CT scans provide detailed 3D images for effective diagnosis, their availability and affordability are limited. Conversely, X-ray imaging is a cost-effective and widely accessible method for routine diagnosis.

Prof. Arijit Sur from the Department of Computer Science and Engineering at IIT Guwahati explains, "Although our proposed model is simple, it serves as a promising starting point for analyzing inexpensive radiographic modalities like X-rays. Our ongoing research focuses on designing efficient Deep Learning models to work with low-resolution radiographic images or even photographs taken from radiographic plates using a smartphone."

The team is committed to reconfiguring these models to enable deployment on resource-constrained devices, ensuring that medical professionals can easily obtain initial but accurate diagnostic assessments. This endeavor has the potential to address the shortage of skilled personnel in this field, particularly in rural areas of India. The research has been accepted for publication in the journal Multimedia Tools and Applications and was carried out by Mr. Rohit Kumar Jain, a recently graduated MTech Data Science student, under the joint supervision of Prof. Arijit Sur and Dr. Palash Ghosh. The research team also includes former Ph.D. students of Prof. Sur at IIT Guwahati, Dr. Prasen Kumar Sharma, and Dr. Sibaji Gaj.