Lumbar Spine

Abstract

Many people as they age deal with lumbar spine degenerative illnesses, which are somewhat frequent. These disorders include the degradation of intervertebral discs, vertebrae, and other elements of the spine that causes issues including subarticular stenosis, neural foraminal constriction, and spinal canal stenosis. For clinicians to grasp and identify these problems, magnetic resonance imaging (MRI) among other techniques is essential. This work aims to create an image recognition system for lumbar spine MRI data to identify and classify degenerative disorders. Different methods applied for identifying and categorizing spinal anomalies from the L1 to L5 vertebrae are investigated in this work. The project consists in several phases of medical image processing covering preprocessing, segmentation, feature extraction, and condition categorization into normal, moderate, or severe categories. Python is used for the last implementation, training and testing using a freely available Kaggle dataset. Next.js is designed front-end with FastAPI integrated to provide seamless interaction. This project runs YoloV8 and MobileNet-V3 with gated attention since its lightweight character makes it perfect for running on CPUs and GPUs. Using Next.js, we have created a website whereby users may upload MRI images for examination. By spotting damaged or aberrant spinal locations, the method offers precise results and insight into disorders including Left Neural Foraminal Narrowing, Right Neural Foraminal Narrowing, Left Subarticular Stenosis, Right Subarticular Stenosis, and Spinal Canal Stenosis. A chatbot is also included into the website to help consumers with lumbar spine related questions. The process starts with MRI picture preprocessing including grayscale conversion, scaling, contrast enhancement. Then the lumbar spine area from L1/ L2 to L5/S1 is segmented with masks or bounding boxes. Following segmentation, the images feed the model for feature extraction and classification. The result is a probability matrix showing the degree and kind of disorder at every spinal level. The paper also offers evaluation measures, graphic findings, difficulties faced, and suggestions for next development.