Hybrid neural network for non-image-based knee osteoarthritis prediction

Abstract

Osteoarthritis (OA) of the knee is a common degenerative joint condition that adversely affects millions of people worldwide. Early detection and forecasting risks of knee OA can help with prompt interventions and individualized treatment plans to slow the disease's progression. 2 million patients' worth of Taiwanese data was sampled (2001–2015). 1,068,464 comprised control subjects, while 132,594 patients had Knee Osteoarthritis. Over the course of a three-year period, we sequentially used diagnoses, medication, age, and sex to build a feature matrix. In this study, we developed a risk prediction model using a hybrid strategy (CNN-FFNN) that combines a convolutional neural network (CNN) and a feed-forward neural network (FFNN). The performance of the hybrid approach is also evaluated using a variety of performance indicators. Age and sex were excluded from the list of significant disease variables, and drugs like antacids, cough relievers, and stimulants demonstrated discriminative efficacy. The proposed methodology may help medical personnel identify people who are most at risk of getting knee OA, allowing for preventative measures and individualized treatment regimens to lessen the impact of this crippling ailment.