Pulmonary Nodules Classification In Computed Tomography Images Using Faster R-CNN

Abstract

Pulmonary nodules are the primary symptom of lung cancer, which has a high mortality rate anywhere in the globe. Radiologists are able to spend less time on false positives and missed diagnoses because to automatic pulmonary nodule identification. We suggest using a more efficient variant of the Faster R-CNN algorithm to identify these pulmonary nodules in early stages. A more efficient Faster R-CNN algorithm is able to identify pulmonary nodules, and this is shown by using the training set. Theoretically, modification of the parameters involved in a network may lead to both structural and detection improvements. This research proposes an enhanced and optimized approach for identifying pulmonary nodules, which, on average, increases detection accuracy by over 20% compared to previous, more conventional techniques. The Faster R-CNN-based technique of pulmonary nodule detection demonstrated high accuracy in our experiments, suggesting it may have practical use in the diagnosis of pulmonary illness. This technique may be of additional use to radiologists and academics working on improving the pulmonary nodule’s detection lesion.