M.S. AAI Capstone Chronicles 2024

CNN Lung Disease Classification

17

Conclusion Model Performance and Problem Application While the results in this research show promise, they also indicate that further improvements are necessary before the model can be reliably applied in clinical settings for chest X-ray classification. The current performance level suggests that our model could potentially assist radiologists by providing an initial screening or a second opinion, but it is not yet sufficiently accurate for autonomous diagnosis. Deep learning models for medical imaging need to achieve high accuracy, and our dataset only consists of frontal radiographs and as Rajpurkar et al. (2017) noted, up to 15% of accurate diagnosis

175