An Efficient Fine-tuned GoogleNet Model for Multiclass Classification of Blood Cell Cancer

Abstract

In this study, a novel use of the GoogleNet model to detect blood cell cancer is used. The results are highly impressive--the accuracy was as high as 99.83 %. Drawing power from deep learning abilities, mainly the characteristics of the GoogleNet architecture, our model shows unprecedented levels of precision at cell-level scanning for cancerous cells in blood samples. This remarkable accuracy is supplemented by a loss of just 0.9148 at epoch 15, which serves to highlight the strength and soundness of this approach. The research includes a broad-based study of blood cell images, including various forms and phases of cancer. The GoogleNet model is exquisitely honed, with an exceptional ability to perceive fine patterns such as those associated with blood cell cancers. The performance of the model is accordingly rigorously tested, achieving as high a level of accuracy for testing purposes means that it has the potential to serve as an effective diagnostic tool. In particular, the relatively low value of loss at epoch 15 shows that the model has converged and done a good job pulling out features in learning from this data. The accuracy rate is a record-low 99.83 %, putting the GoogleNet at forefront in the world of blood cell cancer diagnosis, enabling high precision early detection of hematologic malignancy. In addition to providing assistance in the development of more effective diagnostic approaches within oncology, this research further highlights how deep learning has begun changing health care. The results here offer paths towards using such models in clinical practice, which should increase the accuracy of diagnosis and timeliness of interventions for individuals with blood cell cancers.

Topics & Concepts