Essay Sample on Deep Learning: A Crucial Component of Artificial Intelligence

Introduction

Due to the rapid growth rate in the IT sector, various technological advances have become major of interest, for example, deep learning. Generally, deep learning can be defined as a function of artificial intelligence that models the human brain working in creating patterns and processing data for utilization in decision making processes Yamashita et al. (2018). Deep learning is advancing faster and is becoming a critical instrument in the applications of artificial intelligence. For instance, in areas like natural language processing, computer vision, and speech recognition, it has demonstrated an enormous potentiality of showing astonishing results. Among some models of deep learning, CNN (Convolutional Neural Networks) is the most developed algorithm Yamashita et al. (2018). CNN is a classification of artificial neural networks that are dominating in several tasks of computer vision, and it creates significant interests across various domains such as radiology. Deep learning performs stands out mainly in problem-related to image classification. The main functionality of the image classification feature is classifying a particular picture based on different sets of probable categories. From the perspective of deep learning, the problem of image classification can be worked out by the use of transfer learning. This assignment focuses on explaining why a pre-trained CNN model preferred for our application and what results are expected from using a patch-level CNN classifier to extract features with heatmaps that could help bring out hidden findings in mammograms

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Save 25%

One of the commonly used approaches in computer vision is using transfer learning as it allows one to develop an accurate model in less time possible. It works by leveraging learnings done before and avoids the process of starting the learning from scratch. In performing computer vision tasks, transfer learning is expressed typically through the utilization of pre-trained models. A pre-trained model is a model that is trained on a large dataset to offer solutions similar to a specific problem that needs solving. Pre-trained CNN models are usually preferred in the process of transfer learning. For example, in our proposed medical application. This is because of its easiness and high performance in training over the past years. Usually, CNN has two parts; a convolutional base and classifier layer. The main functionalities of these two parts are to generate image features, and image classification based on the features detected, respectively. Repurposing a particular pre-trained CNN model as per your needs (mammography) necessitates the removal of the original classifier layer and put in a new layer fitting your purpose. This increases the model's accuracy, thus improving the quality of CNN decisions. An example of reviewed pieces of literature that provides an in-depth understanding of the use of CNN pre-training models in medical imaging is such as Yamashita et al. (2018) and Sahiner et al. (2019).

Various studies, for example, Fonseca et al. (2015) and Becker et al. (2017), have illustrated the higher potentiality of applying the methods of CNNs through the use of deep learning in mammography. They depicted the potential significance which CNNs had of developing a new predicting risk detector within a short period that has an advanced performance of assessing and detecting an abnormal symptom early from negative mammography. Becker et al. (2017) achieved an AUC of 0.82 in comparison to the AUC of between 0.79-to-0.87 of experienced radiologists by exploring the workflow automatic composition classification of a breast through the use of CNN that featured both extraction and vector machine classifier for support. This study is kind of similar to the proposed medical imaging. Thus, the expected results for our proposed imaging model in comparison to the original data are expected to be between AUC of 0.79-to-0.87, taking into consideration the disadvantages of overfitting and the use of small data sets.

References

Becker, A. S., Marcon, M., Ghafoor, S., Wurnig, M. C., Frauenfelder, T., & Boss, A. (2017). Deep learning in mammography: diagnostic accuracy of a multipurpose image analysis software in the detection of breast cancer. Investigative radiology, 52(7), 434-440. Retrieved from; https://journals.lww.com/investigativeradiology/Abstract/2017/07000/Deep_Learning_in_Mammography__Diagnostic_Accuracy.7.aspx

Fonseca, P., Mendoza, J., Wainer, J., Ferrer, J., Pinto, J., Guerrero, J., & Castaneda, B. (2015, March). Automatic breast density classification using a convolutional neural network architecture search procedure. In Medical Imaging 2015: Computer-Aided Diagnosis (Vol. 9414, p. 941428). International Society for Optics and Photonics. Retrieved from; https://www.spiedigitallibrary.org/conference-proceedings-of-spie/9414/941428/Automatic-breast-density-classification-using-a-convolutional-neural-network-architecture/10.1117/12.2081576.short

Sahiner, B., Pezeshk, A., Hadjiiski, L. M., Wang, X., Drukker, K., Cha, K. H., ... & Giger, M. L. (2019). Deep learning in medical imaging and radiation therapy. Medical Physics, 46(1), e1-e36. Retrieved from; https://aapm.onlinelibrary.wiley.com/doi/full/10.1002/mp.13264

Yamashita, R., Nishio, M., Do, R. K. G., & Togashi, K. (2018). Convolutional neural networks: an overview and application in radiology. Insights into Imaging, 9(4), 611-629. Retrieved from; https://link.springer.com/article/10.1007/s13244-018-0639-9