Importance of Deep Learning Models in the Medical Imaging Field

Introduction

The brain is the main controller organ of the body that regulates the functioning of other organs. The central organ of the nervous system is the brain that is protected by skull bones. The main parts of brain structure are the brainstem, cerebrum, and cerebellum shown in Figure 1.The brain collects, organizes, and distributes information throughout the body. The main functions of the brain are to process the information collected by sense organs, regularization of breathing and blood pressure, and release hormones in the body. A healthy brain works fast and spontaneously. Any kind of problem inside the brain can make the working of the body difficult. The complex structure of the brain makes it difficult to find any kind of abnormality in it. It requires a lot of experience and precise knowledge to diagnose issues in the brain.

Medical Imaging applications provide health professionals detail and fast information about the internal organs and tissues of patients. There are so many types of medical imaging technologies that give information about different body parts. Medical imaging techniques help diagnose diseases like pneumonia, brain injuries, cancer, internal bleeding, and many other issues. Large amount of data is generated by medical imaging applications. High quality of imaging can advances medical decision making and can lessen the unnecessary procedures. Anatomical and physiological database are created with help of these techniques. Timely diagnosis of disease plays a very important part for recovery of patient from disease. For early diagnosis and other information that needed of patient’s body parts, medical imaging plays significant role here. Medical imaging procedures like X-ray images, Computed Tomography (CT) scans, and MR images are supporting medical teams in diagnosing disease or any kind of abnormality in distinct body parts of patients to determine what procedure should be followed for an early recovery. The Magnetic Resonance Imaging (MRI) technique is practiced in radiology for investigating the human body through MR scans. MR images give specific information of internal tissues that can help in the diagnosis of disease, injury or tumor, etc. The powerful magnetic field is applied to produce pictures through an MRI scanner. MRI-generated pictures are more detailed compare to other techniques, like Ultrasound or CT.

Digital Image processing methods like image classification and segmentation are extremely important techniques to follow in the image processing field. Segmentation is the method of generating distinct sets of pixels in an image with pixels sharing common characteristics included in the same set. In many cases, the entire image is not required for additional processing that is why with the help of the segmentation method the region of interest (ROI) can be acquired within the targeted image, and the useless part is discarded. The complex structure of the brain makes it’s quite challenging for detailed segmentation of the brain region for the investigation of pathological tissues. Researchers are attempting to develop automatic image segmentation and classification models to assist the medical team in finding the appearance of any sort of abnormality in scanned images. The feature of self-leaning and fast processing of deep learning technology is raising its use in several fields including the medical field too. There has been a lot of work accomplished on segmentation of brain region and brain tumor detection on brain MRI scans and it’s still attracting many researchers to extend their study in this field and develop models with high efficiency. The employment of deep learning models is increasing the performance levels on segmentation of brain MRI in multiple measurement aspects and defeating the performance of many machine learning models.

Figure 1.

Three main parts of brain

This chapter explicates the significance of deep learning methods in neuroimaging area. This chapter covers previous studies that have been conducting to segment the brain region or detect brain tumor, using deep learning models while MRI scan data sets are common in all elected studies. This chapter also presents a skimpy knowledge of digital image processing, brain imaging techniques and, deep learning technique. This study encourages the readers to know how deep learning models can perform well in image segmentation tasks by examining the previously done works and to provide contribution in the medical imaging field.