Prediction in Medicine: The Impact of Machine Learning on Healthcare

A Scientific Implementation for Medical Images to Detect and Classify Various Diseases Using Machine Learning

Author(s):

Pp: 248-270 (23)

DOI: 10.2174/9789815305128124010016

* (Excluding Mailing and Handling)

Abstract

Reconstruction of medical images is imperative for the comprehension of clinical anomalies. Various processes and techniques are employed to generate efficient anatomical representations of the human body in medical imaging. This technique provides physicians with a visual depiction of internal organs, aiding in the verification of ongoing treatments, accurate diagnoses, and treatment planning. Medical imaging encompasses diverse methods such as ultrasound, X-rays, MRI, and CT scans, with the choice depending on the condition of the ailment, such as kidney stone diseases, breast cancer, and brain tumors. However, the quality of medical images can becompromised due to different sources of noise and blurriness. This chapter introduces an advanced image processing methodology to detect diseases in medical images, particularly brain tumors, kidney stones, and breast cancer using ultrasound and MRI images. The proposed approach involves converting RGB medical images into grayscale, removing labels, and adjusting image intensity to enhance the contrast of biomedical images. Median filtering is applied to eliminate noise, and the Discrete Wavelet Transform (DWT) is utilized for brain tumor detection. The filtered medical image output is subjected to morphological and k-means clustering segmentation. To classify the images into two categories benign and malignant, Convolutional Neural Network (CNN) classifiers are employed. The final system analysis involves evaluating, specificity, accuracy, and sensitivity through the preparation of a confusion matrix. The classification system demonstrates an accuracy of approximately 95%. This presented technique holds potential in supporting doctors with early detection for precise patient treatment. 

© 2024 Bentham Science Publishers | Privacy Policy