These days Machine Learning has been very much useful in Healthcare sector. In particularly, our proposed framework focuses on one of the most common condition known as Deep Vein Thrombosis(DVT). Deep vein blood clots typically form in your thigh or lower leg, but they can also develop in other areas of your body. Deep vein thrombosis (DVT) is a blood clot commonly found in deep veins of the lower extremities. Deep vein thrombosis (DVT) is a serious condition that occurs when a blood clot forms in a vein located deep inside your body. A blood clot is a clump of blood that's turned to a solid state. Every year, 60,000-100,000 Americans die of complications arising from DVT. Patients who get CT scans are frequently discharged before a radiologist looks at the scan. Early and automated detection is critical for lowering fatalities and in regions with few radiologists. In our proposed system, the input image is trained through several layers. Initially, the input image is subjected to canny feature detector. After feature detection the feature mapissent to the convolution layer where Rectified Linear Unit (ReLu) activation function is used. The second layer which we use is the max pool layer where the maximum value from each of the feature map obtained from previous layer is detected. The third layer is the flatten layer where the feature map is reduce to 1D. The feature maps obtained through all these layers are connected in the final layer which is Fully Converted layer. Also Dropout layer is used to avoid over fitting. Unet architecture is found to give more accuracy than any other traditional models and also it is one of the modern architectures employed in Machine Learning Domain.
Key words: Images, Modern, Neural, Network ,Technique etc
|
