Sayali Kandarkar

According to the World Health Organization, approximately 1 million children die each year due to preterm birth complications (1). Many survivors face a lifetime of disability, including learning disabilities and visual and hearing problems. Inequalities in survival rates around the world are stark. In low-income settings, half of the babies born at or below 32 weeks (2 months early) die due to a lack of feasible, cost-effective care, such as warmth, breastfeeding support, and primary care for… Show more

According to the World Health Organization, approximately 1 million children die each year due to preterm birth complications (1). Many survivors face a lifetime of disability, including learning disabilities and visual and hearing problems. Inequalities in survival rates around the world are stark. In low-income settings, half of the babies born at or below 32 weeks (2 months early) die due to a lack of feasible, cost-effective care, such as warmth, breastfeeding support, and primary care for infections and breathing difficulties. India is the country with the most significant number of preterm births (2). One of the reasons could be the lack of doctors to examine the medical conditions of pregnant women manually. Over 44% of WHO Member States in India reported less than one physician per 1,000 population (3). Hence, developing countries such as India need Artificial Intelligence (AI) systems that would assist the clinicians and eventually require less to no manual intervention to predict such conditions. Several studies have reported that cervical assessment on transvaginal sonography may help predict preterm delivery (4,5). In this paper, a segmentation technique is proposed to predict preterm birth using Convolutional Neural Networks (CNNs). The proposed model is trained on a dataset of 1334 transvaginal ultrasonic 2D images. According to the result, the U-net-based CNN approach achieved more promising results than the current state-of-the-art techniques. Show less

Brain tumor segmentation can be used to separate healthy tissues from cancerous tissues accurately.
Usually, in clinical settings, a brain MRI (Magnetic Resonance Imaging) scan is used to detect
tumors, but manually segmenting out a tumor is a time-consuming process and requires highly skilled
and experienced radiologists. If a malignant tumor is not detected in the early stages of cancer, it can
cost a life. Hence, early detection of tumors in a short amount of time would be a… Show more

Brain tumor segmentation can be used to separate healthy tissues from cancerous tissues accurately.
Usually, in clinical settings, a brain MRI (Magnetic Resonance Imaging) scan is used to detect
tumors, but manually segmenting out a tumor is a time-consuming process and requires highly skilled
and experienced radiologists. If a malignant tumor is not detected in the early stages of cancer, it can
cost a life. Hence, early detection of tumors in a short amount of time would be a great boon to
humanity. This paper proposes a brain tumor segmentation technique using a Deep Convolutional
Neural Network (CNN). The type of Convolutional Neural Network proposed is called U-NET. The
U-NET architecture uses semantic segmentation that labels every pixel of an image rather than just
detecting objects. So, if we can segment out an image, we would know which pixel belongs to what
part of a human’s anatomy. It can help us efficiently detect all types of tumors and irregularities,
which would help the radiologists and surgeons with detection and surgery. Show less