Chromosomal Abnormal detection Using Invariant Structural Cascade Segmentation based on Deep vectorized Scaling Neural Network

Chromosomal anomaly causes the development of a fetal growth defect, which is identified in ultrasound images by analysing Down syndrome samples. Researchers have developed various methods for analysing chromosomal data, employing soft computing optimised with an ML approach and examining micro Down syndrome samples. Still, detecting genetic disorders in the fetus is difficult. However, it is necessary to identify differences in chromosomes that are genetically based accurately. Since feature variants project similar dimensional structures, the resulting recognition errors may prevent accurate anomaly detection. To address these issues, this research develops an Invariant Structural Cascade Segmentation (ISCS) based on a Deep Vectorised Scaling Neural Network (DVSNN) to automatically detect genetically based chromosome defects in ultrasound images for early diagnosis. Initially, the Fetus Ultrasound image dataset is collected, and an Adaptive Median Filter (AMF) is applied to obtain a noise-free fetal image. After pre-processing, the Invariant Structural Cascade Segmentation (ISCS) method is employed to segment the image. This method determines features such as texture, colour, and pixel intensity. Subsequently, the Angular Vector Projection (AVP) technique is applied to analyse cell structural variance. Afterwards, the Histogram Colour Equaliser (HCE) approach is used to select irregularity cells using K-counts and feature weights. Based on feature weight, the proposed DVSNN algorithm is used to categorize the risk of genetic disorders based on chromosomal abnormality. Our approval showed empowering results when coordinating chromosomes, with noticeable and undetectable banding patterns. Therefore, the proposed method helps address concerns in genetic-based chromosome image analysis, such as segmentation and cytogenetic classification. This effectively identifies chromosome abnormalities in micro image cells to detect patients at risk. Compared to traditional approaches, the proposed method improves classification accuracy, precision, and recall rate.