Karyotyping of human chromosomes has become an important clinical procedure in the screening and diagnosis of cancers and genetic disorders. Karyotyping is a standard technique utilized to classify metaphase chromosomes into 24 types and represents the essential part of the chromosome trisomy abnormality recognition. A trisomy is an abnormality type where there are three copies of a particular chromosome, instead of two in normal cases. To improve the speed and accuracy of the classification process, we propose a four-step automated system that can detect any of the six most common trisomies. The first step is the chromosome’s image enhancement to prepare the image before performing the detection of metaphase spreads and classifying the banded chromosome. The second step is chromosome segmentation to differentiate the chromosome from 'non- chromosome' or background. The third is feature extraction. The chromosome’s features: chromosome length, relative length, chromosome area, relative area, centromere index, and density profile are extracted to classify the chromosomes using an assembled local images database consisting of 157 images involving 7266 chromosomes. The fourth step is chromosome classification, where a multi-layer neural network classifier with autoencoder classifies the chromosomes into targeted chromosomes classes. The testing recognition correctness rate was 95%.