The World Health Organization (WHO) reports ColoRectal Cancer (CRC) as the third most common cancer in Algeria and also the third most common cause of cancer-related deaths worldwide. ColoRectal Adenocarcinoma (CRA) is a common form of CRC where histopathology analysis remains the optimal method for distinguishing between different types of colon cancer, by examining slides of the patient tissue specimens under a microscope. However, this technique is very demanding and time consuming particularly for low-grade cancers, and may lead to the risk of misdiagnosis. Consequently, computers systems are becoming increasingly valuable in alleviating the diagnostic burden for medical professionals. But, medical images are complex in nature, which make their analysis and segmentation a very challenging task. In this work, we tackle CRA image segmentation by combining an object detection deep learning architecture with a traditional segmentation algorithm. Indeed, we adapt the Mask R-CNN architecture to predict instance masks for each gland from histological images. Next, we apply the watershed algorithm to handle the overlapping predicted gland masks. This approach yields a satisfactory result on the GlaS dataset with an F1 score of 0.87, a Dice coefficient of 0.86, and an IoU of 0.78. Despite the promising results, we faced, when conducting this study, two main challenge: the heavy computational requirements of the Mask R-CNN model and the limited availability of histological CRC images.