Canned mackerel is popular and has become a vital commodity in the modern consumer market. It is important to preserve product value, maintaining the intact appearance of the mackerel. Although a sampling inspection can be performed by opening cans during shipment, this approach renders the product unsellable. There are required methods that enable judgments regarding quality without the opening of cans. Bones could be visualized, but the canned fish meat surrounding the bones remains challenging to observe through X-ray computed tomography (CT). This study confirmed that machine learning could effectively discern these minute discrepancies. This study revealed that a reasonably accurate discrimination could be achieved with approximately 35 pieces of training data. These results will be applications in identifying cans with issues like peeling skin or flesh, effectively discerning substantial differences.