For people suffering from nocturnal epileptic seizures it is crucial to have a system that can detect such seizures in real-time. In this paper, we present a preclinical demonstrator for real-time detection of nocturnal seizures based on the heart rate. The system is built on the VITRUVIUS body sensor platform, which consists of a body hub (a smart phone) and sensors communicating via Bluetooth. The seizure detection application running on the body hub has an online classifier that is triggered by an adaptable cumulative sum (CUSUM) algorithm. In case of an event, an alarm message can be sent to the seizure monitoring application or to the caregiver's phone. We present the architecture of the system with emphasis on the detection algorithm and the applications. The off-line evaluation of the system on five patients achieved a sensitivity of 95% and a positive predictive value of 85%. The preliminary results are encouraging and the system will be implemented in a clinical trial at Kempenhaeghe epilepsy clinic in the Netherlands.