This research introduces a compact Ultra-Wideband (UWB) antipodal Vivaldi antenna with a U-slot tailored for biomedical applications. Utilizing an elliptical tapered patch variation on a 50*32 mm2 substrate achieves its compact design. Parameter optimization, employing the Multi-Layer Stacked Shallow Attention Neural Network (MLSSANN) with Adaptive Gannet Optimization Algorithm (AGOA), ensures accurate predictions and efficient exploration of the design space. Performance analysis includes metrics like gain, directivity, reflection coefficients, return loss and radiation efficiency. The U-slot variant exhibits higher gain peaking at 24.5 GHz (10 dB) compared to the standard variant peaking near 37 GHz (13.7 dB). The U-slot antenna also shows improved directivity and return loss, with a 13.89 % enhancement in return loss at 3.7 GHz. Moreover, the addition of the slot shifts the lower cut-off frequency from 1,580 MHz to 740 MHz, reducing the antenna size by 41 % while maintaining acceptable radiation characteristics.