The advancements in medical imaging increased the challenges of machine learning models working towards the diagnosis and treatment of medical conditions. The challenges are: (i) the non-availability of a diverse and var-ied dataset, (ii) the class-imbalance in a dataset. This research work focuses on generating medical images to overcome the class imbalance and non-availability in some modalities using a Hybrid Generative AI (GenAI) approach for CT (Computed Tomography) imaging on the Kaggle "Brain CT Images with Intracranial Hemorrhage Masks" dataset. This dataset is an open-source dataset of kaggle and has brain CT images, brain’s bone CT im-ages, and haemorrhage masks (only if a Haemorrhage is present) separately for each patient. This evoked the need for a fusion overlay process, that can fuse the images into one image and a need to balance the images present in each class of haemorrhage. For fusion overlay process, Discrete Wavelet Transform (DWT) technique is applied on the dataset to produce a fused image for each slice of a patient’s brain CT image. The class-imbalance issue is handled using a hybrid GenAI model that comprises of two main models namely: Variational AutoEncoder (VAE) and Generative Adversarial Net-work (GAN); it generates near-to-real synthetic CT image slices of the brain to balance the dataset for each class. The quality of the image fusion and generated images are validated using metrics like Structural Similarity Index Measure (SSIM), Peak Signal-to-Noise Ratio (PSNR) and Entropy. The hybrid model gave an improved performance of 16.36% more than the standalone models in terms of SSIM and PSNR. From the results, it is observed that the hybrid model gives an improved performance using the fusion overlay technique.