Handwriting serves as a means to articulate thoughts, concepts, and language. Throughout time, the scribbled cursive script often associated with medical practitioners has gained widespread acceptance despite its illegibility. The datasets utilized in this study comprise samples of cursive handwriting obtained from doctors and medical students across various clinics and hospitals. The paper introduces a Handwriting Recognition System utilizing a Deep Convolutional Recurrent Neural Network (CRNN) designed to decipher text from images of doctor-written prescriptions, presenting the converted readable text from the original handwriting. The CRNN model is trained using pre-processed input images for handwritten text recognition. The performance evaluation of CRNN, both with and without the Connectionist Temporal Classification (CTC) loss function, is detailed in this study. The findings indicate that the utilization of the CTC loss function resulted in a 92% accuracy rate during training, while without CTC, the model demonstrated an 87% accuracy rate. Validation revealed a rate of 52.5% with CTC and 45% without CTC. A total of 346 images were collected from doctors and medical students.