Text this: An Efficient Explainability of Deep Models on Medical Images