In the machine learning, none of the algorithms are best for a particular type of problem. You have to choose the algorithms based on the complexity of your problem and dataset you use. If it is an image dataset you can use the deep learning algorithms if it is tabular dataset you can use the machine learning algorithms.
You have to just follow the basic machine learning pipeline to calculate the result of algorithm on your dataset if it is not working use the another one. Do follow the following things in each step:
1. EDA
2. Feature Engineering
3. Feature Selection
4. Use the algorithms to calculate the results.
5. Do the hyperparameter tuning using the GridSerachCV, RandomSearchCV, Genetic Algorithms etc.
6. Again calculate the accuracy
If the accuracy satisfy your requirements then deploy this model otherwise choose the next algorithm and repeat the above steps.
In my opinion, the crucial issue is to have an accurate dataset in the first place, from which the model will learn from. As soon as you have it, now you can search for how you want to implement your models.
There are several techniques that perform image classification and segmentation. Therefore it depends on what are you looking to predict (e.g. are you looking to only predict the existence of cancer?
Or, are you looking to predict the exact location of cancer? if So, are you going to predict it as an image or as a volume?
Thank you Abdalraheem Ijjeh Gyanendra Chaubey Vincent Adegoke for the response it was really helpful Abdalraheem Ijjeh am looking to only predict the existence with a high level accuracy my dataset is from https://wiki.cancerimagingarchive.net/display/Public/CBIS-DDSM
As the dataset you indicated contains mammographies (images), Deep Learning (DL) is the best choice. Google already proposed an intelligent system based on DL, which seems to have the best performance on mammographies. You can search.