Medical image detection system, training method and medical analyzation method

A medical image detection system, comprising:

• a memory, configured for storing a neural network model;
• and
• a processor, coupled with the memory, the processor being configured to execute the neural network model, the neural network model comprising:
• a feature extractor for extracting a plurality of intermediate tensors from a medical image, each of the intermediate tensors having a resolution respectively, wherein the feature extractor comprises a plurality of convolutional layers for generating the intermediate tensors with different resolutions, last three layers of the convolutional layers of the feature extractor are deformable convolution layers, sampling points of the deformable convolution layers are not distributed evenly in sample windows of the deformable convolution layers;
• a feature pyramid network (FPN), associated with the feature extractor, for generating a plurality of multi-resolution feature maps according to the intermediate tensors;
• a first output head for generating a global prediction according to the multi-resolution feature maps, the global prediction indicating whether any target object exists in the medical image;
• and
• a second output head for generating a plurality of local predictions according to the multi-resolution feature maps, each of the local predictions indicating a target object predicted location in the medical image,
• wherein the processor is configured to generate output information based on the medical image, the global prediction and the local predictions.