Fig. 1. Network structure

Fig. 2. Backbone feature extraction network

Fig. 3. Spatial attention mechanism

Fig. 4. Multi level feature fusion module

Fig. 5. Down sampling module

Fig. 6. Multi kernel convolution structure

Fig. 7. MRI images of brain tumors. (a) No tumor; (b) glioma; (c) meningioma; (d) pituitary

Fig. 8. Accuracy and loss variation curves. (a) Binary classification accuracy curves; (b) change curves of binary loss; (c) four classification accuracy curves; (d) four category loss change curves

Fig. 9. Diagnosis results of binary classification tasks. (a) No tumor; (b) tumor

Fig. 10. Four classification task diagnosis results. (a) No tumor; (b) glioma; (c) meningioma; (d) pituitary

Fig. 11. Thermograms of various types of tumors. (a) Meningioma; (b) glioma; (c) pituitary