Compared　with　lung　cancer,　liver　cancer　and　other　common　cancers,　hypopharyngeal　cancer　is　a　rare　disease.　Because　the　magnetic　resonance　imaging　(MRI)　of　hypopharyngeal　cancer　is　often　uneven,　fuzzy　and　noisy,　how　to　obtain　useful　information　from　these　MRI　images　is　a　difficult　problem.　It　is　a　major　challenge　to　use　deep　learning　to　detect　the　lesions　of　hypopharyngeal　cancer　through　MRI　images.　First,　the　characteristics　and　causes　of　MRI　images　of　hypopharyngeal　cancer　were　summarized,　the　characteristics　and　application　fields　of　common　target　detection　networks　such　as　Faster-RCNN,　RetinaNet,　FCOS,　and　Cascade-RCNN　were　then　summarized,　and　the　challenges　faced　by　the　application　of　target　detection　networks　in　the　localization　of　hypopharyngeal　cancer　lesions　were　analyzed.　The　effective　solutions:　deformable　convolution　and　application　of　customized　anchors　were　introduced.　Then,　the　common　semantic　segmentation　networks　were　introduced,　and　the　challenges　of　applying　these　semantic　segmentation　networks　to　the　segmentation　of　hypopharyngeal　cancer　lesions　were　analyzed.　Finally,　the　target　detection　network　and　semantic　segmentation　network　mentioned　above　were　summarized,　and　the　future　work　of　target　detection　and　semantic　segmentation　of　hypopharyngeal　cancer　medical　images　was　prospected.　©　2024　Beijing　University　of　Technology.　All　rights　reserved.