{"id":1815,"date":"2026-03-09T08:27:37","date_gmt":"2026-03-09T08:27:37","guid":{"rendered":"https:\/\/hria.utcluj.ro\/?p=1815"},"modified":"2026-06-15T12:56:27","modified_gmt":"2026-06-15T12:56:27","slug":"deep-learning-in-imagistica-medicala-segmentare-multimodala-si-diagnostic-asistat","status":"publish","type":"post","link":"https:\/\/hria.utcluj.ro\/en\/deep-learning-in-imagistica-medicala-segmentare-multimodala-si-diagnostic-asistat\/","title":{"rendered":"Deep learning in medical imaging: multimodal segmentation and assisted diagnosis"},"content":{"rendered":"

Recent HRIA research, conducted in partnership with medical universities, aims to improve diagnostic accuracy through neural architectures optimized for various types of scans (CT, MRI, X-ray, OCT).<\/p>\n\n\n\n

Vascular Network Segmentation and Radiographic Analysis<\/strong><\/h3>\n\n\n\n

Structural improvements are introduced to standard models to increase sensitivity in detecting fine anatomical structures:<\/p>\n\n\n\n

Angio-OCT (Retinal Networks):<\/strong> The I-MedSAM<\/strong> model was used, augmented with a Frangi module<\/strong>. This modification led to a +1%<\/strong> improvement in Dice, Precision, and Sensitivity metrics for vascular network segmentation.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

Chest Radiographs:<\/strong> An improved YOLO_V12 architecture<\/strong> was developed for detecting thoracic pathologies (cardiomegaly, aortic dilation, pulmonary nodules, and pleural effusion).<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

Scoliosis Assessment:<\/strong> Use of YOLO OBB<\/strong> (Oriented Bounding Boxes) combined with a modified I-MedSAM model for automatic Cobb angle calculation.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

Tumor Recognition through Domain Adaptation Learning (DAL)<\/strong><\/h3>\n\n\n\n

A major contribution consists in the use of domain adaptation learning techniques for liver and pancreatic tumor recognition:<\/p>\n\n\n\n

    \n
  • Dataset:<\/strong> Contrast-enhanced CT images from patients with liver tumors (45 patients) and pancreatic tumors (30 patients).<\/li>\n\n\n\n
  • Methodology:<\/strong> A mixed dataset was used, followed by fine-tuning through DAL<\/strong>.<\/li>\n\n\n\n
  • Performance:<\/strong> Models such as ResNet50, ResNet101, InceptionV3, and EfficientNetB0<\/strong> were evaluated comparatively. The results indicate that the DAL approach (Acc_DAL) consistently outperforms the accuracy of models trained in a standard way or on mixed datasets.<\/li>\n<\/ul>\n\n\n\n
    \n
    \n
    \"\"
    Liver tumors \u2013 maximum accuracy of 89.2%<\/figcaption><\/figure>\n<\/div><\/div>\n\n\n\n
    \n
    \"\"
    Pancreatic tumors \u2013 maximum accuracy of 98.9%<\/figcaption><\/figure>\n<\/div><\/div>\n<\/div>\n\n\n\n

    Research Directions and Assisted Diagnosis (UMF Partnership)<\/strong><\/h3>\n\n\n\n

    In collaboration with UMF Cluj, AI-assisted diagnostic systems are being developed for complex liver pathologies:<\/p>\n\n\n\n

      \n
    • Focal Lesions:<\/strong> Classification of hepatic lesions using abbreviated MRI protocols.<\/li>\n\n\n\n
    • Hepatic Fibrosis:<\/strong> Classification of fibrosis stage through multimodal analysis (CT\/MRI\/Ultrasound).<\/li>\n\n\n\n
    • Treatment Monitoring:<\/strong> Evaluation of treatment response for hepatic lesions using CT scans.<\/li>\n\n\n\n
    • Portal Hypertension:<\/strong> Automated diagnosis of portal hypertension using CT data.<\/li>\n<\/ul>","protected":false},"excerpt":{"rendered":"

      Recent HRIA research, conducted in partnership with medical universities, aims to improve diagnostic accuracy through neural architectures optimized for various types of scans (CT, MRI, X-ray, OCT). Vascular Network Segmentation and Radiographic Analysis. Structural improvements are introduced to standard models to increase sensitivity in detecting fine anatomical structures: Angio-OCT (Retinal Networks): The\u2026 <\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"neve_meta_sidebar":"","neve_meta_container":"default","neve_meta_enable_content_width":"on","neve_meta_content_width":100,"neve_meta_title_alignment":"","neve_meta_author_avatar":"off","neve_post_elements_order":"[\"content\"]","neve_meta_disable_header":"off","neve_meta_disable_footer":"","neve_meta_disable_title":"","_themeisle_gutenberg_block_has_review":false,"footnotes":""},"categories":[7],"tags":[],"class_list":["post-1815","post","type-post","status-publish","format-standard","hentry","category-rezultate"],"_links":{"self":[{"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/posts\/1815","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/comments?post=1815"}],"version-history":[{"count":1,"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/posts\/1815\/revisions"}],"predecessor-version":[{"id":1821,"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/posts\/1815\/revisions\/1821"}],"wp:attachment":[{"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/media?parent=1815"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/categories?post=1815"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hria.utcluj.ro\/en\/wp-json\/wp\/v2\/tags?post=1815"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}