{"id":24857255,"url":"https://github.com/blackhatinside/mtech_researchwork","last_synced_at":"2025-03-26T15:41:06.935Z","repository":{"id":261782749,"uuid":"844506479","full_name":"blackhatinside/Mtech_ResearchWork","owner":"blackhatinside","description":"Brain Legion Segmentation using Python","archived":false,"fork":false,"pushed_at":"2025-03-25T04:29:01.000Z","size":10690,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-03-25T05:27:09.822Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/blackhatinside.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-08-19T12:00:31.000Z","updated_at":"2025-03-01T05:33:24.000Z","dependencies_parsed_at":"2024-11-08T12:31:20.915Z","dependency_job_id":"b42510d7-9f1a-4cc9-944b-1baf77a0b9fe","html_url":"https://github.com/blackhatinside/Mtech_ResearchWork","commit_stats":null,"previous_names":["blackhatinside/mtech_researchwork"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/blackhatinside%2FMtech_ResearchWork","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/blackhatinside%2FMtech_ResearchWork/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/blackhatinside%2FMtech_ResearchWork/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/blackhatinside%2FMtech_ResearchWork/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/blackhatinside","download_url":"https://codeload.github.com/blackhatinside/Mtech_ResearchWork/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":245683301,"owners_count":20655537,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2025-01-31T17:16:42.905Z","updated_at":"2025-03-26T15:41:06.923Z","avatar_url":"https://github.com/blackhatinside.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Mtech_ResearchWork\nBrain Legion Segmentation using Python\n\n\nHere's a detailed summary of the research project I am working on in my MTech Thesis work and also information about the ISLES22 dataset:\n\nResearch Project Overview:\nThe project focuses on brain lesion segmentation using class-aware augmentation techniques combined with an Attention U-Net architecture. The key objectives were:\n- Addressing class imbalance and data scarcity in brain lesion datasets\n- Implementing optimized segmentation through Attention U-Net\n- Examining effects of various augmentation methods\n\nISLES22 Dataset Characteristics:\n- Multi-center MRI dataset designed for stroke lesion segmentation\n- 400 total cases (250 training, 150 test samples)\n- Images in NIfTI format converted to PNG (112x112 pixels)\n- Contains DWI (Diffusion-weighted imaging) scans with corresponding lesion masks\n- Ground truth masks are binary (white for lesion, black for non-lesion)\n- Lesions can vary in sizes and shapes, may be as small as a single pixel\n\nKey Methodology:\n1. Data Preprocessing:\n- Conversion from NIfTI to PNG format\n- Intensity normalization using Min-Max scaling\n- Standardized image resizing to 112x112 pixels\n\n2. Class-aware Augmentation:\nThe data was categorized into 5 classes based on lesion size:\n- C1: 1-50 pixels (2477 images)\n- C2: 51-100 pixels (637 images)\n- C3: 101-150 pixels (413 images)\n- C4: 151-200 pixels (253 images)\n- C5: \u003e200 pixels (1047 images)\n\nEach class received specific augmentation strategies:\n- Smaller lesions: More aggressive rotations and transformations\n- Larger lesions: Conservative changes to preserve clinical validity\n\n3. Results:\n- Initial dataset expanded from 4,827 to 13,174 images\n(c1 - 2477 images, c2 - 2548 images, c4 - 2530 images, c4 - 1047 images, c3 - 2478 images)\n- Improved Dice score from 0.6651 to 0.7307 with augmentation\n- Best performance achieved by U-Net with augmentation (0.7451 Dice score)\n\nThe project demonstrated that class-aware augmentation effectively addresses imbalance issues while maintaining clinical relevance, leading to improved segmentation accuracy especially for challenging lesion cases.\n\nThe ISLES22 dataset proved valuable for this research due to its:\n- Diverse lesion characteristics (size, shape, location)\n- Multi-vendor nature reflecting real clinical scenarios\n- High-quality annotations\n- Standardized evaluation metrics\n- Public availability for reproducible research\n\nThe dataset's inherent challenges (class imbalance, lesion variability) made it ideal for testing advanced augmentation strategies and segmentation architectures.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fblackhatinside%2Fmtech_researchwork","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fblackhatinside%2Fmtech_researchwork","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fblackhatinside%2Fmtech_researchwork/lists"}