AI HER2 Slide Diagnosis Software
Budget: ₹37,500 – ₹75,000 INR
I am Dr Krishn Kumar Verma, currently pursuing an MD in Pathology at Gajra Raja Medical College, Gwalior. For my thesis—“Role of AI in HER2-neu Positive Breast Carcinoma”—I need a Windows desktop application with a clean, intuitive GUI that can:
• i will provide Ai model with multiple slides images and AI model will make a digital whole slide image by Grid model or mosaic technique
• ingest whole-slide images (WSI) of breast tissue that I supply
• train an AI model exclusively on the HER2-neu positive cases provided
• run inference on new slides, displaying the system’s diagnosis and confidence score side-by-side with the human diagnosis for easy comparison
• allow me to export these paired results for statistical analysis and inclusion in my dissertation
The workflow I envision is simple: my two supervisors and I will first review each slide manually; I will then feed the same slide to the application and compare outcomes, noting whether the AI result alters our final opinion. Clear visual cues—such as heat-maps or bounding boxes—will be valuable but are not mandatory if they complicate development.
Core requirements
• Platform: Windows 10/11 standalone installer (no web deployment)
• Interface: Graphical only; no command line interaction for routine use
• Compatibility with common WSI formats (SVS, NDPI or TIFF—happy to agree on one during kickoff)
• A straightforward way for me to add more training data later without rewriting code
Deliverables I need at hand by project close:
1. The compiled Windows application (.exe or installer) ready to run offline
2. Full source code with brief inline comments
3. A short user manual covering installation, training workflow, inference, and result export
4. A concise technical note describing the model architecture, dataset split, and performance metrics—material I can cite in my thesis
I am flexible on the underlying tech stack—Python with PyTorch or TensorFlow is common in digital pathology, but I am open to alternatives if you can justify them. What matters is reliability, reproducibility, and clear documentation.
If you have experience in digital pathology, computer vision, or medical image analysis, I would love to see examples of similar work and discuss timelines. Feel free to message me with any clarifying questions; I can share sample slides as soon as an NDA is in place so we can confirm format compatibility before development begins.
• i will provide Ai model with multiple slides images and AI model will make a digital whole slide image by Grid model or mosaic technique
• ingest whole-slide images (WSI) of breast tissue that I supply
• train an AI model exclusively on the HER2-neu positive cases provided
• run inference on new slides, displaying the system’s diagnosis and confidence score side-by-side with the human diagnosis for easy comparison
• allow me to export these paired results for statistical analysis and inclusion in my dissertation
The workflow I envision is simple: my two supervisors and I will first review each slide manually; I will then feed the same slide to the application and compare outcomes, noting whether the AI result alters our final opinion. Clear visual cues—such as heat-maps or bounding boxes—will be valuable but are not mandatory if they complicate development.
Core requirements
• Platform: Windows 10/11 standalone installer (no web deployment)
• Interface: Graphical only; no command line interaction for routine use
• Compatibility with common WSI formats (SVS, NDPI or TIFF—happy to agree on one during kickoff)
• A straightforward way for me to add more training data later without rewriting code
Deliverables I need at hand by project close:
1. The compiled Windows application (.exe or installer) ready to run offline
2. Full source code with brief inline comments
3. A short user manual covering installation, training workflow, inference, and result export
4. A concise technical note describing the model architecture, dataset split, and performance metrics—material I can cite in my thesis
I am flexible on the underlying tech stack—Python with PyTorch or TensorFlow is common in digital pathology, but I am open to alternatives if you can justify them. What matters is reliability, reproducibility, and clear documentation.
If you have experience in digital pathology, computer vision, or medical image analysis, I would love to see examples of similar work and discuss timelines. Feel free to message me with any clarifying questions; I can share sample slides as soon as an NDA is in place so we can confirm format compatibility before development begins.