Alpha particles are highly lethal to cancerous cells, creating complex double-strand DNA breaks. Only a few hits to the cell nucleus are required to kill the cancer cell. However, as a result of their short-range in tissue, Alpha particles had been unsuitable for the treatment of cancer.

Alpha DaRT overcomes this range limitation and enables Alpha radiation for the treatment of solid tumors. The treatment is delivered by intratumoral insertion of the Alpha DaRT seed that contains Radium-224 atoms embedded below its surface. In the process of decay, Radium-224 releases its short-lived Alpha-emitting atoms into the tumor. By diffusion and convection, these atoms disperse to a therapeutically significant range of several millimeters, delivering a high dose of radiation inside the tumor.

The detection of seeds in CT images is a critical step in treatment planning and is currently performed manually. However, this process is prone to inter-and intra-observer variability. Automation would increase accuracy and efficiency while reducing cost.

AlphaTau has created a synthetic CT dataset for McMedHacks. Their unique challenge includes the detection of seeds from CT images. More information is coming soon.

Detection and segmentation are staples of deep learning-based medical image analysis. During the McMedHacks series, we have shown many of you some of the most popular deep learning architectures such as U-Net, ResNet, MaskRCNN, and more applied to various problems across imaging modalities. Now it is your turn to use a deep learning-based detection and/or segmentation algorithm.

Interestingly we can use the same principles used to distinguish dogs from cats to classify tissues into healthy, benign, and malignant, diagnose Alzheimer's disease, and much more. This prize is awarded to the submission that has the best implementation of a deep learning-based classification of medical images.

In almost every medical image analysis pipeline, visualizing images, labels and other metadata is crucial for clinical integration. An intuitive visualization is a great tool for physicians to understand and interpret medical images. We challenge you to create a clever and or beautiful way of visualizing medical images and or associated information.

GANs are a powerful tool to generate synthetic images. In medical image analysis, GANs can be used to create new training data to overcome the traditional medical image data limitations, to generate contours of organs, and much more. This prize goes to the best implementation of GANs for medical image analysis.

Automatically considered: Deep learning-based medical image analysis faces many significant challenges ranging from inter-institutional data variability to the issue of patient-data sharing restrictions. This prize is awarded to the most innovative solution to a medical image analysis-specific challenge.

Automatically considered: Many of our participants are beginners in the field of deep learning and medical image analysis. This award goes to participants that showed the most growth and improvement over the course of McMedHacks.

Automatically considered: This award goes to the submission that was most upvoted by the McMedHacks community.

Automatically considered: We know that 48 hours is not enough time to create a fully functioning deep learning application. This award goes to the best unfinished project submission.