Radiation Treatment Planning Software with Integrated Computer Aided Tumor Detection

Unmet Need

Radiation is an effective component of cancer therapeutics and is utilized in approximately 50% of all cancer cases. For treatment to be effective, a treatment plan must be created that takes into consideration diagnostic images and scans as well as previous treatment history. However, tumors must be large enough to be detectable by both diagnostic imaging and the physician. Additionally, it can be difficult to distinguish between previously treated tumors and newer tumors. This can lead to incorrect retreatment or omission of treatment for the patient. There is a clear need for improvements in radiation treatment planning that can improve accuracy and efficiency of tumor detection and treatment.

Technology

Duke inventors have developed an integrated software for radiation treatment planning. This is intended to be used by physicians to create more accurate and effective radiation treatment plans for patients with cancer. This software would fit in seamlessly with the existing workflow that physicians already follow. Once diagnostics images of the patient are taken, they will be integrated into a radiation treatment planning system where the tumors will be identified and marked for treatment. This new technology would integrate computer aided detection (CAD) algorithms into the existing planning software to assist physicians in identifying tumors that are difficult to detect. Additionally, this new software will have the ability to keep track of which tumors have already received treatment and which tumors are treatment agnostic. This CAD algorithm was developed by training a convolutional neural network on a dataset of brain metastases MRIs that were not detected during routine clinical care. This has been demonstrated to have a sensitivity of 94% for prospectively identified metastases and 80% for retrospectively identified metastases which escaped detection by physicians during their routine clinical workflow.

Other Applications

With continued training and validation, this technology could be adapted for use in a diagnostic setting, as it can identify tumors that are difficult to detect with the naked eye.

Advantages

• Saves physicians time by performing lesion identification and contouring first
• Has the ability to act as a double-checking mechanism and can assist physicians in finding overlooked tumors
• Reduces potential treatment errors by cataloguing previously treated tumors