CARINA MEDICAL LLC — Department of Health and Human Services SBIR Phase I: NIBIB

Description

ABSTRACT As early detection and better treatment have increased cancer patient survival ratesthe importance of protecting normal organs during radiation treatment is drawing more attentionCancers in the thoracic regionwhich include lungesophagealthymusmesothelioma and breast cancersare among the most pervasive and deadly cancersThe protection of normal thoracic organs including lungsheartesophagus and spinal cord is critical in reducing long term toxicity in such cancersTo avoid excessively high radiation doses to such organs at riskOARsthey are required to be correctly segmented from simulation computed tomographyCTscans during radiation treatment planning to get an accurate dosage distributionDespite tremendous effort into the development of semior fully automatic segmentation solutionscurrent automated segmentation softwaremostly using the atlas based methodshas not yet reached the level of accuracy and robustness required for clinical usageThereforein current practicesignificant manual efforts are still required in the OAR segmentation processManual contouring suffers from interand intra observer variability as well as institutional variability where different sites adopt distinct contouring atlases and labeling criteria and thus leads to inaccuracy and variability in OAR segmentationWhen OARs are very close to the treatment targetsegmentation errors as small as a few millimeters can have a statistically significant impact on dosimetry distribution and outcomeIn additionit is also costly and time consuming as it can takehours of a clinicianstime to segment major thoracic organs due to the large number of axial slices requiredThe associated human efforts would significantly increase if adaptive radiation therapyARTis used as OARs from two or more simulation CT scans need to be segmented to adjust treatment plansIn recent yearsthe rapid development of deep learning methods has revolutionized many computer vision areas and the adoption of deep learning in medical applications has shown great successBased on a deep learning based algorithm we developed that achieved better than human performance and rankedst inAmerican Association of Physicist in Medicine Thoracic Auto segmentation Challengea thoracic OAR auto segmentation product will be developed in this project with the two aimsimprove and validate the deep learning based automatic thoracic organ segmentation algorithm on a larger clinical data setandincorporate this algorithm into a preliminary product that fits into the clinical workflowWith this productthe segmentation accuracy can be improvedleading to more robust treatment plans in protecting normal organs and improved long term patient outcomeFurthermorethe time and cost of radiation treatment planning can be greatly reducedcontributing to a more affordable cancer treatment and reduced healthcare burden NARRATIVE As early detection and better treatment have increased cancer patient survival ratesthe importance of protecting normal organs during radiation treatment is drawing more attentionTo avoid excessively high radiation doses to such organs at riskOARsthey are required to be correctly segmented from simulation computed tomographyCTscansA deep learning based thoracic OAR auto segmentation product developed in this project can improve the segmentation accuracy and reduce the time and cost of radiation treatment planning as compared with the current manual processleading to improved long term patient outcome and reduced cancer treatment cost