Evaluation of a delineation software for cardiac atlas-based autosegmentation: An example of the use of artificial intelligence in modern radiotherapy

PURPOSE

The primary objective of this work was to implement and evaluate a cardiac atlas-based autosegmentation technique based on the "Workflow Box" software (Mirada Medical, Oxford UK), in order to delineate cardiac substructures according to European Society of Therapeutic Radiation Oncology (ESTRO) guidelines; review and comparison with other cardiac atlas-based autosegmentation algorithms published to date.

MATERIALS AND METHODS

Of an atlas of data set from 20 breast cancer patients' CT scans with recontoured cardiac substructures creation according to the ESTRO guidelines. Performance evaluation on a validation data set consisting of 20 others CT scans acquired in the same treatment position: cardiac substructure were automatically contoured by the Mirada system, using the implemented cardiac atlas, and simultaneously manually contoured by a radiation oncologist. The Dice similarity coefficient was used to evaluate the concordance level between the manual and the automatic segmentations.

RESULTS

Dice similarity coefficient value was 0.95 for the whole heart and 0.80 for the four cardiac chambers. Average Dice similarity coefficient value for the left ventricle walls was 0.50, ranging between 0.34 for the apical wall and 0.70 for the lateral wall. Compared to manual contours, autosegmented substructure volumes were significantly smaller, with the exception of the left ventricle. Coronary artery segmentation was unsuccessful. Performances were overall similar to other published cardiac atlas-based autosegmentation algorithms.

CONCLUSION

The evaluated cardiac atlas-based autosegmentation technique, using the Mirada software, demonstrated acceptable performance for cardiac cavities delineation. However, algorithm improvement is still needed in order to develop efficient and trusted cardiac autosegmentation working tools for daily practice.

Evaluation of a software for automatic delineation of the mammary gland and organs at risk in patients treated for breast cancer in lateral position

PURPOSE OF THE STUDY

The purpose of this study was to assess the potential for improvement of routine breast and organs at risk contouring in patients treated in the lateral decubitus position using Workflow Box™ (Mirada Medical™, UK) automatic contouring software.

MATERIALS AND METHODS

Automatic contouring of the breast by this software is currently based on an atlas created from isodoses representing 95% of the prescribed dose in a population of patients previously treated at institut Curie. Forty cases of breast cancer (20 right breasts and 20 left breasts) were contoured by three radiation oncologists specialized in breast cancer, allowing the creation of a new atlas in the automatic contouring software. This study assessed the quality of contouring in 20 patients (ten right breasts and ten left breasts) by comparing manual contouring performed by the expert radiation oncologists (reference) with those generated by the old and new atlases developed at Institut Curie. The accuracy of contouring was assessed by overlap volume and the associated standard deviations.

RESULTS

Breast contouring based on the new atlas and by radiation oncologists presented a mean overlap of 0.80±0.09 for the right breast and 0.81±0.06 for the left breast. By comparing volumes of interest contoured by radiation oncologists and those obtained from the old atlas, mean overlap volumes were 0.75±0.08 and 0.74±0.13 for the right and left breasts, respectively. Twenty cases (ten right breasts and ten left breasts) used to create the new atlas were also reprocessed by this same atlas in order to assess the quality of automatic contouring of the breast when the case was already known to the software. The mean overlap volume in this case was 0.84±0.08 for the right breast and 0.83±0.06 for the left breast. Finally, after automatic contouring of organs at risk by means of the new atlas, the mean overlap volume was 0.87±0.04 for the heart and 0.93 for each lung (±0.05 for the right lung and±0.04 for the left lung).

CONCLUSION

Workflow Box™ automatic contouring software, based on our new atlas provides reliable and clinically relevant organs at risk and breast contouring. The contours proposed by the software from the new atlas were better than those obtained with the previous atlas based on 95% isodoses obtained from old treatment plans. This software has therefore become more efficient, justifying its use in routine clinical practice for breast cancer contouring in patients treated in the lateral decubitus position. Investigations are currently underway to develop a fully automated process to ensure reliable, robust and operator-independent contouring and breast cancer treatment dosimetry in the lateral decubitus position. Promising preliminary results have already been obtained.