Leveraging the power of AI in cancer care

Compared to humans, AI doesn’t tire. It always delivers the same standardized quality, even after a long day at work. That is why certain tasks which take a long time in the clinical routine can be automized and standardized to relieve physicians from tedious and repetitive tasks while at the same time providing patients with precise diagnosis and treatment. This is crucial in cancer care, since the earlier and the more accurately cancer is diagnosed, the higher the chances of cure. Another benefit of earlier diagnosis and treatment: if more patients are being treated earlier, the financial impact of cancer can be greatly reduced.[2] For one, the cost for early cancer treatment is much less. Secondly, patients will be able to return to work earlier, require less post-treatments and benefit from better quality of life in general. 

In the field of early detection and diagnosis, Siemens Healthineers has developed an array of “automated helpers” to speed up workflows – with some of them specifically targeting cancer. Based on deep-learning algorithms, this family of “AI-Rad Companion” applications¹ is able to support a variety of functionalities such as highlighting abnormalities, segmenting anatomies and comparing results to reference values. 

With AI automating certain steps, more time can be dedicated to individual patients who then might benefit from an early diagnosis and, if necessary, treatment. Studies have shown workflow improvements by reducing the mean interpretation time for radiologists by around 22 percent when AI support is provided for the reading of Chest CT images.[3]

Lung cancer is the most common cancer in men worldwide and also the deadliest cancer type.[4] When it comes to the actual diagnosis and the reading of medical images, AI can help the radiologist by automatically detecting and marking lung nodules on CT images. After the segmentation of the nodules, the AI-Rad Companion Chest CT automatically calculates their volume and the maximum two- and three-dimensional diameter. This supports radiologists in focusing on these conspicuous areas. If these prove to be malignant tumors, the treatment might begin earlier. “AI has the possibility to shorten the turnaround time and increase the cost effectiveness of lung cancer screening”, explains Prof. Philippe Grenier from Hôpital Foch, France. 

In case there is a finding, and the patient starts therapy, this software can also be used in the follow-up sessions. Tumor monitoring and comparison over time is very labor-intensive and therefore costly. With follow-up, patients can be rescanned according to specific grids to produce comparable images. Another advantage of the AI-based Chest CT algorithm: used in dedicated programs for certain risk groups, e.g., smokers, it can also deliver incidental findings for other areas, that are included in the scan field, such as enlarged diameters of the aorta, because the AI algorithm automatically analyzes the complete CT image of the chest and does not only focus on the lung like a radiologist would normally do. 

In the field of prostate cancer where Magnetic Resonance Imaging (MRI) is typically the imaging modality of choice, an AI-Rad Companion Prostate MR has been specifically developed to provide radiologists with biopsy support. The software performs an automated segmentation as well as an automated volume estimation of the prostate. When the prostate specific antigen (PSA) value is known, the AI can calculate the PSA density based on that. The radiologist can manually mark and characterize lesions and other targets and add comments. Segmentations, targets found as well as burnt-in contours can all be exported for the urologist to be fused with ultrasound images as biopsy guidance.

The AI-Rad Companion family of AI-powered solutions¹ also help in the fight against cancer – from radiology support for lung and prostate to automatic contouring of the organs in therapy planning.

To increase accuracy and speed up the early detection of breast cancer, Siemens Healthineers has integrated a special AI-powered algorithm into its mammography reading software.² And it is helping on two fronts. Firstly, trained on more than one million multi-vendor data from all over the world, this AI tool helps radiologists prioritize cases with a higher chance of malignancy by triaging them on an accent score from 1 to 10. This means that more suspicious cases can be prioritized, and patients can get their diagnosis and subsequent potential treatment earlier. 
Secondly, the software offers local region analysis for dedicated areas and works in an interactive way to help radiologists with more accurate reading.[5] If radiologists see an anomaly on a mammogram or tomosynthesis, they can click on the suspicious region and then get an estimation on a score from 1 to 95 how high the chance is that a malignancy is present. “Because screening is becoming more precise, fewer women have to come back to the hospital for further analysis while the same sensitivity is maintained.” says radiologist Ritse Mann, MD, of Radboud University Medical Center, Nijmegen, the Netherlands.

If a patient has to undergo radiation therapy, a thorough treatment plan is key, since the goal is to destroy cancerous cells while preserving healthy tissue and organ function to get the best outcome for the patient. Creating such a plan is complex. RapidPlan™ knowledge-based planning from Varian, a Siemens Healthineers company, is a machine-learning tool that studies best practices from past successful treatment plans and creates knowledge-based treatment models that are applied to improve the treatment plans for future patients. These RapidPlan models help to quickly generate and validate new high-quality treatment plans based on shared expertise. 
Furthermore, prior to starting the actual radiation therapy, the treating radiation oncologists need to contour the organs-at-risk (OAR) based on CT images to ensure a more precise dose distribution and optimization of radiotherapy treatments thereby sparing healthy OAR located near a tumor, from unnecessary radiation. To have this done automatically and without cumbersome manual work, Siemens Healthineers offers various AI-based autocontouring solutions³, like the AI-Rad Companion Organs RT. “The standardized algorithms deliver the same precision as an experienced radiation therapy expert – always returning consistent results. Manual contouring can vary from user to user, which affects treatment”, explains Manuel Algara López, MD, from Hospital del Mar in Barcelona, Spain.

But the role of AI doesn’t stop with treatment planning – it can also support individualized treatment. Over the course of radiation treatment for cancer, which can last anywhere from one to seven weeks, there are anatomical changes in the tumor as well as in the surrounding healthy tissues. “Adaptive therapy” is a way to account for these changes in a matter of minutes: The Ethos Therapy™ system by Varian, a Siemens Healthineers company, is capable of generating a new radiotherapy treatment plan for every individual patient every day based on current anatomical images taken just prior to treatment, instead of basing an entire course of treatment on one CT scan that is generated days or even weeks before treatment begins. Ethos thereby enables the clinical team to offer a more targeted treatment that minimizes the impact on other tissues and organs. 

“Being able to adapt treatments based on what we see every day is already showing huge benefits,” said Trent Aland, Group Director of Medical Physics for Icon Group, Australia's largest dedicated cancer care provider. “Since our clinical teams have a mix of skills, including treating prostate, head and neck, and thorax, we see an opportunity to develop and offer adaptive workflows for a large range of patients. We’re also rethinking whether we need to impose burdensome preparations on our patients, like a full bladder and an empty bowel.” 

All these examples show how artificial intelligence can support healthcare professionals in their daily fight against cancer to provide their patients with earlier and more precise diagnoses and faster and more accurate therapy planning and therapy.