AI-DRIVEN CLINICAL DECISION SUPPORT

for Cancer Treatment

Clinical decision support (CDS) software provides clinicians, staff, and patients with the knowledge and patient-specific information empowered by Artificial Intelligence. The information is presented at appropriate times to support subsequent diagnostic and treatment clinical decisions for patients.

DESIGN PROCESS

The design process is iterative and best thought of as a cycle of exploring opportunity areas, creating potential solutions, and testing those creations to quickly gather feedback to move ideas and prototypes to solutions and plans that can be implemented. Cross-functional collaboration is at the heart of the design process.

USER RESEARCH

We focus on clinical decision support for prostate cancer treatment in this project. The collaboration sites with Johns Hopkins hospital and Radboud University Nijmegen were set up. Bi-weekly user interviews were conducted with the users: Radiologist, Urologist, NM Physician, etc. to understand their current workflow, pain points, and getting feedback on the concepts we developed.   
  

UNDERSTAND

Why are we focusing on prostate cancer treatment? The new cases of prostate cancer are increasing every year. It was estimated that 26,730 deaths from this disease would occur in 2017. The annual prostate cancer cost is very high.  

WORKFLOW ANALYSIS

WHAT IS THE PROBLEM? 

“There is a massive under grading with systematic Biopsies, therefore a lot of surgeries are done because of this. We achieve 95% correlation between targeted Biopsies and surgery results.”

- Geert Litjens,  Radboud University, Diagnostic Image Analysis Group

COMPETITOR ANALYSIS

We researched the current market competitors on how they handle the clinical decision-making support software and also research on the AI supported decision-making process. The leanings are summarized below.

IDEATION

We took the approach of diverging the concepts, analyzing the pros and cons of each concept, then converging to one concept. By this method, we ideated each perspective of the application and had the opportunity to present alternatives to the user. 

SKETCH

We ideated the initial concept by sketching out the workflow and the important information along the process. This is a very fast way to document and visualize our thoughts after getting feedback from the user.

CONCEPTS

We designed 3 different concepts to show different information architecture. The different concept has its focus on use cases. By analyzing and testing the different concepts, we converged on one concept. 

WIREFRAMES

We went through many iterations when designing the software. The current wireframes include Dashboard, Diagnosis, Therapy pages. From the main pages, the user will be able to drill down to the detail pages in the next level.

HIGH FIDELITY PROTOTYPE

By working with the Engineers and Front-end Developers in the Princeton office, we developed the working prototype which runs the real patient data for demonstration and testing purposes. We also developed Axure prototype to showcase the interaction and workflows of the software.

TESTING

We have been testing and getting feedback from the user every week by showing the prototypes. The feedbacks show the benefit of the data integration and easy-to-use interface for the clinical decision support process. 

USE CASE

The use case below shows how an end-user would accomplish a specific task using the interface. Different types of users are involved in the use case: Urologist, Radiologist, Nuclear Medicine Physician, Patient, and other specialists. The story shows how the CDS software will help the users to make a decision for Prostate Cancer diagnosis and treatment. 

UROLOGIST

View the Dashboard

A 60-year-old patient with the process of active surveillance: Gleason score of the last Biopsy = 7; PSA= 15 ng/ml. The Patient gets a referral to the Hospital for staging. The patient checks in, the Urologist opens the CDS and finds a prepared summary of the Patient history, including comorbidities, Medication, Family history, Lab values, Pathology results.

Plan the Exams

The urologist discusses the next steps with the Patient. The patient will get an MRI of the Pelvis and a Bone scan to perform a screening of possible Bone metastasis.

RADIOLOGIST

Perform the Exam

The radiologist opens up the CDS and works with the relevant data (Lab values, Biopsy results, Patient history, Comorbidities, genetic disposition, etc.) to perform a precise diagnosis of the Pelvis MRI. The report of the MRI exam is sent to CDS. 

Validate Findings

The radiologist goes to the Diagnosis page. The recent exam report is open and the findings are extracted from the report and organized in the CDS tool. The radiologist can review and validate the findings. 

Findings are Visualized

By clicking validate button, the report is closed. The exam findings including previous imaging exams and biopsy exams are overlaid and visualized in the cartoon automatically.

NUCLEAR MEDICINE PHYSICIAN

Perform the Exam

The NM physician opens the CDS to work with the relevant data (Anamnesis if bone pain is described by the Patient, Patient history, genetic disposition, Biopsy results, etc.) The NM physician diagnoses that no bone metastasis is detected. The Bone Scan report is sent to the CDS tool. Similar to the Radiologist, the NM physician validates the finding and the finding is visualized on the global staging. 

UROLOGIST

Review the Dashboard

The Patient goes to the Urologist to review the results of the examinations. At this stage, the cancer stage and risk are calculated by the system. 

Review the Diagnosis

The urologist goes to the staging page and checks the result of the Pelvis MRI and Bone Scan. The tumor is intracapsular, with no enlarged Lymph nodes in the pelvic area; the result is a PI-RADS score of 5. The Urologist edits the diagnosis when needed. 

Recommend the Therapy

By clicking the therapy page, the Urologist gets therapy recommendations. The patient value and side effects of each therapy are visualized to help the patient and Urologist make the decision. The urologist recommends a radical prostatectomy to approach cancer curative. The urologist can communicate with a senior Urologist about the therapy recommendation. 

Detail Pages

The user can drill down to the detail pages from the dashboard to review more information such as a previous exam, risk assessment, previous exam data, family history, guidelines, etc.