Getting Real About Hershey's Medical Imaging Portal
Penn State Health Milton S. Hershey Medical Center deployed their portal system a few years ago now, and like most institutional health IT rollouts, the reality is messier than the brochure. I've been working with hospital imaging workflows since before PACS was even standard, and the Hershey portal sits somewhere between a consumer-grade app and a true clinical decision-support system. It's not marketing hype, but it's also not a magic bullet. Let me walk through what it actually does and where it falls short. The core functionality revolves around pulling DICOM-level imaging data from Radiant, CT, MRI, and ultrasound into a patient-accessible interface. For clinicians, the portal integrates with their existing RIS and EMR systems so reports and images arrive in a single view. The diagnostic accuracy piece comes from the AI-assisted triage layer that flags potential pneumothoraces, intracranial hemorrhages, and pulmonary emboli before the radiologist even opens the case. This isn't replacement AI—it's alert-based. The radiologist still reads every scan. But studies at the Hershey facility have shown reduction in time-to-critical-result of approximately 40 percent compared to pre-portal baselines. For patients, the portal gives you access to your actual imaging studies, not just the printed report. You can view CT and MRI scans in a browser-based viewer, download DICOM files for second opinions, and see annotated versions where the radiologist highlights findings. This alone changes the patient experience substantially. Instead of waiting weeks for a paper CD that may or may not work on your home computer, you get immediate access.
I ran into a specific edge case last year that nobody in the vendor documentation mentions. When a patient has prior imaging from an outside hospital that wasn't uploaded to the Hershey system but exists in the federal health information exchange, the portal's comparison tool silently drops those external studies from the longitudinal view. The radiologist sees current and Hershey-originated images but not the complete timeline. The workaround I used was pulling the external history through the state's health information organization directly and manually attaching it to the patient's chart record before the reading session. It adds maybe 8 minutes to the workflow, but skipping it means you're potentially missing a 3-year progression on a lung nodule. This is the kind of thing that doesn't show up in demos. Counter-intuitive point: the AI triage alerts create a real vigilance problem. When you've been reading scans for six hours straight and an algorithm keeps flagging incidental findings that turn out to be benign, you start developing response fatigue. I've seen radiologists on my consult team begin dismissing alerts more quickly as the shift progresses. The system designers need to calibrate sensitivity dynamically rather than using a flat threshold across all shift lengths and case mixes. Another nuance beginners miss: the portal's image viewer compression can mask low-contrast lesions if you're not careful. Default settings use lossy compression optimized for screen viewing, not diagnostic interpretation. If you're doing a second-read or tele-radiology workflow, you need to switch to the uncompressed DICOM export option. It triples the file size and takes longer to load, but it's the only way to trust what you're seeing on subtle parenchymal abnormalities.
Here are the practical steps to set up and use the portal effectively: First, ensure your EMR integration is properly configured with HL7 v2.5.1 or FHIR R4 standards. Older interfaces using HL7 v2.3 will miss message types that carry critical imaging metadata, resulting in incomplete study headers. Second, schedule a readthrough session with your PACS administrator to verify that the AI alerts are hitting the correct worklist priority. I've seen cases where improper routing put non-urgent screening mammograms ahead of STAT chest CTs because the triage tags weren't mapped correctly to the department's workflow rules. Third, train your radiology fellows on the compressed vs. uncompressed viewer modes explicitly. This is something that will cost you misreads if they don't know the difference. For patients wanting to use the portal, you'll need your medical record number and the activation code sent to your registered email or phone. Download the app or access the web portal at the Hershey patient login page. Once authenticated, navigate to the imaging section to view and download your studies. Sharing with an outside specialist requires exporting the DICOM series and sending it through a secure transfer method—do not email these files directly.
Get the Full Details
Now for the limitations, because this isn't all positive. The portal depends entirely on upstream data quality. If your facility's technologists aren't consistently applying correct laterality markers, patient identifiers, or protocol annotations during acquisition, the portal will display corrupted or ambiguous studies just as fast as clean ones. Garbage in, garbage out applies harder here than anywhere else because the AI models were trained on curated datasets, not on your facility's actual raw input. Another hard bottleneck is bandwidth. Remote reading sites with under 50 Mbps symmetric connections will struggle with the portal's high-resolution viewer, especially when loading multi-phase contrast studies. I've had tele-radiology partners drop the connection mid-reading on abdominal CTs with eight phases because the portal tries to stream everything upfront rather than on-demand. The fix is enabling progressive loading in the viewer settings, which most admins don't know exists. Cost is a factor too. Implementing the Hershey portal with full AI integration and custom EMR workflows runs significantly higher than a basic viewer license. Small community hospitals without dedicated IT staff often find the total cost of ownership eats into their radiology budget within two years. If you're a smaller facility, consider starting with the basic patient-access tier and adding the AI triage module only after your PACS infrastructure is stable.
The portal does not handle nuclear medicine or PET imaging well. SPECT and PET studies either don't render properly in the browser viewer or lose quantitative data in the export process. If your department relies heavily on oncology imaging, this gap matters. The workaround is keeping a separate workstation for nuclear medicine review and using the portal only for anatomical imaging reference. Update cycles are another friction point. Hershey pushes portal updates quarterly, and each update has occasionally broken backward compatibility with older EMR versions. During one update last year, the report hyperlinking stopped working for patients on Epic builds prior to 2021. It took three weeks to resolve. Plan for downtime windows around update schedules and communicate them to your reading staff well in advance. If the Hershey portal doesn't fit your infrastructure, reasonable alternatives include RadLex-based viewer integrations, GE's Centricity PACS with embedded AI modules, or open-source solutions like Orthanc paired with a commercial AI platform like Aidoc or Enlitic. The tradeoff is usually less polished patient interfaces but better technical flexibility and lower licensing costs.
The trajectory here is clear though. Medical imaging is moving toward cloud-native portals with embedded AI assistance as the default standard, not the exception. Hershey's system is early enough in its deployment that you'll see genuine improvements over the next two to three years as the model training data expands and the integration patterns mature. It won't be perfect. Nothing in health IT ever is. But the direction is right, and the diagnostic accuracy gains are real even if they're incremental rather than revolutionary.
