Cervical cancer remains one of the deadliest cancers among women worldwide, and MRI is the main tool doctors use to judge how far the disease has spread, particularly whether the tumour has invaded the tissue around the cervix. This distinction decides the treatment: surgery if not, chemotherapy and radiotherapy if so. In practice, however, MRI scans are often shortened for patient comfort, leaving only partial data and producing images that are blurred or distorted exactly where the diagnosis matters most. This raises a practical question: can a clear, faithful image be rebuilt from this incomplete and noisy data, while also telling doctors which parts of the reconstruction can be trusted?

To address this, the project developed PnP-AMP, a reconstruction method combining a mathematical algorithm that tracks the remaining noise at every step with a pretrained AI model that recognises realistic anatomy and cleans the image accordingly. The two combine to adjust automatically to the exact noise level at each step, with no manual tuning, and to produce a confidence map showing region by region how much the reconstruction can be trusted. This map is not arbitrary: the regions it flags as uncertain consistently match where the reconstruction differs most from the true image, mainly around tissue boundaries and areas affected by missing data, letting a doctor see at a glance which parts of the scan deserve a second look before a decision as consequential as choosing between surgery and chemoradiotherapy.

Tested on real MRI scans of cervical cancer patients, accelerated to a degree comparable to everyday hospital practice, the method improved image quality dramatically over classical approaches and, compared against every reconstruction strategy examined in this project, achieved both the best average accuracy and the most consistent performance across all test cases.

This work is more than a theoretical exercise. By improving MRI image quality without lengthening the scan, and by clearly flagging uncertain regions, it offers a practical tool for settings where radiologists are scarce relative to patient needs, such as much of sub-Saharan Africa, contributing directly to the fight against a disease that is largely preventable yet continues to claim lives precisely where access to quality diagnostic tools is hardest to find.