The clinic at St. Luke's Hospital in Bradford, UK, has improved its patient intake by 30% following the implementation of the DERM AI system, now treating 32 patients per session instead of 24. Additionally, unnecessary biopsies have been reduced by about 10%, according to a report by BBC.

Developed by the British firm Skin Analytics, the DERM algorithm analyzes images of suspicious moles and other skin lesions to assess the likelihood of cancer.

A medical staff member captures three images of a suspicious area using a smartphone equipped with a dermatoscope attachment. The system processes these images in just a few minutes.

Source: BBC.

Patients diagnosed with benign conditions are given recommendations for care, while those suspected of having cancer are referred to a specialist on the same day.

Every diagnosis made by the algorithm is verified by a physician at St. Luke's Hospital.

“This has significantly affected all bottlenecks — from patient wait times to unnecessary tests and appointment delays,” stated plastic surgeon and clinical lead for skin cancer, Zakir Sharif.

The DERM system was launched at the clinic on April 30. Bradford Teaching Hospitals became the first NHS trust in West Yorkshire to adopt this technology, which is now in use at 40 different organizations.

According to the trust, approximately 5,000 referrals for suspected skin cancer are received annually in Bradford, with only about 8% ultimately diagnosed as malignant.

Evaluating DERM's Accuracy

Skin Analytics claims that DERM has a negative predictive value (NPV) of 99.9% for melanoma, meaning that 99.9% of cases identified as "non-melanoma" by the system are indeed not melanoma.

This metric assesses the reliability of excluding a diagnosis rather than the overall accuracy of the diagnosis.

The company reports that DERM's sensitivity to all types of skin cancer, reflecting the proportion of actual cases identified by the system, is 98%. These figures are derived from post-marketing surveillance.

An independent evaluation under the AI in Health and Care Award program indicated a sensitivity of 97% for melanoma.

Another study commissioned by NHS England, which analyzed 33,700 real cases, reported an NPV of 99.8% for melanoma using the system, compared to 98.9% for dermatologists during in-person consultations.

The UK's National Institute for Health and Care Excellence has conditionally recommended DERM for use in the NHS for three years, during which additional data on the system's clinical and economic effectiveness will be gathered.

In July, it was announced that two NHS trusts would begin using the PinPoint AI blood test for women suspected of having uterine cancer.

In a group of 3,313 participants in the study, the machine learning model demonstrated a 99% accuracy rate in identifying and ruling out gynecological cancers.

Follow ForkLog on social media

Telegram (main channel) Facebook X