The accuracy of radiological diagnosis of benign, primarily and secondarily malignant gastric ulcers and their correlation with three simplified radiological types.

Two radiologists individually gave 'blind' diagnoses of benignancy or malignancy to 145 radiographic representations of gastric ulcers, all with gastrectomy histological proof of benignancy or malignancy. A radiological diagnosis of malignancy was shown to be 98-6% reliable. A diagnosis of benignancy was less reliable (85-4% true positives) but this was a surgical series containing thereby an abnormally high proportion (over a third) of malignant ulcers. A correct diagnosis was mde in 99-5% of the proven benign ulcers, but in only 68-6% of the proven malignant ulcers. Analysis of the pathology reports on the malignant ulcers has shown that it was mainly those arising on the basis of pre-existing benign ulcers (i.e. secondarily malignant ulcers) that were wrongly diagnosed radiologically, thus suggesting that they have fundamentally different radiographic appearance from that of primarily malignant ulcers. It was then found that the ulcers could be divided into three simplified radiological types based on their in-profile appearances: type 1 (projecting), type 2 (intraluminal) and type 3 (so shallow as to be neither projecting nor intraluminal). Statistical association with the histological types led to the following conclusions: (a) radiological type 2 or 3 ulcers should be diagnosed as almost certainly malignant (usually primarily malignant) and (b) radiological type 1 ulcers are almost always benign but should be followed up until healed because the rare secondarily malignant ulcers are also usually of this type.