NEURAL-NETWORK COMPUTER-PROGRAM TO DETERMINE PHOTOREFRACTIVE KERATECTOMY NOMOGRAMS

Purpose: To evaluate a commercially available neural network program f or calculation of photorefractive keratectomy treatment nomograms. Set ting: University referral refractive surgery clinic. Methods: PRK/LASI K Brain(TM), a commercial neural network computer program, was trained using the demographics, preoperalive clinical data, surgical paramete rs, and 1 year postoperative clinical data of 44 patients treated with a Summit Technology excimer laser using a 5.0 mm optical zone. The ne ural-network-derived nomogram was compared with the standard treatment nomogram for each patient The relative contribution of age, sex, kera tometry, and intraocular pressure. (IOP) to the predicted nomograms wa s also assessed. Results: Nomograms produced by the neural network wer e qualitatively similar to the standard nomogram.The sequence of data entry during training affected the network's predictions. Entry ordere d by Outcome (as opposed to entry by chronological order) yielded a no mogram that was more consistent with the 1 standard nomogram. However, both outcome- and chronologically ordered network-derived nomograms d iverged from the standard nomogram in individual Ir patients, includin g a subset for whom use of the standard nomogram yielded desired refra ctive results (within 0.25 diopter of emmetropia). Further analysis of the neural-network-derived nomograms revealed marked sensitivity to s ex, age, keratometry readings, and IOP. Conclusions: Neural networks o ffer a potential means of individualizing treatment nomograms to accou nt for patient demographics, preoperative examination, surgeon style, and equipment bias. However, a data set of 44 patients was not surgeon style, and equipment bias. However, a data set of 44 patients was not sufficient to train the PRK/IASIK Brain network to accurately predict treatment parameters in individual cases in the training set. A large r training set or a different learning algorithm may be required to im prove the neural network's performance.