Artikel
New Computer-assisted Methods to Predict Wound Healing
Suche in Medline nach
Autoren
Veröffentlicht: | 21. April 2016 |
---|
Gliederung
Text
Background: Patients suffering from chronic wound healing disorders need around 17 consultations on average. The diagnostic and therapeutic costs add up to 16.000 € (18.000 US-$) for patients above 2 years of disease. Since incidence of diabetes mellitus, peripheral vascular and cancer diseases are rising, a higher amount of patient with chronic wound diseases are to be expected. In these patients, an accurate evaluation of the current wound healing is difficult. Diseases of the patient such as diabetes mellitus, renal insufficiency or peripheral vascular diseases, and current therapies such as chemotherapy or immunosuppression influence the wound healing in various ways, which leads to overtherapy by unnecessary visitations or wound products. To address this problem, we wanted to develop a prediction for wound healing in order to evaluate and adapt the current therapy.
Materials and methods: 7 postoperative wounds with a follow-up of 2 weeks were selected. Morphological features such as necrosis, bleeding, fibrin film and reddening due to infections were annotated for further analysis. Artefacts such as sutures were not considered. We developed an algorithm, which used a stochastic spatial Markov model, which was trained by the annotated pairs of wound pictures. The simulation calculated the probability of the wound morphology to change e.g. from granulated wound ground to fibrin film and vice versa. We used references points to equalize changes of photograph angle and lighting of the pictures. The simulation was performed 5 times. Afterwards, the relative most frequent morphological state of the wound was depicted and compared to the real follow-up photograph.
Results: 6 wound pictures could be successfully simulated. The algorithm predicted the possible expansion or decrease of mostly the fibrin film. First visual results show a good similarity (see Figure 1 [Fig. 1]). Due to methodological and computational issues, numerical results could not be calculated.
Conclusion: We successfully could predict the wound healing. However, pictures have to be taken in an almost identical angle and with same lightning. Accuracy will be increased by adding more patients and using patient data to create subgroups of risk profiles. This system may lead to a better prognosis of chronic wounds for an optimized treatment planning. Compliance may increase, since patients are more informed about when their wounds are probable to be close or suitable for secondary wound closure.
Acknowledgement: The current study was conducted within the setting of the Collaborative Research Center 125: Cognition Guided Surgery funded by the German Research Foundation (DFG).